Subversion Repositories planix.SVN

Rev

Rev 2 | Details | Compare with Previous | Last modification | View Log | RSS feed

Rev Author Line No. Line
2 - 1 
<?xml version="1.0" encoding="utf-8"?>
2 
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
3 
"http://www.w3.org/TR/html4/loose.dtd">
4 
<html>
5 
<head>
6 
<meta http-equiv=Content-Type content="text/html; charset=utf8">
7 
<title>The Text Editor sam</title>
8 
</meta>
9 
</head>
10 
<body>
11 
<p style="margin-top: 0; margin-bottom: 0.50in"></p>
12 
<p style="margin-top: 0; margin-bottom: 0.21in"></p>
13 
 
14 
<p style="line-height: 1.4em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: center;">
15 
<span style="font-size: 12pt"><b>The Text Editor </b></span><span style="font-size: 12pt"><tt>sam</tt></span><span style="font-size: 12pt"><b></b></span></p>
16 
<p style="margin-top: 0; margin-bottom: 0.21in"></p>
17 
 
18 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
19 
<p style="line-height: 1.4em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: center;">
20 
<span style="font-size: 10pt"><i>Rob Pike</i></span></p>
21 
<p style="line-height: 1.4em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: center;">
22 
<span style="font-size: 10pt"><i>rob@plan9.bell-labs.com</i></span></p>
23 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
24 
 
25 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
26 
<p style="margin-top: 0; margin-bottom: 0.33in"></p>
27 
<p style="line-height: 1.4em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: center;">
28 
<span style="font-size: 10pt"><i>ABSTRACT</i></span></p>
29 
<p style="margin-top: 0; margin-bottom: 0.19in"></p>
30 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
31 
<p style="line-height: 1.2em; margin-left: 1.50in; text-indent: 0.50in; margin-right: 1.50in; margin-top: 0; margin-bottom: 0; text-align: justify;">
32 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
33 
is an interactive multi-file text editor intended for
34 
bitmap displays.
35 
A textual command language
36 
supplements the mouse-driven, cut-and-paste interface
37 
to make complex or
38 
repetitive editing tasks easy to specify.
39 
The language is characterized by the composition of regular expressions
40 
to describe the structure of the text being modified.
41 
The treatment of files as a database, with changes logged
42 
as atomic transactions, guides the implementation and
43 
makes a general &lsquo;undo&rsquo; mechanism straightforward.
44 
</span><span style="font-size: 10pt"></span><span style="font-size: 10pt"></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
45 
<p style="line-height: 1.2em; margin-left: 1.50in; text-indent: 0.35in; margin-right: 1.50in; margin-top: 0; margin-bottom: 0; text-align: justify;">
46 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
47 
is implemented as two processes connected by a low-bandwidth stream,
48 
one process handling the display and the other the editing
49 
algorithms.  Therefore it can run with the display process
50 
in a bitmap terminal and the editor on a local host,
51 
with both processes on a bitmap-equipped host, or with
52 
the display process in the terminal and the editor in a
53 
remote host.
54 
By suppressing the display process,
55 
it can even run without a bitmap terminal.
56 
</span><span style="font-size: 10pt"></span><span style="font-size: 10pt"></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
57 
<p style="line-height: 1.2em; margin-left: 1.50in; text-indent: 0.35in; margin-right: 1.50in; margin-top: 0; margin-bottom: 0; text-align: justify;">
58 
<span style="font-size: 10pt">This paper is reprinted from Software&mdash;Practice and Experience,
59 
Vol 17, number 11, pp. 813-845, November 1987.
60 
The paper has not been updated for the Plan 9 manuals.  Although
61 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
62 
has not changed much since the paper was written, the system around it certainly has.
63 
Nonetheless, the description here still stands as the best introduction to the editor.
64 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
65 
 
66 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
67 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
68 
<p style="margin-top: 0; margin-bottom: 0.50in"></p>
69 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
70 
<span style="font-size: 10pt"><b>Introduction
71 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
72 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
73 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
74 
is an interactive text editor that combines cut-and-paste interactive editing with
75 
an unusual command language based on the composition of regular expressions.
76 
It is written as two programs: one, the &lsquo;host part,&rsquo; runs on a UNIX system
77 
and implements the command language and provides file access; the other, the
78 
&lsquo;terminal part,&rsquo; runs asynchronously
79 
on a machine with a mouse and bitmap display
80 
and supports the display and interactive editing.
81 
The host part may be even run in isolation on an ordinary terminal
82 
to edit text using the command
83 
language, much like a traditional line editor,
84 
without assistance from a mouse or display.
85 
Most often,
86 
the terminal part runs on a Blit<sup></sup></span><sup><span style="font-size: 6pt">1</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt"> terminal
87 
(actually on a Teletype DMD 5620, the production version of the Blit), whose
88 
host connection is an ordinary 9600 bps RS232 link;
89 
on the SUN computer the host and display processes run on a single machine,
90 
connected by a pipe.
91 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
92 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
93 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
94 
edits uninterpreted
95 
ASCII text.
96 
It has no facilities for multiple fonts, graphics or tables,
97 
unlike MacWrite,<sup></sup></span><sup><span style="font-size: 6pt">2</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt"> Bravo,<sup></sup></span><sup><span style="font-size: 6pt">3</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt"> Tioga<sup></sup></span><sup><span style="font-size: 6pt">4</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
98 
or Lara.<sup></sup></span><sup><span style="font-size: 6pt">5</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
99 
Also unlike them, it has a rich command language.
100 
(Throughout this paper, the phrase
101 
</span><span style="font-size: 10pt"><i>command language
102 
</i></span><span style="font-size: 10pt">refers to
103 
textual commands; commands activated from the mouse form the
104 
</span><span style="font-size: 10pt"><i>mouse</i></span><span style="font-size: 10pt">
105 
</span><span style="font-size: 10pt"><i>language.</i></span><span style="font-size: 10pt">)
106 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
107 
developed as an editor for use by programmers, and tries to join
108 
the styles of the UNIX text editor
109 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt"><sup></sup></span><sup><span style="font-size: 6pt">6,7</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
110 
with that of interactive cut-and-paste editors by
111 
providing a comfortable mouse-driven interface
112 
to a program with a solid command language driven by regular expressions.
113 
The command language developed more than the mouse language, and
114 
acquired a notation for describing the structure of files
115 
more richly than as a sequence of lines,
116 
using a dataflow-like syntax for specifying changes.
117 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
118 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
119 
<span style="font-size: 10pt">The interactive style was influenced by
120 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">,<sup></sup></span><sup><span style="font-size: 6pt">1</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
121 
an early cut-and-paste editor for the Blit, and by
122 
</span><span style="font-size: 10pt"><tt>mux</tt></span><span style="font-size: 10pt">,<sup></sup></span><sup><span style="font-size: 6pt">8</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
123 
the Blit window system.
124 
</span><span style="font-size: 10pt"><tt>Mux</tt></span><span style="font-size: 10pt">
125 
merges the original Blit window system,
126 
</span><span style="font-size: 10pt"><tt>mpx</tt></span><span style="font-size: 10pt">,<sup></sup></span><sup><span style="font-size: 6pt">1</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
127 
with cut-and-paste editing, forming something like a
128 
multiplexed version of
129 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
130 
that edits the output of (and input to) command sessions rather than files.
131 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
132 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
133 
<span style="font-size: 10pt">The first part of this paper describes the command language, then the mouse
134 
language, and explains how they interact.
135 
That is followed by a description of the implementation,
136 
first of the host part, then of the terminal part.
137 
A principle that influenced the design of
138 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
139 
is that it should have no explicit limits, such as upper limits on
140 
file size or line length.
141 
A secondary consideration is that it be efficient.
142 
To honor these two goals together requires a method for efficiently
143 
manipulating
144 
huge strings (files) without breaking them into lines,
145 
perhaps while making thousands of changes
146 
under control of the command language.
147 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">&rsquo;s
148 
method is to
149 
treat the file as a transaction database, implementing changes as atomic
150 
updates.  These updates may be unwound easily to &lsquo;undo&rsquo; changes.
151 
Efficiency is achieved through a collection of caches that minimizes
152 
disc traffic and data motion, both within the two parts of the program
153 
and between them.
154 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
155 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
156 
<span style="font-size: 10pt">The terminal part of
157 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
158 
is fairly straightforward.
159 
More interesting is how the two halves of the editor stay
160 
synchronized when either half may initiate a change.
161 
This is achieved through a data structure that organizes the
162 
communications and is maintained in parallel by both halves.
163 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
164 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
165 
<span style="font-size: 10pt">The last part of the paper chronicles the writing of
166 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
167 
and discusses the lessons that were learned through its development and use.
168 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
169 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
170 
<span style="font-size: 10pt">The paper is long, but is composed largely of two papers of reasonable length:
171 
a description of the user interface of
172 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
173 
and a discussion of its implementation.
174 
They are combined because the implementation is strongly influenced by
175 
the user interface, and vice versa.
176 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
177 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
178 
<span style="font-size: 10pt"><b>The Interface
179 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
180 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
181 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
182 
is a text editor for multiple files.
183 
File names may be provided when it is invoked:
184 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
185 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
186 
<span style="font-size: 9pt"><tt>sam file1 file2 ...</tt></span></p>
187 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
188 
 
189 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
190 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
191 
<span style="font-size: 10pt">and there are commands
192 
to add new files and discard unneeded ones.
193 
Files are not read until necessary
194 
to complete some command.
195 
Editing operations apply to an internal copy
196 
made when the file is read; the UNIX file associated with the copy
197 
is changed only by an explicit command.
198 
To simplify the discussion, the internal copy is here called a
199 
</span><span style="font-size: 10pt"><i>file</i></span><span style="font-size: 10pt">,
200 
while the disc-resident original is called a
201 
</span><span style="font-size: 10pt"><i>disc file.
202 
</i></span><span style="font-size: 10pt"></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
203 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
204 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
205 
is usually connected to a bitmap display that presents a cut-and-paste
206 
editor driven by the mouse.
207 
In this mode, the command language is still available:
208 
text typed in a special window, called the
209 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
210 
</span><span style="font-size: 10pt"><i>window,</i></span><span style="font-size: 10pt">
211 
is interpreted
212 
as commands to be executed in the current file.
213 
Cut-and-paste editing may be used in any window &mdash; even in the
214 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
215 
window to construct commands.
216 
The other mode of operation, invoked by starting
217 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
218 
with the option
219 
</span><span style="font-size: 10pt"><tt>-d</tt></span><span style="font-size: 10pt">
220 
(for &lsquo;no download&rsquo;),
221 
does not use the mouse or bitmap display, but still permits
222 
editing using the textual command language, even on an ordinary terminal,
223 
interactively or from a script.
224 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
225 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
226 
<span style="font-size: 10pt">The following sections describe first the command language (under
227 
</span><span style="font-size: 10pt"><tt>sam\fP-d
228 
and in the
229 
</tt></span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt"><tt>
230 
window), and then the mouse interface.
231 
These two languages are nearly independent, but connect through the
232 
</tt></span><span style="font-size: 10pt"><i>current</i></span><span style="font-size: 10pt"><tt>
233 
</tt></span><span style="font-size: 10pt"><i>text,</i></span><span style="font-size: 10pt"><tt>
234 
described below.
235 
</tt></span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
236 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
237 
<span style="font-size: 10pt"><b>The Command Language
238 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
239 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
240 
<span style="font-size: 10pt">A file consists of its contents, which are an array of characters
241 
(that is, a string); the
242 
</span><span style="font-size: 10pt"><i>name</i></span><span style="font-size: 10pt">
243 
of the associated disc file; the
244 
</span><span style="font-size: 10pt"><i>modified bit
245 
</i></span><span style="font-size: 10pt">that states whether the contents match those of
246 
the disc file;
247 
and a substring of the contents, called the
248 
</span><span style="font-size: 10pt"><i>current text
249 
</i></span><span style="font-size: 10pt">or
250 
</span><span style="font-size: 10pt"><i>dot</i></span><span style="font-size: 10pt">
251 
(see Figures 1 and 2).
252 
If the current text is a null string, dot falls between characters.
253 
The
254 
</span><span style="font-size: 10pt"><i>value</i></span><span style="font-size: 10pt">
255 
of dot is the location of the current text; the
256 
</span><span style="font-size: 10pt"><i>contents</i></span><span style="font-size: 10pt">
257 
of dot are the characters it contains.
258 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
259 
imparts to the text no two-dimensional interpretation such as columns
260 
or fields; text is always one-dimensional.
261 
Even the idea of a &lsquo;line&rsquo; of text as understood by most UNIX programs
262 
&mdash; a sequence of characters terminated by a newline character &mdash;
263 
is only weakly supported.
264 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
265 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
266 
<span style="font-size: 10pt">The
267 
</span><span style="font-size: 10pt"><i>current file
268 
</i></span><span style="font-size: 10pt">is the file to which editing commands refer.
269 
The current text is therefore dot in the current file.
270 
If a command doesn&rsquo;t explicitly name a particular file or piece of text,
271 
the command is assumed to apply to the current text.
272 
For the moment, ignore the presence of multiple files and consider
273 
editing a single file.
274 
</span><span style="font-size: 10pt"></span></p><center><img src="fig1.gif" /></center>
275 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
276 
<span style="font-size: 8pt"><i>Figure 1. A typical
277 
</i></span><span style="font-size: 8pt"><tt>sam</tt></span><span style="font-size: 8pt"><i>
278 
screen, with the editing menu presented.
279 
The
280 
</i></span><span style="font-size: 8pt"><tt>sam</tt></span><span style="font-size: 8pt"><i>
281 
(command language) window is in the middle, with file windows above and below.
282 
(The user interface makes it easy to create these abutting windows.)
283 
The partially obscured window is a third file window.
284 
The uppermost window is that to which typing and mouse operations apply,
285 
as indicated by its heavy border.
286 
Each window has its current text highlighted in reverse video.
287 
The
288 
</i></span><span style="font-size: 8pt"><tt>sam</tt></span><span style="font-size: 8pt"><i>
289 
window&rsquo;s current text is the null string on the last visible line,
290 
indicated by a vertical bar.
291 
See also Figure 2.
292 
</i></span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
293 
<p style="margin-top: 0; margin-bottom: 0.02in"></p>
294 
 
295 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
296 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
297 
<span style="font-size: 10pt">Commands have one-letter names.
298 
Except for non-editing commands such as writing
299 
the file to disc, most commands make some change
300 
to the text in dot and leave dot set to the text resulting from the change.
301 
For example, the delete command,
302 
</span><span style="font-size: 10pt"><tt>d</tt></span><span style="font-size: 10pt">,
303 
deletes the text in dot, replacing it by the null string and setting dot
304 
to the result.
305 
The change command,
306 
</span><span style="font-size: 10pt"><tt>c</tt></span><span style="font-size: 10pt">,
307 
replaces dot by text delimited by an arbitrary punctuation character,
308 
conventionally
309 
a slash.  Thus,
310 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
311 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
312 
<span style="font-size: 9pt"><tt>c/Peter/</tt></span></p>
313 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
314 
 
315 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
316 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
317 
<span style="font-size: 10pt">replaces the text in dot by the string
318 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">.
319 
Similarly,
320 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
321 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
322 
<span style="font-size: 9pt"><tt>a/Peter/</tt></span></p>
323 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
324 
 
325 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
326 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
327 
<span style="font-size: 10pt">(append) adds the string after dot, and
328 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
329 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
330 
<span style="font-size: 9pt"><tt>i/Peter/</tt></span></p>
331 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
332 
 
333 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
334 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
335 
<span style="font-size: 10pt">(insert) inserts before dot.
336 
All three leave dot set to the new text,
337 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">.
338 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
339 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
340 
<span style="font-size: 10pt">Newlines are part of the syntax of commands:
341 
the newline character lexically terminates a command.
342 
Within the inserted text, however, newlines are never implicit.
343 
But since it is often convenient to insert multiple lines of text,
344 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
345 
has a special
346 
syntax for that case:
347 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
348 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
349 
<span style="font-size: 9pt"><tt>a</tt></span></p>
350 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
351 
<span style="font-size: 9pt"><tt>some lines of text</tt></span></p>
352 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
353 
<span style="font-size: 9pt"><tt>to be inserted in the file,</tt></span></p>
354 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
355 
<span style="font-size: 9pt"><tt>terminated by a period</tt></span></p>
356 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
357 
<span style="font-size: 9pt"><tt>on a line by itself</tt></span></p>
358 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
359 
<span style="font-size: 9pt"><tt>.</tt></span></p>
360 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
361 
 
362 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
363 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
364 
<span style="font-size: 10pt">In the one-line syntax, a newline character may be specified by a C-like
365 
escape, so
366 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
367 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
368 
<span style="font-size: 9pt"><tt>c/\n/</tt></span></p>
369 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
370 
 
371 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
372 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
373 
<span style="font-size: 10pt">replaces dot by a single newline character.
374 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
375 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
376 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
377 
also has a substitute command,
378 
</span><span style="font-size: 10pt"><tt>s</tt></span><span style="font-size: 10pt">:
379 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
380 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
381 
<span style="font-size: 9pt"><tt>s/</tt></span><span style="font-size: 9pt"><i>expression</i></span><span style="font-size: 9pt"><tt>/</tt></span><span style="font-size: 9pt"><i>replacement</i></span><span style="font-size: 9pt"><tt>/</tt></span></p>
382 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
383 
 
384 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
385 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
386 
<span style="font-size: 10pt">substitutes the replacement text for the first match, in dot,
387 
of the regular expression.
388 
Thus, if dot is the string
389 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">,
390 
the command
391 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
392 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
393 
<span style="font-size: 9pt"><tt>s/t/st/</tt></span></p>
394 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
395 
 
396 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
397 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
398 
<span style="font-size: 10pt">changes it to
399 
</span><span style="font-size: 10pt"><tt>Pester</tt></span><span style="font-size: 10pt">.
400 
In general,
401 
</span><span style="font-size: 10pt"><tt>s</tt></span><span style="font-size: 10pt">
402 
is unnecessary, but it was inherited from
403 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">
404 
and it has some convenient variations.
405 
For instance, the replacement text may include the matched text,
406 
specified by
407 
</span><span style="font-size: 10pt"><tt>&amp;</tt></span><span style="font-size: 10pt">:
408 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
409 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
410 
<span style="font-size: 9pt"><tt>s/Peter/Oh, &amp;, &amp;, &amp;, &amp;!/</tt></span></p>
411 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
412 
 
413 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
414 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
415 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
416 
<span style="font-size: 10pt">There are also three commands that apply programs
417 
to text:
418 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
419 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
420 
<span style="font-size: 9pt"><tt>&lt; </tt></span><span style="font-size: 9pt"><i>UNIX program</i></span><span style="font-size: 9pt"><tt></tt></span></p>
421 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
422 
 
423 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
424 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
425 
<span style="font-size: 10pt">replaces dot by the output of the UNIX program.
426 
Similarly, the
427 
</span><span style="font-size: 10pt"><tt>&gt;</tt></span><span style="font-size: 10pt">
428 
command
429 
runs the program with dot as its standard input, and
430 
</span><span style="font-size: 10pt"><tt>|</tt></span><span style="font-size: 10pt">
431 
does both.  For example,
432 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
433 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
434 
<span style="font-size: 9pt"><tt>| sort</tt></span></p>
435 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
436 
 
437 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
438 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
439 
<span style="font-size: 10pt">replaces dot by the result of applying the standard sorting utility to it.
440 
Again, newlines have no special significance for these
441 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
442 
commands.
443 
The text acted upon and resulting from these commands is not necessarily
444 
bounded by newlines, although for connection with UNIX programs,
445 
newlines may be necessary to obey conventions.
446 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
447 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
448 
<span style="font-size: 10pt">One more command:
449 
</span><span style="font-size: 10pt"><tt>p</tt></span><span style="font-size: 10pt">
450 
prints the contents of dot.
451 
Table I summarizes
452 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
453 
commands.
454 
</span><span style="font-size: 10pt"></span></p><center><img src="sam0.png"></center>
455 
</center>
456 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
457 
<p style="margin-top: 0; margin-bottom: 0.02in"></p>
458 
 
459 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
460 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
461 
<span style="font-size: 10pt">The value of dot may be changed by
462 
specifying an
463 
</span><span style="font-size: 10pt"><i>address</i></span><span style="font-size: 10pt">
464 
for the command.
465 
The simplest address is a line number:
466 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
467 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
468 
<span style="font-size: 9pt"><tt>3</tt></span></p>
469 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
470 
 
471 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
472 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
473 
<span style="font-size: 10pt">refers to the third line of the file, so
474 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
475 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
476 
<span style="font-size: 9pt"><tt>3d</tt></span></p>
477 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
478 
 
479 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
480 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
481 
<span style="font-size: 10pt">deletes the third line of the file, and implicitly renumbers
482 
the lines so the old line 4 is now numbered 3.
483 
(This is one of the few places where
484 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
485 
deals with lines directly.)
486 
Line
487 
</span><span style="font-size: 10pt"><tt>0</tt></span><span style="font-size: 10pt">
488 
is the null string at the beginning of the file.
489 
If a command consists of only an address, a
490 
</span><span style="font-size: 10pt"><tt>p</tt></span><span style="font-size: 10pt">
491 
command is assumed, so typing an unadorned
492 
</span><span style="font-size: 10pt"><tt>3</tt></span><span style="font-size: 10pt">
493 
prints line 3 on the terminal.
494 
There are a couple of other basic addresses:
495 
a period addresses dot itself; and
496 
a dollar sign
497 
(</span><span style="font-size: 10pt"><tt>$</tt></span><span style="font-size: 10pt">)
498 
addresses the null string at the end of the file.
499 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
500 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
501 
<span style="font-size: 10pt">An address is always a single substring of the file.
502 
Thus, the address
503 
</span><span style="font-size: 10pt"><tt>3</tt></span><span style="font-size: 10pt">
504 
addresses the characters
505 
after the second newline of
506 
the file through the third newline of the file.
507 
A
508 
</span><span style="font-size: 10pt"><i>compound address
509 
</i></span><span style="font-size: 10pt">is constructed by the comma operator
510 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
511 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
512 
<span style="font-size: 9pt"><i>address1</i></span><span style="font-size: 9pt"><tt>,</tt></span><span style="font-size: 9pt"><i>address2</i></span><span style="font-size: 9pt"><tt></tt></span></p>
513 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
514 
 
515 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
516 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
517 
<span style="font-size: 10pt">and addresses the substring of the file from the beginning of
518 
</span><span style="font-size: 10pt"><i>address1</i></span><span style="font-size: 10pt">
519 
to the end of
520 
</span><span style="font-size: 10pt"><i>address2</i></span><span style="font-size: 10pt">.
521 
For example, the command
522 
</span><span style="font-size: 10pt"><tt>3,5p</tt></span><span style="font-size: 10pt">
523 
prints the third through fifth lines of the file and
524 
</span><span style="font-size: 10pt"><tt>.,$d</tt></span><span style="font-size: 10pt">
525 
deletes the text from the beginning of dot to the end of the file.
526 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
527 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
528 
<span style="font-size: 10pt">These addresses are all absolute positions in the file, but
529 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
530 
also has relative addresses, indicated by
531 
</span><span style="font-size: 10pt"><tt>+</tt></span><span style="font-size: 10pt">
532 
or
533 
</span><span style="font-size: 10pt"><tt>-</tt></span><span style="font-size: 10pt">.
534 
For example,
535 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
536 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
537 
<span style="font-size: 9pt"><tt>$-3</tt></span></p>
538 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
539 
 
540 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
541 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
542 
<span style="font-size: 10pt">is the third line before the end of the file and
543 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
544 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
545 
<span style="font-size: 9pt"><tt>.+1</tt></span></p>
546 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
547 
 
548 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
549 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
550 
<span style="font-size: 10pt">is the line after dot.
551 
If no address appears to the left of the
552 
</span><span style="font-size: 10pt"><tt>+</tt></span><span style="font-size: 10pt">
553 
or
554 
</span><span style="font-size: 10pt"><tt>-</tt></span><span style="font-size: 10pt">,
555 
dot is assumed;
556 
if nothing appears to the right,
557 
</span><span style="font-size: 10pt"><tt>1</tt></span><span style="font-size: 10pt">
558 
is assumed.
559 
Therefore,
560 
</span><span style="font-size: 10pt"><tt>.+1</tt></span><span style="font-size: 10pt">
561 
may be abbreviated to just a plus sign.
562 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
563 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
564 
<span style="font-size: 10pt">The
565 
</span><span style="font-size: 10pt"><tt>+</tt></span><span style="font-size: 10pt">
566 
operator acts relative to the end of its first argument, while the
567 
</span><span style="font-size: 10pt"><tt>-</tt></span><span style="font-size: 10pt">
568 
operator acts relative to the beginning.  Thus
569 
</span><span style="font-size: 10pt"><tt>.+1</tt></span><span style="font-size: 10pt">
570 
addresses the first line after dot,
571 
</span><span style="font-size: 10pt"><tt>.-</tt></span><span style="font-size: 10pt">
572 
addresses the first line before dot, and
573 
</span><span style="font-size: 10pt"><tt>+-</tt></span><span style="font-size: 10pt">
574 
refers to the line containing the end of dot.  (Dot may span multiple lines, and
575 
</span><span style="font-size: 10pt"><tt>+</tt></span><span style="font-size: 10pt">
576 
selects the line after the end of dot, then
577 
</span><span style="font-size: 10pt"><tt>-</tt></span><span style="font-size: 10pt">
578 
backs up one line.)
579 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
580 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
581 
<span style="font-size: 10pt">The final type of address is a regular expression, which addresses the
582 
text matched by the expression.  The expression is enclosed in slashes, as in
583 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
584 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
585 
<span style="font-size: 9pt"><tt>/</tt></span><span style="font-size: 9pt"><i>expression</i></span><span style="font-size: 9pt"><tt>/</tt></span></p>
586 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
587 
 
588 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
589 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
590 
<span style="font-size: 10pt">The expressions are the same as those in the UNIX program
591 
</span><span style="font-size: 10pt"><tt>egrep</tt></span><span style="font-size: 10pt">,<sup></sup></span><sup><span style="font-size: 6pt">6,7</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
592 
and include closures, alternations, and so on.
593 
They find the
594 
</span><span style="font-size: 10pt"><i>leftmost longest
595 
</i></span><span style="font-size: 10pt">string that matches the expression, that is,
596 
the first match after the point where the search is started,
597 
and if more than one match begins at the same spot, the longest such match.
598 
(I assume familiarity with the syntax for regular expressions in UNIX programs.<sup></sup></span><sup><span style="font-size: 6pt">9</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">)
599 
For example,
600 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
601 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
602 
<span style="font-size: 9pt"><tt>/x/</tt></span></p>
603 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
604 
 
605 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
606 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
607 
<span style="font-size: 10pt">matches the next
608 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
609 
character in the file,
610 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
611 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
612 
<span style="font-size: 9pt"><tt>/xx*/</tt></span></p>
613 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
614 
 
615 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
616 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
617 
<span style="font-size: 10pt">matches the next run of one or more
618 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">&rsquo;s,
619 
and
620 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
621 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
622 
<span style="font-size: 9pt"><tt>/x|Peter/</tt></span></p>
623 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
624 
 
625 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
626 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
627 
<span style="font-size: 10pt">matches the next
628 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
629 
or
630 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">.
631 
For compatibility with other UNIX programs, the &lsquo;any character&rsquo; operator,
632 
a period,
633 
does not match a newline, so
634 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
635 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
636 
<span style="font-size: 9pt"><tt>/.*/</tt></span></p>
637 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
638 
 
639 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
640 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
641 
<span style="font-size: 10pt">matches the text from dot to the end of the line, but excludes the newline
642 
and so will not match across
643 
the line boundary.
644 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
645 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
646 
<span style="font-size: 10pt">Regular expressions are always relative addresses.
647 
The direction is forwards by default,
648 
so
649 
</span><span style="font-size: 10pt"><tt>/Peter/</tt></span><span style="font-size: 10pt">
650 
is really an abbreviation for
651 
</span><span style="font-size: 10pt"><tt>+/Peter/</tt></span><span style="font-size: 10pt">.
652 
The search can be reversed with a minus sign, so
653 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
654 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
655 
<span style="font-size: 9pt"><tt></tt></span><span style="font-size: 9pt"><tt>-/Peter/</tt></span><span style="font-size: 9pt"><tt></tt></span></p>
656 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
657 
 
658 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
659 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
660 
<span style="font-size: 10pt">finds the first
661 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">
662 
before dot.
663 
Regular expressions may be used with other address forms, so
664 
</span><span style="font-size: 10pt"><tt>0+/Peter/</tt></span><span style="font-size: 10pt">
665 
finds the first
666 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">
667 
in the file and
668 
</span><span style="font-size: 10pt"><tt>$-/Peter/</tt></span><span style="font-size: 10pt">
669 
finds the last.
670 
Table II summarizes
671 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
672 
addresses.
673 
</span><span style="font-size: 10pt"></span></p><center><img src="sam1.png"></center>
674 
</center>
675 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
676 
<p style="margin-top: 0; margin-bottom: 0.02in"></p>
677 
 
678 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
679 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
680 
<span style="font-size: 10pt">The language discussed so far will not seem novel
681 
to people who use UNIX text editors
682 
such as
683 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">
684 
or
685 
</span><span style="font-size: 10pt"><tt>vi</tt></span><span style="font-size: 10pt">.<sup></sup></span><sup><span style="font-size: 6pt">9</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
686 
Moreover, the kinds of editing operations these commands allow, with the exception
687 
of regular expressions and line numbers,
688 
are clearly more conveniently handled by a mouse-based interface.
689 
Indeed,
690 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
691 
mouse language (discussed at length below) is the means by which
692 
simple changes are usually made.
693 
For large or repetitive changes, however, a textual language
694 
outperforms a manual interface.
695 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
696 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
697 
<span style="font-size: 10pt">Imagine that, instead of deleting just one occurrence of the string
698 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">,
699 
we wanted to eliminate every
700 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">.
701 
What&rsquo;s needed is an iterator that runs a command for each occurrence of some
702 
text.
703 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">&rsquo;s
704 
iterator is called
705 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">,
706 
for extract:
707 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
708 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
709 
<span style="font-size: 9pt"><tt>x/</tt></span><span style="font-size: 9pt"><i>expression</i></span><span style="font-size: 9pt"><tt>/ </tt></span><span style="font-size: 9pt"><i>command</i></span><span style="font-size: 9pt"><tt></tt></span></p>
710 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
711 
 
712 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
713 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
714 
<span style="font-size: 10pt">finds all matches in dot of the specified expression, and for each
715 
such match, sets dot to the text matched and runs the command.
716 
So to delete all the
717 
</span><span style="font-size: 10pt"><tt>Peters:</tt></span><span style="font-size: 10pt">
718 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
719 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
720 
<span style="font-size: 9pt"><tt>0,$ x/Peter/ d</tt></span></p>
721 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
722 
 
723 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
724 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
725 
<span style="font-size: 10pt">(Blanks in these examples are to improve readability;
726 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
727 
neither requires nor interprets them.)
728 
This searches the entire file
729 
(</span><span style="font-size: 10pt"><tt>0,$</tt></span><span style="font-size: 10pt">)
730 
for occurrences of the string
731 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">,
732 
and runs the
733 
</span><span style="font-size: 10pt"><tt>d</tt></span><span style="font-size: 10pt">
734 
command with dot set to each such occurrence.
735 
(By contrast, the comparable
736 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">
737 
command would delete all
738 
</span><span style="font-size: 10pt"><i>lines</i></span><span style="font-size: 10pt">
739 
containing
740 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">;
741 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
742 
deletes only the
743 
</span><span style="font-size: 10pt"><tt>Peters</tt></span><span style="font-size: 10pt">.)
744 
The address
745 
</span><span style="font-size: 10pt"><tt>0,$</tt></span><span style="font-size: 10pt">
746 
is commonly used, and may be abbreviated to just a comma.
747 
As another example,
748 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
749 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
750 
<span style="font-size: 9pt"><tt>, x/Peter/ p</tt></span></p>
751 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
752 
 
753 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
754 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
755 
<span style="font-size: 10pt">prints a list of
756 
</span><span style="font-size: 10pt"><tt>Peters,</tt></span><span style="font-size: 10pt">
757 
one for each appearance in the file, with no intervening text (not even newlines
758 
to separate the instances).
759 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
760 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
761 
<span style="font-size: 10pt">Of course, the text extracted by
762 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
763 
may be selected by a regular expression,
764 
which complicates deciding what set of matches is chosen &mdash;
765 
matches may overlap.  This is resolved by generating the matches
766 
starting from the beginning of dot using the leftmost-longest rule,
767 
and searching for each match starting from the end of the previous one.
768 
Regular expressions may also match null strings, but a null match
769 
adjacent to a non-null match is never selected; at least one character
770 
must intervene.
771 
For example,
772 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
773 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
774 
<span style="font-size: 9pt"><tt>, c/AAA/</tt></span></p>
775 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
776 
<span style="font-size: 9pt"><tt>x/B*/ c/-/</tt></span></p>
777 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
778 
<span style="font-size: 9pt"><tt>, p</tt></span></p>
779 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
780 
 
781 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
782 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
783 
<span style="font-size: 10pt">produces as output
784 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
785 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
786 
<span style="font-size: 9pt"><tt>-A-A-A-</tt></span></p>
787 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
788 
 
789 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
790 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
791 
<span style="font-size: 10pt">because the pattern
792 
</span><span style="font-size: 10pt"><tt>B*</tt></span><span style="font-size: 10pt">
793 
matches the null strings separating the
794 
</span><span style="font-size: 10pt"><tt>A</tt></span><span style="font-size: 10pt">&rsquo;s.
795 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
796 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
797 
<span style="font-size: 10pt">The
798 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
799 
command has a complement,
800 
</span><span style="font-size: 10pt"><tt>y</tt></span><span style="font-size: 10pt">,
801 
with similar syntax, that executes the command with dot set to the text
802 
</span><span style="font-size: 10pt"><i>between</i></span><span style="font-size: 10pt">
803 
the matches of the expression.
804 
For example,
805 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
806 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
807 
<span style="font-size: 9pt"><tt>, c/AAA/</tt></span></p>
808 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
809 
<span style="font-size: 9pt"><tt>y/A/ c/-/</tt></span></p>
810 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
811 
<span style="font-size: 9pt"><tt>, p</tt></span></p>
812 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
813 
 
814 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
815 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
816 
<span style="font-size: 10pt">produces the same result as the example above.
817 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
818 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
819 
<span style="font-size: 10pt">The
820 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
821 
and
822 
</span><span style="font-size: 10pt"><tt>y</tt></span><span style="font-size: 10pt">
823 
commands are looping constructs, and
824 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
825 
has a pair of conditional commands to go with them.
826 
They have similar syntax:
827 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
828 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
829 
<span style="font-size: 9pt"><tt>g/</tt></span><span style="font-size: 9pt"><i>expression</i></span><span style="font-size: 9pt"><tt>/ </tt></span><span style="font-size: 9pt"><i>command</i></span><span style="font-size: 9pt"><tt></tt></span></p>
830 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
831 
 
832 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
833 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
834 
<span style="font-size: 10pt">(guard)
835 
runs the command exactly once if dot contains a match of the expression.
836 
This is different from
837 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">,
838 
which runs the command for
839 
</span><span style="font-size: 10pt"><i>each</i></span><span style="font-size: 10pt">
840 
match:
841 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
842 
loops;
843 
</span><span style="font-size: 10pt"><tt>g</tt></span><span style="font-size: 10pt">
844 
merely tests, without changing the value of dot.
845 
Thus,
846 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
847 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
848 
<span style="font-size: 9pt"><tt>, x/Peter/ d</tt></span></p>
849 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
850 
 
851 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
852 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
853 
<span style="font-size: 10pt">deletes all occurrences of
854 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">,
855 
but
856 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
857 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
858 
<span style="font-size: 9pt"><tt>, g/Peter/ d</tt></span></p>
859 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
860 
 
861 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
862 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
863 
<span style="font-size: 10pt">deletes the whole file (reduces it to a null string) if
864 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">
865 
occurs anywhere in the text.
866 
The complementary conditional is
867 
</span><span style="font-size: 10pt"><tt>v</tt></span><span style="font-size: 10pt">,
868 
which runs the command if there is
869 
</span><span style="font-size: 10pt"><i>no</i></span><span style="font-size: 10pt">
870 
match of the expression.
871 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
872 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
873 
<span style="font-size: 10pt">These control-structure-like commands may be composed to construct more
874 
involved operations.  For example, to print those lines of text that
875 
contain the string
876 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">:
877 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
878 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
879 
<span style="font-size: 9pt"><tt>, x/.*\n/ g/Peter/ p</tt></span></p>
880 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
881 
 
882 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
883 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
884 
<span style="font-size: 10pt">The
885 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
886 
breaks the file into lines, the
887 
</span><span style="font-size: 10pt"><tt>g</tt></span><span style="font-size: 10pt">
888 
selects those lines containing
889 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">,
890 
and the
891 
</span><span style="font-size: 10pt"><tt>p</tt></span><span style="font-size: 10pt">
892 
prints them.
893 
This command gives an address for the
894 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
895 
command (the whole file), but because
896 
</span><span style="font-size: 10pt"><tt>g</tt></span><span style="font-size: 10pt">
897 
does not have an explicit address, it applies to the value of
898 
dot produced by the
899 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
900 
command, that is, to each line.
901 
All commands in
902 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
903 
except for the command to write a file to disc use dot for the
904 
default address.
905 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
906 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
907 
<span style="font-size: 10pt">Composition may be continued indefinitely.
908 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
909 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
910 
<span style="font-size: 9pt"><tt>, x/.*\n/ g/Peter/ v/SaltPeter/ p</tt></span></p>
911 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
912 
 
913 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
914 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
915 
<span style="font-size: 10pt">prints those lines containing
916 
</span><span style="font-size: 10pt"><tt>Peter</tt></span><span style="font-size: 10pt">
917 
but
918 
</span><span style="font-size: 10pt"><i>not</i></span><span style="font-size: 10pt">
919 
those containing
920 
</span><span style="font-size: 10pt"><tt>SaltPeter</tt></span><span style="font-size: 10pt">.
921 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
922 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
923 
<span style="font-size: 10pt"><b>Structural Regular Expressions
924 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
925 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
926 
<span style="font-size: 10pt">Unlike other UNIX text editors,
927 
including the non-interactive ones such as
928 
</span><span style="font-size: 10pt"><tt>sed</tt></span><span style="font-size: 10pt">
929 
and
930 
</span><span style="font-size: 10pt"><tt>awk</tt></span><span style="font-size: 10pt">,<sup></sup></span><sup><span style="font-size: 6pt">7</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
931 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
932 
is good for manipulating files with multi-line &lsquo;records.&rsquo;
933 
An example is an on-line phone book composed of records,
934 
separated by blank lines, of the form
935 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
936 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
937 
<span style="font-size: 9pt"><tt>Herbert Tic</tt></span></p>
938 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
939 
<span style="font-size: 9pt"><tt>44 Turnip Ave., Endive, NJ</tt></span></p>
940 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
941 
<span style="font-size: 9pt"><tt>201-5555642</tt></span></p>
942 
<p style="margin-top: 0; margin-bottom: 0.15in"></p>
943 
 
944 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
945 
<span style="font-size: 9pt"><tt>Norbert Twinge</tt></span></p>
946 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
947 
<span style="font-size: 9pt"><tt>16 Potato St., Cabbagetown, NJ</tt></span></p>
948 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
949 
<span style="font-size: 9pt"><tt>201-5553145</tt></span></p>
950 
<p style="margin-top: 0; margin-bottom: 0.15in"></p>
951 
 
952 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
953 
<span style="font-size: 9pt"><tt>...</tt></span></p>
954 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
955 
 
956 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
957 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
958 
<span style="font-size: 10pt">The format may be encoded as a regular expression:
959 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
960 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
961 
<span style="font-size: 9pt"><tt>(.+\n)+</tt></span></p>
962 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
963 
 
964 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
965 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
966 
<span style="font-size: 10pt">that is, a sequence of one or more non-blank lines.
967 
The command to print Mr. Tic&rsquo;s entire record is then
968 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
969 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
970 
<span style="font-size: 9pt"><tt>, x/(.+\n)+/ g/^Herbert Tic$/ p</tt></span></p>
971 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
972 
 
973 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
974 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
975 
<span style="font-size: 10pt">and that to extract just the phone number is
976 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
977 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
978 
<span style="font-size: 9pt"><tt>, x/(.+\n)+/ g/^Herbert Tic$/ x/^[0-9]*-[0-9]*\n/ p</tt></span></p>
979 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
980 
 
981 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
982 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
983 
<span style="font-size: 10pt">The latter command breaks the file into records,
984 
chooses Mr. Tic&rsquo;s record,
985 
extracts the phone number from the record,
986 
and finally prints the number.
987 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
988 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
989 
<span style="font-size: 10pt">A more involved problem is that of
990 
renaming a particular variable, say
991 
</span><span style="font-size: 10pt"><tt>n</tt></span><span style="font-size: 10pt">,
992 
to
993 
</span><span style="font-size: 10pt"><tt>num</tt></span><span style="font-size: 10pt">
994 
in a C program.
995 
The obvious first attempt,
996 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
997 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
998 
<span style="font-size: 9pt"><tt>, x/n/ c/num/</tt></span></p>
999 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1000 
 
1001 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1002 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1003 
<span style="font-size: 10pt">is badly flawed: it changes not only the variable
1004 
</span><span style="font-size: 10pt"><tt>n</tt></span><span style="font-size: 10pt">
1005 
but any letter
1006 
</span><span style="font-size: 10pt"><tt>n</tt></span><span style="font-size: 10pt">
1007 
that appears.
1008 
We need to extract all the variables, and select those that match
1009 
</span><span style="font-size: 10pt"><tt>n</tt></span><span style="font-size: 10pt">
1010 
and only
1011 
</span><span style="font-size: 10pt"><tt>n</tt></span><span style="font-size: 10pt">:
1012 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1013 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1014 
<span style="font-size: 9pt"><tt>, x/[A-Za-z_][A-Za-z_0-9]*/ g/n/ v/../ c/num/</tt></span></p>
1015 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1016 
 
1017 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1018 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1019 
<span style="font-size: 10pt">The pattern
1020 
</span><span style="font-size: 10pt"><tt>[A-Za-z_][A-Za-z_0-9]*</tt></span><span style="font-size: 10pt">
1021 
matches C identifiers.
1022 
Next
1023 
</span><span style="font-size: 10pt"><tt>g/n/</tt></span><span style="font-size: 10pt">
1024 
selects those containing an
1025 
</span><span style="font-size: 10pt"><tt>n</tt></span><span style="font-size: 10pt">.
1026 
Then
1027 
</span><span style="font-size: 10pt"><tt>v/../</tt></span><span style="font-size: 10pt">
1028 
rejects those containing two (or more) characters, and finally
1029 
</span><span style="font-size: 10pt"><tt>c/num/</tt></span><span style="font-size: 10pt">
1030 
changes the remainder (identifiers
1031 
</span><span style="font-size: 10pt"><tt>n</tt></span><span style="font-size: 10pt">)
1032 
to
1033 
</span><span style="font-size: 10pt"><tt>num</tt></span><span style="font-size: 10pt">.
1034 
This version clearly works much better, but there may still be problems.
1035 
For example, in C character and string constants, the sequence
1036 
</span><span style="font-size: 10pt"><tt>\n</tt></span><span style="font-size: 10pt">
1037 
is interpreted as a newline character, and we don&rsquo;t want to change it to
1038 
</span><span style="font-size: 10pt"><tt>\num.</tt></span><span style="font-size: 10pt">
1039 
This problem can be forestalled with a
1040 
</span><span style="font-size: 10pt"><tt>y</tt></span><span style="font-size: 10pt">
1041 
command:
1042 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1043 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1044 
<span style="font-size: 9pt"><tt>, y/\\n/ x/[A-Za-z_][A-Za-z_0-9]*/ g/n/ v/../ c/num/</tt></span></p>
1045 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1046 
 
1047 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1048 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1049 
<span style="font-size: 10pt">(the second
1050 
</span><span style="font-size: 10pt"><tt>\</tt></span><span style="font-size: 10pt">
1051 
is necessary because of lexical conventions in regular expressions),
1052 
or we could even reject character constants and strings outright:
1053 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1054 
<p style="line-height: 1.1em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1055 
<span style="font-size: 9pt"><tt>,y/&rsquo;[^&rsquo;]*&rsquo;/ y/"[^"]*"/ x/[A-Za-z_][A-Za-z_0-9]*/ g/n/ v/../ c/num/</tt></span></p>
1056 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1057 
 
1058 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1059 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1060 
<span style="font-size: 10pt">The
1061 
</span><span style="font-size: 10pt"><tt>y</tt></span><span style="font-size: 10pt">
1062 
commands in this version exclude from consideration all character constants
1063 
and strings.
1064 
The only remaining problem is to deal with the possible occurrence of
1065 
</span><span style="font-size: 10pt"><tt>\&rsquo;</tt></span><span style="font-size: 10pt">
1066 
or
1067 
</span><span style="font-size: 10pt"><tt>\"</tt></span><span style="font-size: 10pt">
1068 
within these sequences, but it&rsquo;s easy to see how to resolve this difficulty.
1069 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1070 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1071 
<span style="font-size: 10pt">The point of these composed commands is successive refinement.
1072 
A simple version of the command is tried, and if it&rsquo;s not good enough,
1073 
it can be honed by adding a clause or two.
1074 
(Mistakes can be undone; see below.
1075 
Also, the mouse language makes it unnecessary to retype the command each time.)
1076 
The resulting chains of commands are somewhat reminiscent of
1077 
shell pipelines.<sup></sup></span><sup><span style="font-size: 6pt">7</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
1078 
Unlike pipelines, though, which pass along modified
1079 
</span><span style="font-size: 10pt"><i>data</i></span><span style="font-size: 10pt">,
1080 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1081 
commands pass a
1082 
</span><span style="font-size: 10pt"><i>view</i></span><span style="font-size: 10pt">
1083 
of the data.
1084 
The text at each step of the command is the same, but which pieces
1085 
are selected is refined step by step until the correct piece is
1086 
available to the final step of the command line, which ultimately makes the change.
1087 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1088 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1089 
<span style="font-size: 10pt">In other UNIX programs, regular expressions are used only for selection,
1090 
as in the
1091 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1092 
</span><span style="font-size: 10pt"><tt>g</tt></span><span style="font-size: 10pt">
1093 
command, never for extraction as in the
1094 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
1095 
or
1096 
</span><span style="font-size: 10pt"><tt>y</tt></span><span style="font-size: 10pt">
1097 
command.
1098 
For example, patterns in
1099 
</span><span style="font-size: 10pt"><tt>awk</tt></span><span style="font-size: 10pt"><sup></sup></span><sup><span style="font-size: 6pt">7</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
1100 
are used to select lines to be operated on, but cannot be used
1101 
to describe the format of the input text, or to handle newline-free text.
1102 
The use of regular expressions to describe the structure of a piece
1103 
of text rather than its contents, as in the
1104 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
1105 
command,
1106 
has been given a name:
1107 
</span><span style="font-size: 10pt"><i>structural regular expressions.
1108 
</i></span><span style="font-size: 10pt">When they are composed, as in the above example,
1109 
they are pleasantly expressive.
1110 
Their use is discussed at greater length elsewhere.<sup></sup></span><sup><span style="font-size: 6pt">10</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
1111 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1112 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1113 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1114 
<span style="font-size: 10pt"><b>Multiple files
1115 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1116 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1117 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
1118 
has a few other commands, mostly relating to input and output.
1119 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1120 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1121 
<span style="font-size: 9pt"><tt>e discfilename</tt></span></p>
1122 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1123 
 
1124 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1125 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1126 
<span style="font-size: 10pt">replaces the contents and name of the current file with those of the named
1127 
disc file;
1128 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1129 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1130 
<span style="font-size: 9pt"><tt>w discfilename</tt></span></p>
1131 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1132 
 
1133 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1134 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1135 
<span style="font-size: 10pt">writes the contents to the named disc file; and
1136 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1137 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1138 
<span style="font-size: 9pt"><tt>r discfilename</tt></span></p>
1139 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1140 
 
1141 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1142 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1143 
<span style="font-size: 10pt">replaces dot with the contents of the named disc file.
1144 
All these commands use the current file&rsquo;s name if none is specified.
1145 
Finally,
1146 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1147 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1148 
<span style="font-size: 9pt"><tt>f discfilename</tt></span></p>
1149 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1150 
 
1151 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1152 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1153 
<span style="font-size: 10pt">changes the name associated with the file and displays the result:
1154 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1155 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1156 
<span style="font-size: 9pt"><tt>&rsquo;-. discfilename</tt></span></p>
1157 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1158 
 
1159 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1160 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1161 
<span style="font-size: 10pt">This output is called the file&rsquo;s
1162 
</span><span style="font-size: 10pt"><i>menu line,
1163 
</i></span><span style="font-size: 10pt">because it is the contents of the file&rsquo;s line in the button 3 menu (described
1164 
in the
1165 
next section).
1166 
The first three characters are a concise notation for the state of the file.
1167 
The apostrophe signifies that the file is modified.
1168 
The minus sign indicates the number of windows
1169 
open on the file (see the next section):
1170 
</span><span style="font-size: 10pt"><tt>-</tt></span><span style="font-size: 10pt">
1171 
means none,
1172 
</span><span style="font-size: 10pt"><tt>+</tt></span><span style="font-size: 10pt">
1173 
means one, and
1174 
</span><span style="font-size: 10pt"><tt>*</tt></span><span style="font-size: 10pt">
1175 
means more than one.
1176 
Finally, the period indicates that this is the current file.
1177 
These characters are useful for controlling the
1178 
</span><span style="font-size: 10pt"><tt>X</tt></span><span style="font-size: 10pt">
1179 
command, described shortly.
1180 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1181 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1182 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
1183 
may be started with a set of disc files (such as all the source for
1184 
a program) by invoking it with a list of file names as arguments, and
1185 
more may be added or deleted on demand.
1186 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1187 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1188 
<span style="font-size: 9pt"><tt>B discfile1 discfile2 ...</tt></span></p>
1189 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1190 
 
1191 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1192 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1193 
<span style="font-size: 10pt">adds the named files to
1194 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
1195 
list, and
1196 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1197 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1198 
<span style="font-size: 9pt"><tt>D discfile1 discfile2 ...</tt></span></p>
1199 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1200 
 
1201 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1202 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1203 
<span style="font-size: 10pt">removes them from
1204 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
1205 
memory (without effect on associated disc files).
1206 
Both these commands have a syntax for using the shell<sup></sup></span><sup><span style="font-size: 6pt">7</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
1207 
(the UNIX command interpreter) to generate the lists:
1208 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1209 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1210 
<span style="font-size: 9pt"><tt>B &lt;echo *.c</tt></span></p>
1211 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1212 
 
1213 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1214 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1215 
<span style="font-size: 10pt">will add all C source files, and
1216 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1217 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1218 
<span style="font-size: 9pt"><tt>B &lt;grep -l variable *.c</tt></span></p>
1219 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1220 
 
1221 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1222 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1223 
<span style="font-size: 10pt">will add all C source files referencing a particular variable
1224 
(the UNIX command
1225 
</span><span style="font-size: 10pt"><tt>grep\fP-l
1226 
lists all files in its arguments that contain matches of
1227 
the specified regular expression).
1228 
Finally,
1229 
</tt></span><span style="font-size: 10pt"><tt>D</tt></span><span style="font-size: 10pt"><tt>
1230 
without arguments deletes the current file.
1231 
</tt></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1232 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1233 
<span style="font-size: 10pt">There are two ways to change which file is current:
1234 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1235 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1236 
<span style="font-size: 9pt"><tt>b filename</tt></span></p>
1237 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1238 
 
1239 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1240 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1241 
<span style="font-size: 10pt">makes the named file current.
1242 
The
1243 
</span><span style="font-size: 10pt"><tt>B</tt></span><span style="font-size: 10pt">
1244 
command
1245 
does the same, but also adds any new files to
1246 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
1247 
list.
1248 
(In practice, of course, the current file
1249 
is usually chosen by mouse actions, not by textual commands.)
1250 
The other way is to use a form of address that refers to files:
1251 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1252 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1253 
<span style="font-size: 9pt"><tt>"</tt></span><span style="font-size: 9pt"><i>expression</i></span><span style="font-size: 9pt"><tt>" </tt></span><span style="font-size: 9pt"><i>address</i></span><span style="font-size: 9pt"><tt></tt></span></p>
1254 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1255 
 
1256 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1257 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1258 
<span style="font-size: 10pt">refers to the address evaluated in the file whose menu line
1259 
matches the expression (there must be exactly one match).
1260 
For example,
1261 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1262 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1263 
<span style="font-size: 9pt"><tt>"peter.c" 3</tt></span></p>
1264 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1265 
 
1266 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1267 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1268 
<span style="font-size: 10pt">refers to the third line of the file whose name matches
1269 
</span><span style="font-size: 10pt"><tt>peter.c</tt></span><span style="font-size: 10pt">.
1270 
This is most useful in the move
1271 
(</span><span style="font-size: 10pt"><tt>m</tt></span><span style="font-size: 10pt">)
1272 
and copy
1273 
(</span><span style="font-size: 10pt"><tt>t</tt></span><span style="font-size: 10pt">)
1274 
commands:
1275 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1276 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1277 
<span style="font-size: 9pt"><tt>0,$ t "peter.c" 0</tt></span></p>
1278 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1279 
 
1280 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1281 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1282 
<span style="font-size: 10pt">makes a copy of the current file at the beginning of
1283 
</span><span style="font-size: 10pt"><tt>peter.c</tt></span><span style="font-size: 10pt">.
1284 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1285 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1286 
<span style="font-size: 10pt">The
1287 
</span><span style="font-size: 10pt"><tt>X</tt></span><span style="font-size: 10pt">
1288 
command
1289 
is a looping construct, like
1290 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">,
1291 
that refers to files instead of strings:
1292 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1293 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1294 
<span style="font-size: 9pt"><tt>X/</tt></span><span style="font-size: 9pt"><i>expression</i></span><span style="font-size: 9pt"><tt>/ </tt></span><span style="font-size: 9pt"><i>command</i></span><span style="font-size: 9pt"><tt></tt></span></p>
1295 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1296 
 
1297 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1298 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1299 
<span style="font-size: 10pt">runs the command in all
1300 
files whose menu lines match the expression.  The best example is
1301 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1302 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1303 
<span style="font-size: 9pt"><tt>X/&rsquo;/ w</tt></span></p>
1304 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1305 
 
1306 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1307 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1308 
<span style="font-size: 10pt">which writes to disc all modified files.
1309 
</span><span style="font-size: 10pt"><tt>Y</tt></span><span style="font-size: 10pt">
1310 
is the complement of
1311 
</span><span style="font-size: 10pt"><tt>X</tt></span><span style="font-size: 10pt">:
1312 
it runs the command on all files whose menu lines don&rsquo;t match the expression:
1313 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1314 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1315 
<span style="font-size: 9pt"><tt>Y/\.c/ D</tt></span></p>
1316 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1317 
 
1318 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1319 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1320 
<span style="font-size: 10pt">deletes all files that don&rsquo;t have
1321 
</span><span style="font-size: 10pt"><tt>.c</tt></span><span style="font-size: 10pt">
1322 
in their names, that is, it keeps all C source files and deletes the rest.
1323 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1324 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1325 
<span style="font-size: 10pt">Braces allow commands to be grouped, so
1326 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1327 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1328 
<span style="font-size: 9pt"><tt>{</tt></span></p>
1329 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1330 
<span style="font-size: 9pt"><tt>    </tt></span><span style="font-size: 9pt"><i>command1</i></span><span style="font-size: 9pt"><tt></tt></span></p>
1331 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1332 
<span style="font-size: 9pt"><tt>    </tt></span><span style="font-size: 9pt"><i>command2</i></span><span style="font-size: 9pt"><tt></tt></span></p>
1333 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1334 
<span style="font-size: 9pt"><tt>}</tt></span></p>
1335 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1336 
 
1337 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1338 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1339 
<span style="font-size: 10pt">is syntactically a single command that runs two commands.
1340 
Thus,
1341 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1342 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1343 
<span style="font-size: 9pt"><tt>X/\.c/ ,g/variable/ {</tt></span></p>
1344 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1345 
<span style="font-size: 9pt"><tt>    f</tt></span></p>
1346 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1347 
<span style="font-size: 9pt"><tt>    , x/.*\n/ g/variable/ p</tt></span></p>
1348 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1349 
<span style="font-size: 9pt"><tt>}</tt></span></p>
1350 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1351 
 
1352 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1353 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1354 
<span style="font-size: 10pt">finds all occurrences of
1355 
</span><span style="font-size: 10pt"><tt>variable</tt></span><span style="font-size: 10pt">
1356 
in C source files, and prints
1357 
out the file names and lines of each match.
1358 
The precise semantics of compound operations is discussed in the implementation
1359 
sections below.
1360 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1361 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1362 
<span style="font-size: 10pt">Finally,
1363 
the undo command,
1364 
</span><span style="font-size: 10pt"><tt>u</tt></span><span style="font-size: 10pt">,
1365 
undoes the last command,
1366 
no matter how many files were affected.
1367 
Multiple undo operations move further back in time, so
1368 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1369 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1370 
<span style="font-size: 9pt"><tt>u</tt></span></p>
1371 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1372 
<span style="font-size: 9pt"><tt>u</tt></span></p>
1373 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1374 
 
1375 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1376 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1377 
<span style="font-size: 10pt">(which may be abbreviated
1378 
</span><span style="font-size: 10pt"><tt>u2</tt></span><span style="font-size: 10pt">)
1379 
undoes the last two commands.  An undo may not be undone, however, nor
1380 
may any command that adds or deletes files.
1381 
Everything else is undoable, though, including for example
1382 
</span><span style="font-size: 10pt"><tt>e</tt></span><span style="font-size: 10pt">
1383 
commands:
1384 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1385 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1386 
<span style="font-size: 9pt"><tt>e filename</tt></span></p>
1387 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1388 
<span style="font-size: 9pt"><tt>u</tt></span></p>
1389 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1390 
 
1391 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1392 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1393 
<span style="font-size: 10pt">restores the state of the file completely, including its name, dot,
1394 
and modified bit.  Because of the undo, potentially dangerous commands
1395 
are not guarded by confirmations.  Only
1396 
</span><span style="font-size: 10pt"><tt>D</tt></span><span style="font-size: 10pt">,
1397 
which destroys the information necessary to restore itself, is protected.
1398 
It will not delete a modified file, but a second
1399 
</span><span style="font-size: 10pt"><tt>D</tt></span><span style="font-size: 10pt">
1400 
of the same file will succeed regardless.
1401 
The
1402 
</span><span style="font-size: 10pt"><tt>q</tt></span><span style="font-size: 10pt">
1403 
command, which exits
1404 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">,
1405 
is similarly guarded.
1406 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
1407 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1408 
<span style="font-size: 10pt"><b>Mouse Interface
1409 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1410 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1411 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
1412 
is most commonly run
1413 
connected to a bitmap display and mouse for interactive editing.
1414 
The only difference in the command language
1415 
between regular, mouse-driven
1416 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1417 
and
1418 
</span><span style="font-size: 10pt"><tt>sam\fP-d
1419 
is that if an address
1420 
is provided without a command,
1421 
</tt></span><span style="font-size: 10pt"><tt>sam\fP-d
1422 
will print the text referenced by the address, but
1423 
regular
1424 
</tt></span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt"><tt>
1425 
will highlight it on the screen &mdash; in fact,
1426 
dot is always highlighted (see Figure 2).
1427 
</tt></span><span style="font-size: 10pt"></span></p><center><img src="fig3.gif" /></center>
1428 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1429 
<span style="font-size: 8pt"><i>Figure 2. A
1430 
</i></span><span style="font-size: 8pt"><tt>sam</tt></span><span style="font-size: 8pt"><i>
1431 
window.  The scroll bar down the left
1432 
represents the file, with the bubble showing the fraction
1433 
visible in the window.
1434 
The scroll bar may be manipulated by the mouse for convenient browsing.
1435 
The current text,
1436 
which is highlighted, need not fit on a line.  Here it consists of one partial
1437 
line, one complete line, and final partial line.
1438 
</i></span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
1439 
<p style="margin-top: 0; margin-bottom: 0.02in"></p>
1440 
 
1441 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
1442 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1443 
<span style="font-size: 10pt">Each file may have zero or more windows open on the display.
1444 
At any time, only one window in all of
1445 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1446 
is the
1447 
</span><span style="font-size: 10pt"><i>current window,
1448 
</i></span><span style="font-size: 10pt">that is, the window to which typing and mouse actions refer;
1449 
this may be the
1450 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1451 
window (that in which commands may be typed)
1452 
or one of the file windows.
1453 
When a file has multiple windows, the image of the file in each window
1454 
is always kept up to date.
1455 
The current file is the last file affected by a command,
1456 
so if the
1457 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1458 
window is current,
1459 
the current window is not a window on the current file.
1460 
However, each window on a file has its own value of dot,
1461 
and when switching between windows on a single file,
1462 
the file&rsquo;s value of dot is changed to that of the window.
1463 
Thus, flipping between windows behaves in the obvious, convenient way.
1464 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1465 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1466 
<span style="font-size: 10pt">The mouse on the Blit has three buttons, numbered left to right.
1467 
Button 3 has a list of commands to manipulate windows,
1468 
followed by a list of &lsquo;menu lines&rsquo; exactly as printed by the
1469 
</span><span style="font-size: 10pt"><tt>f</tt></span><span style="font-size: 10pt">
1470 
command, one per file (not one per window).
1471 
These menu lines are sorted by file name.
1472 
If the list is long, the Blit menu software will make it more manageable
1473 
by generating a scrolling menu instead of an unwieldy long list.
1474 
Using the menu to select a file from the list makes that file the current
1475 
file, and the most recently current window in that file the current window.
1476 
But if that file is already current, selecting it in the menu cycles through
1477 
the windows on the file; this simple trick avoids a special menu to
1478 
choose windows on a file.
1479 
If there is no window open on the file,
1480 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1481 
changes the mouse cursor to prompt the user to create one.
1482 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1483 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1484 
<span style="font-size: 10pt">The commands on the button 3 menu are straightforward (see Figure 3), and
1485 
are like the commands to manipulate windows in
1486 
</span><span style="font-size: 10pt"><tt>mux</tt></span><span style="font-size: 10pt">,<sup></sup></span><sup><span style="font-size: 6pt">8</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
1487 
the Blit&rsquo;s window system.
1488 
</span><span style="font-size: 10pt"><tt>New</tt></span><span style="font-size: 10pt">
1489 
makes a new file, and gives it one empty window, whose size is determined
1490 
by a rectangle swept by the mouse.
1491 
</span><span style="font-size: 10pt"><tt>Zerox</tt></span><span style="font-size: 10pt">
1492 
prompts for a window to be selected, and
1493 
makes a clone of that window; this is how multiple windows are created on one file.
1494 
</span><span style="font-size: 10pt"><tt>Reshape</tt></span><span style="font-size: 10pt">
1495 
changes the size of the indicated window, and
1496 
</span><span style="font-size: 10pt"><tt>close</tt></span><span style="font-size: 10pt">
1497 
deletes it.  If that is the last window open on the file,
1498 
</span><span style="font-size: 10pt"><tt>close</tt></span><span style="font-size: 10pt">
1499 
first does a
1500 
</span><span style="font-size: 10pt"><tt>D</tt></span><span style="font-size: 10pt">
1501 
command on the file.
1502 
</span><span style="font-size: 10pt"><tt>Write</tt></span><span style="font-size: 10pt">
1503 
is identical to a
1504 
</span><span style="font-size: 10pt"><tt>w</tt></span><span style="font-size: 10pt">
1505 
command on the file; it is in the menu purely for convenience.
1506 
Finally,
1507 
</span><span style="font-size: 10pt"><tt>~~sam~~</tt></span><span style="font-size: 10pt">
1508 
is a menu item that appears between the commands and the file names.
1509 
Selecting it makes the
1510 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1511 
window the current window,
1512 
causing subsequent typing to be interpreted as commands.
1513 
</span><span style="font-size: 10pt"></span></p><center><img src="fig2.gif" /></center>
1514 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1515 
<span style="font-size: 8pt"><i>Figure 3. The menu on button 3.
1516 
The black rectangle on the left is a scroll bar; the menu is limited to
1517 
the length shown to prevent its becoming unwieldy.
1518 
Above the
1519 
</i></span><span style="font-size: 8pt"><tt>~~sam~~</tt></span><span style="font-size: 8pt"><i>
1520 
line is a list of commands;
1521 
beneath it is a list of files, presented exactly as with the
1522 
</i></span><span style="font-size: 8pt"><tt>f</tt></span><span style="font-size: 8pt"><i>
1523 
command.
1524 
</i></span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
1525 
<p style="margin-top: 0; margin-bottom: 0.02in"></p>
1526 
 
1527 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
1528 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1529 
<span style="font-size: 10pt">When
1530 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1531 
requests that a window be swept, in response to
1532 
</span><span style="font-size: 10pt"><tt>new</tt></span><span style="font-size: 10pt">,
1533 
</span><span style="font-size: 10pt"><tt>zerox</tt></span><span style="font-size: 10pt">
1534 
or
1535 
</span><span style="font-size: 10pt"><tt>reshape</tt></span><span style="font-size: 10pt">,
1536 
it changes the mouse cursor from the usual arrow to a box with
1537 
a small arrow.
1538 
In this state, the mouse may be used to indicate an arbitrary rectangle by
1539 
pressing button 3 at one corner and releasing it at the opposite corner.
1540 
More conveniently,
1541 
button 3 may simply be clicked,
1542 
whereupon
1543 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1544 
creates the maximal rectangle that contains the cursor
1545 
and abuts the
1546 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1547 
window.
1548 
By placing the
1549 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1550 
window in the middle of the screen, the user can define two regions (one above,
1551 
one below) in which stacked fully-overlapping
1552 
windows can be created with minimal fuss (see Figure 1).
1553 
This simple user interface trick makes window creation noticeably easier.
1554 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1555 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1556 
<span style="font-size: 10pt">The cut-and-paste editor is essentially the same as that in Smalltalk-80.<sup></sup></span><sup><span style="font-size: 6pt">11</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
1557 
The text in dot is always highlighted on the screen.
1558 
When a character is typed it replaces dot, and sets dot to the null
1559 
string after the character.  Thus, ordinary typing inserts text.
1560 
Button 1 is used for selection:
1561 
pressing the button, moving the mouse, and lifting the button
1562 
selects (sets dot to) the text between the points where the
1563 
button was pressed and released.
1564 
Pressing and releasing at the same point selects a null string; this
1565 
is called clicking.  Clicking twice quickly, or
1566 
</span><span style="font-size: 10pt"><i>double clicking,
1567 
</i></span><span style="font-size: 10pt">selects larger objects;
1568 
for example, double clicking in a word selects the word,
1569 
double clicking just inside an opening bracket selects the text
1570 
contained in the brackets (handling nested brackets correctly),
1571 
and similarly for
1572 
parentheses, quotes, and so on.
1573 
The double-clicking rules reflect a bias toward
1574 
programmers.
1575 
If
1576 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1577 
were intended more for word processing, double-clicks would probably
1578 
select linguistic structures such as sentences.
1579 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1580 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1581 
<span style="font-size: 10pt">If button 1 is pressed outside the current window, it makes the indicated
1582 
window current.
1583 
This is the easiest way to switch between windows and files.
1584 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1585 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1586 
<span style="font-size: 10pt">Pressing button 2 brings up a menu of editing functions (see Figure 4).
1587 
These mostly apply to the selected text:
1588 
</span><span style="font-size: 10pt"><tt>cut</tt></span><span style="font-size: 10pt">
1589 
deletes the selected text, and remembers it in a hidden buffer called the
1590 
</span><span style="font-size: 10pt"><i>snarf buffer,
1591 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>paste</tt></span><span style="font-size: 10pt">
1592 
replaces the selected text by the contents of the snarf buffer,
1593 
</span><span style="font-size: 10pt"><tt>snarf</tt></span><span style="font-size: 10pt">
1594 
just copies the selected text to the snarf buffer,
1595 
</span><span style="font-size: 10pt"><tt>look</tt></span><span style="font-size: 10pt">
1596 
searches forward for the next literal occurrence of the selected text, and
1597 
</span><span style="font-size: 10pt"><tt>&lt;mux&gt;</tt></span><span style="font-size: 10pt">
1598 
exchanges snarf buffers with the window system in which
1599 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1600 
is running.
1601 
Finally, the last regular expression used appears as a menu entry
1602 
to search
1603 
forward for the next occurrence of a match for the expression.
1604 
</span><span style="font-size: 10pt"></span></p><center><img src="fig4.gif" /></center>
1605 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1606 
<span style="font-size: 8pt"><i>Figure 4. The menu on button 2.
1607 
The bottom entry tracks the most recently used regular expression, which may
1608 
be literal text.
1609 
</i></span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
1610 
<p style="margin-top: 0; margin-bottom: 0.02in"></p>
1611 
 
1612 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
1613 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1614 
<span style="font-size: 10pt">The relationship between the command language and the mouse language is
1615 
entirely due to the equality of dot and the selected text chosen
1616 
with button 1 on the mouse.
1617 
For example, to make a set of changes in a C subroutine, dot can be
1618 
set by double clicking on the left brace that begins the subroutine,
1619 
which sets dot for the command language.
1620 
An address-free command then typed in the
1621 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1622 
window will apply only to the text between the opening and closing
1623 
braces of the function.
1624 
The idea is to select what you want, and then say what you want
1625 
to do with it, whether invoked by a menu selection or by a typed command.
1626 
And of course, the value of dot is highlighted on
1627 
the display after the command completes.
1628 
This relationship between mouse interface and command language
1629 
is clumsy to explain, but comfortable, even natural, in practice.
1630 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
1631 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1632 
<span style="font-size: 10pt"><b>The Implementation
1633 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1634 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1635 
<span style="font-size: 10pt">The next few sections describe how
1636 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1637 
is put together, first the host part,
1638 
then the inter-component communication,
1639 
then the terminal part.
1640 
After explaining how the command language is implemented,
1641 
the discussion follows (roughly) the path of a character
1642 
from the temporary file on disc to the screen.
1643 
The presentation centers on the data structures,
1644 
because that is how the program was designed and because
1645 
the algorithms are easy to provide, given the right data
1646 
structures.
1647 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
1648 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1649 
<span style="font-size: 10pt"><b>Parsing and execution
1650 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1651 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1652 
<span style="font-size: 10pt">The command language is interpreted by parsing each command with a
1653 
table-driven recursive
1654 
descent parser, and when a complete command is assembled, invoking a top-down
1655 
executor.
1656 
Most editors instead employ a simple character-at-a-time
1657 
lexical scanner.
1658 
Use of a parser makes it
1659 
easy and unambiguous to detect when a command is complete,
1660 
which has two advantages.
1661 
First, escape conventions such as backslashes to quote
1662 
multiple-line commands are unnecessary;  if the command isn&rsquo;t finished,
1663 
the parser keeps reading.  For example, a multiple-line append driven by an
1664 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
1665 
command is straightforward:
1666 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1667 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1668 
<span style="font-size: 9pt"><tt>x/.*\n/ g/Peter/ a</tt></span></p>
1669 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1670 
<span style="font-size: 9pt"><tt>one line about Peter</tt></span></p>
1671 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1672 
<span style="font-size: 9pt"><tt>another line about Peter</tt></span></p>
1673 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1674 
<span style="font-size: 9pt"><tt>.</tt></span></p>
1675 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1676 
 
1677 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1678 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1679 
<span style="font-size: 10pt">Other UNIX editors would require a backslash after all but the last line.
1680 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1681 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1682 
<span style="font-size: 10pt">The other advantage is specific to the two-process structure of
1683 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">.
1684 
The host process must decide when a command is completed so the
1685 
command interpreter can be called.  This problem is easily resolved
1686 
by having the lexical analyzer read the single stream of events from the
1687 
terminal, directly executing all typing and mouse commands,
1688 
but passing to the parser characters typed to the
1689 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1690 
command window.
1691 
This scheme is slightly complicated by the availability of cut-and-paste
1692 
editing in the
1693 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1694 
window, but that difficulty is resolved by applying the rules
1695 
used in
1696 
</span><span style="font-size: 10pt"><tt>mux</tt></span><span style="font-size: 10pt">:
1697 
when a newline is typed to the
1698 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1699 
window, all text between the newline and the previously typed newline
1700 
is made available to the parser.
1701 
This permits arbitrary editing to be done to a command before
1702 
typing newline and thereby requesting execution.
1703 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1704 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1705 
<span style="font-size: 10pt">The parser is driven by a table because the syntax of addresses
1706 
and commands is regular enough
1707 
to be encoded compactly.  There are few special cases, such as the
1708 
replacement text in a substitution, so the syntax of almost all commands
1709 
can be encoded with a few flags.
1710 
These include whether the command allows an address (for example,
1711 
</span><span style="font-size: 10pt"><tt>e</tt></span><span style="font-size: 10pt">
1712 
does not), whether it takes a regular expression (as in
1713 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
1714 
and
1715 
</span><span style="font-size: 10pt"><tt>s</tt></span><span style="font-size: 10pt">),
1716 
whether it takes replacement text (as in
1717 
</span><span style="font-size: 10pt"><tt>c</tt></span><span style="font-size: 10pt">
1718 
or
1719 
</span><span style="font-size: 10pt"><tt>i</tt></span><span style="font-size: 10pt">),
1720 
which may be multi-line, and so on.
1721 
The internal syntax of regular expressions is handled by a separate
1722 
parser; a regular expression is a leaf of the command parse tree.
1723 
Regular expressions are discussed fully in the next section.
1724 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1725 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1726 
<span style="font-size: 10pt">The parser table also has information about defaults, so the interpreter
1727 
is always called with a complete tree.  For example, the parser fills in
1728 
the implicit
1729 
</span><span style="font-size: 10pt"><tt>0</tt></span><span style="font-size: 10pt">
1730 
and
1731 
</span><span style="font-size: 10pt"><tt>$</tt></span><span style="font-size: 10pt">
1732 
in the abbreviated address
1733 
</span><span style="font-size: 10pt"><tt>,</tt></span><span style="font-size: 10pt">
1734 
(comma),
1735 
inserts a
1736 
</span><span style="font-size: 10pt"><tt>+</tt></span><span style="font-size: 10pt">
1737 
to the left of an unadorned regular expression in an address,
1738 
and provides the usual default address
1739 
</span><span style="font-size: 10pt"><tt>.</tt></span><span style="font-size: 10pt">
1740 
(dot) for commands that expect an address but are not given one.
1741 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1742 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1743 
<span style="font-size: 10pt">Once a complete command is parsed, the evaluation is easy.
1744 
The address is evaluated left-to-right starting from the value of dot,
1745 
with a mostly ordinary expression evaluator.
1746 
Addresses, like many of the data structures in
1747 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">,
1748 
are held in a C structure and passed around by value:
1749 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1750 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1751 
<span style="font-size: 9pt"><tt>typedef long Posn;    /* Position in a file */</tt></span></p>
1752 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1753 
<span style="font-size: 9pt"><tt>typedef struct Range{</tt></span></p>
1754 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1755 
<span style="font-size: 9pt"><tt>        Posn    p1, p2;</tt></span></p>
1756 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1757 
<span style="font-size: 9pt"><tt>}Range;</tt></span></p>
1758 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1759 
<span style="font-size: 9pt"><tt>typedef struct Address{</tt></span></p>
1760 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1761 
<span style="font-size: 9pt"><tt>        Range   r;</tt></span></p>
1762 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1763 
<span style="font-size: 9pt"><tt>        File    *f;</tt></span></p>
1764 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1765 
<span style="font-size: 9pt"><tt>}Address;</tt></span></p>
1766 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1767 
 
1768 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1769 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1770 
<span style="font-size: 10pt">An address is encoded as a substring (character positions
1771 
</span><span style="font-size: 10pt"><tt>p1</tt></span><span style="font-size: 10pt">
1772 
to
1773 
</span><span style="font-size: 10pt"><tt>p2</tt></span><span style="font-size: 10pt">)
1774 
in a file
1775 
</span><span style="font-size: 10pt"><tt>f</tt></span><span style="font-size: 10pt">.
1776 
(The data type
1777 
</span><span style="font-size: 10pt"><tt>File</tt></span><span style="font-size: 10pt">
1778 
is described in detail below.)
1779 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1780 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1781 
<span style="font-size: 10pt">The address interpreter is an
1782 
</span><span style="font-size: 10pt"><tt>Address</tt></span><span style="font-size: 10pt">-valued
1783 
function that traverses the parse tree describing an address (the
1784 
parse tree for the address has type
1785 
</span><span style="font-size: 10pt"><tt>Addrtree</tt></span><span style="font-size: 10pt">):
1786 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1787 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1788 
<span style="font-size: 9pt"><tt>Address</tt></span></p>
1789 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1790 
<span style="font-size: 9pt"><tt>address(ap, a, sign)</tt></span></p>
1791 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1792 
<span style="font-size: 9pt"><tt>    Addrtree *ap;</tt></span></p>
1793 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1794 
<span style="font-size: 9pt"><tt>    Address a;</tt></span></p>
1795 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1796 
<span style="font-size: 9pt"><tt>    int sign;</tt></span></p>
1797 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1798 
<span style="font-size: 9pt"><tt>{</tt></span></p>
1799 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1800 
<span style="font-size: 9pt"><tt>    Address a2;</tt></span></p>
1801 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1802 
<span style="font-size: 9pt"><tt>    do</tt></span></p>
1803 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1804 
<span style="font-size: 9pt"><tt>        switch(ap-&gt;type){</tt></span></p>
1805 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1806 
<span style="font-size: 9pt"><tt>        case &rsquo;.&rsquo;:</tt></span></p>
1807 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1808 
<span style="font-size: 9pt"><tt>            a=a.f-&gt;dot;</tt></span></p>
1809 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1810 
<span style="font-size: 9pt"><tt>            break;</tt></span></p>
1811 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1812 
<span style="font-size: 9pt"><tt>        case &rsquo;$&rsquo;:</tt></span></p>
1813 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1814 
<span style="font-size: 9pt"><tt>            a.r.p1=a.r.p2=a.f-&gt;nbytes;</tt></span></p>
1815 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1816 
<span style="font-size: 9pt"><tt>            break;</tt></span></p>
1817 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1818 
<span style="font-size: 9pt"><tt>        case &rsquo;"&rsquo;:   </tt></span></p>
1819 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1820 
<span style="font-size: 9pt"><tt>            a=matchfile(a, ap-&gt;aregexp)-&gt;dot; </tt></span></p>
1821 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1822 
<span style="font-size: 9pt"><tt>            break;</tt></span></p>
1823 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1824 
<span style="font-size: 9pt"><tt>        case &rsquo;,&rsquo;:</tt></span></p>
1825 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1826 
<span style="font-size: 9pt"><tt>            a2=address(ap-&gt;right, a, 0);</tt></span></p>
1827 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1828 
<span style="font-size: 9pt"><tt>            a=address(ap-&gt;left, a, 0);</tt></span></p>
1829 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1830 
<span style="font-size: 9pt"><tt>            if(a.f!=a2.f || a2.r.p2&lt;a.r.p1)</tt></span></p>
1831 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1832 
<span style="font-size: 9pt"><tt>                error(Eorder);</tt></span></p>
1833 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1834 
<span style="font-size: 9pt"><tt>            a.r.p2=a2.r.p2;</tt></span></p>
1835 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1836 
<span style="font-size: 9pt"><tt>            return a;</tt></span></p>
1837 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1838 
<span style="font-size: 9pt"><tt>        /* and so on */</tt></span></p>
1839 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1840 
<span style="font-size: 9pt"><tt>        }</tt></span></p>
1841 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1842 
<span style="font-size: 9pt"><tt>    while((ap=ap-&gt;right)!=0);</tt></span></p>
1843 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1844 
<span style="font-size: 9pt"><tt>    return a;</tt></span></p>
1845 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1846 
<span style="font-size: 9pt"><tt>}</tt></span></p>
1847 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1848 
 
1849 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1850 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
1851 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1852 
<span style="font-size: 10pt">Throughout, errors are handled by a non-local
1853 
</span><span style="font-size: 10pt"><tt>goto</tt></span><span style="font-size: 10pt">
1854 
(a
1855 
</span><span style="font-size: 10pt"><tt>setjmp/longjmp</tt></span><span style="font-size: 10pt">
1856 
in C terminology)
1857 
hidden in a routine called
1858 
</span><span style="font-size: 10pt"><tt>error</tt></span><span style="font-size: 10pt">
1859 
that immediately aborts the execution, retracts any
1860 
partially made changes (see the section below on &lsquo;undoing&rsquo;), and
1861 
returns to the top level of the parser.
1862 
The argument to
1863 
</span><span style="font-size: 10pt"><tt>error</tt></span><span style="font-size: 10pt">
1864 
is an enumeration type that
1865 
is translated to a terse but possibly helpful
1866 
message such as &lsquo;?addresses out of order.&rsquo;
1867 
Very common messages are kept short; for example the message for
1868 
a failed regular expression search is &lsquo;?search.&rsquo;
1869 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1870 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1871 
<span style="font-size: 10pt">Character addresses such as
1872 
</span><span style="font-size: 10pt"><tt>#3</tt></span><span style="font-size: 10pt">
1873 
are trivial to implement, as the
1874 
</span><span style="font-size: 10pt"><tt>File</tt></span><span style="font-size: 10pt">
1875 
data structure is accessible by character number.
1876 
However,
1877 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1878 
keeps no information about the position of newlines &mdash; it is too
1879 
expensive to track dynamically &mdash; so line addresses are computed by reading
1880 
the file, counting newlines.  Except in very large files, this has proven
1881 
acceptable: file access is fast enough to make the technique practical,
1882 
and lines are not central to the structure of the command language.
1883 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1884 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1885 
<span style="font-size: 10pt">The command interpreter, called
1886 
</span><span style="font-size: 10pt"><tt>cmdexec</tt></span><span style="font-size: 10pt">,
1887 
is also straightforward.  The parse table includes a
1888 
function to call to interpret a particular command.  That function
1889 
receives as arguments
1890 
the calculated address
1891 
for the command
1892 
and the command tree (of type
1893 
</span><span style="font-size: 10pt"><tt>Cmdtree</tt></span><span style="font-size: 10pt">),
1894 
which may contain information such as the subtree for compound commands.
1895 
Here, for example, is the function for the
1896 
</span><span style="font-size: 10pt"><tt>g</tt></span><span style="font-size: 10pt">
1897 
and
1898 
</span><span style="font-size: 10pt"><tt>v</tt></span><span style="font-size: 10pt">
1899 
commands:
1900 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1901 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1902 
<span style="font-size: 9pt"><tt>int</tt></span></p>
1903 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1904 
<span style="font-size: 9pt"><tt>g_cmd(a, cp)</tt></span></p>
1905 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1906 
<span style="font-size: 9pt"><tt>    Address a;</tt></span></p>
1907 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1908 
<span style="font-size: 9pt"><tt>    Cmdtree *cp;</tt></span></p>
1909 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1910 
<span style="font-size: 9pt"><tt>{</tt></span></p>
1911 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1912 
<span style="font-size: 9pt"><tt>    compile(cp-&gt;regexp);</tt></span></p>
1913 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1914 
<span style="font-size: 9pt"><tt>    if(execute(a.f, a.r.p1, a.r.p2)!=(cp-&gt;cmdchar==&rsquo;v&rsquo;)){</tt></span></p>
1915 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1916 
<span style="font-size: 9pt"><tt>        a.f-&gt;dot=a;</tt></span></p>
1917 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1918 
<span style="font-size: 9pt"><tt>        return cmdexec(a, cp-&gt;subcmd);</tt></span></p>
1919 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1920 
<span style="font-size: 9pt"><tt>    }</tt></span></p>
1921 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1922 
<span style="font-size: 9pt"><tt>    return TRUE;    /* cause execution to continue */</tt></span></p>
1923 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1924 
<span style="font-size: 9pt"><tt>}</tt></span></p>
1925 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
1926 
 
1927 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
1928 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1929 
<span style="font-size: 10pt">(</span><span style="font-size: 10pt"><tt>Compile</tt></span><span style="font-size: 10pt">
1930 
and
1931 
</span><span style="font-size: 10pt"><tt>execute</tt></span><span style="font-size: 10pt">
1932 
are part of the regular expression code, described in the next section.)
1933 
Because the parser and the
1934 
</span><span style="font-size: 10pt"><tt>File</tt></span><span style="font-size: 10pt">
1935 
data structure do most of the work, most commands
1936 
are similarly brief.
1937 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
1938 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1939 
<span style="font-size: 10pt"><b>Regular expressions
1940 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1941 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1942 
<span style="font-size: 10pt">The regular expression code in
1943 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1944 
is an interpreted, rather than compiled on-the-fly, implementation of Thompson&rsquo;s
1945 
non-deterministic finite automaton algorithm.<sup></sup></span><sup><span style="font-size: 6pt">12</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
1946 
The syntax and semantics of the expressions are as in the UNIX program
1947 
</span><span style="font-size: 10pt"><tt>egrep</tt></span><span style="font-size: 10pt">,
1948 
including alternation, closures, character classes, and so on.
1949 
The only changes in the notation are two additions:
1950 
</span><span style="font-size: 10pt"><tt>\n</tt></span><span style="font-size: 10pt">
1951 
is translated to, and matches, a newline character, and
1952 
</span><span style="font-size: 10pt"><tt>@</tt></span><span style="font-size: 10pt">
1953 
matches any character.  In
1954 
</span><span style="font-size: 10pt"><tt>egrep</tt></span><span style="font-size: 10pt">,
1955 
the character
1956 
</span><span style="font-size: 10pt"><tt>.</tt></span><span style="font-size: 10pt">
1957 
matches any character except newline, and in
1958 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
1959 
the same rule seemed safest, to prevent idioms like
1960 
</span><span style="font-size: 10pt"><tt>.*</tt></span><span style="font-size: 10pt">
1961 
from spanning newlines.
1962 
</span><span style="font-size: 10pt"><tt>Egrep</tt></span><span style="font-size: 10pt">
1963 
expressions are arguably too complicated for an interactive editor &mdash;
1964 
certainly it would make sense if all the special characters were two-character
1965 
sequences, so that most of the punctuation characters wouldn&rsquo;t have
1966 
peculiar meanings &mdash; but for an interesting command language, full
1967 
regular expressions are necessary, and
1968 
</span><span style="font-size: 10pt"><tt>egrep</tt></span><span style="font-size: 10pt">
1969 
defines the full regular expression syntax for UNIX programs.
1970 
Also, it seemed superfluous to define a new syntax, since various UNIX programs
1971 
(</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">,
1972 
</span><span style="font-size: 10pt"><tt>egrep</tt></span><span style="font-size: 10pt">
1973 
and
1974 
</span><span style="font-size: 10pt"><tt>vi</tt></span><span style="font-size: 10pt">)
1975 
define too many already.
1976 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1977 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1978 
<span style="font-size: 10pt">The expressions are compiled by a routine,
1979 
</span><span style="font-size: 10pt"><tt>compile</tt></span><span style="font-size: 10pt">,
1980 
that generates the description of the non-deterministic finite state machine.
1981 
A second routine,
1982 
</span><span style="font-size: 10pt"><tt>execute</tt></span><span style="font-size: 10pt">,
1983 
interprets the machine to generate the leftmost-longest match of the
1984 
expression in a substring of the file.
1985 
The algorithm is described elsewhere.<sup></sup></span><sup><span style="font-size: 6pt">12,13</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
1986 
</span><span style="font-size: 10pt"><tt>Execute</tt></span><span style="font-size: 10pt">
1987 
reports
1988 
whether a match was found, and sets a global variable,
1989 
of type
1990 
</span><span style="font-size: 10pt"><tt>Range</tt></span><span style="font-size: 10pt">,
1991 
to the substring matched.
1992 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
1993 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1994 
<span style="font-size: 10pt">A trick is required to evaluate the expression in reverse, such as when
1995 
searching backwards for an expression.
1996 
For example,
1997 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
1998 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
1999 
<span style="font-size: 9pt"><tt>-/P.*r/</tt></span></p>
2000 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
2001 
 
2002 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
2003 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2004 
<span style="font-size: 10pt">looks backwards through the file for a match of the expression.
2005 
The expression, however, is defined for a forward search.
2006 
The solution is to construct a machine identical to the machine
2007 
for a forward search except for a reversal of all the concatenation
2008 
operators (the other operators are symmetric under direction reversal),
2009 
to exchange the meaning of the operators
2010 
</span><span style="font-size: 10pt"><tt>^</tt></span><span style="font-size: 10pt">
2011 
and
2012 
</span><span style="font-size: 10pt"><tt>$</tt></span><span style="font-size: 10pt">,
2013 
and then to read the file backwards, looking for the
2014 
usual earliest longest match.
2015 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2016 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2017 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Execute</tt></span><span style="font-size: 10pt">
2018 
generates only one match each time it is called.
2019 
To interpret looping constructs such as the
2020 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
2021 
command,
2022 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2023 
must therefore synchronize between
2024 
calls of
2025 
</span><span style="font-size: 10pt"><tt>execute</tt></span><span style="font-size: 10pt">
2026 
to avoid
2027 
problems with null matches.
2028 
For example, even given the leftmost-longest rule,
2029 
the expression
2030 
</span><span style="font-size: 10pt"><tt>a*</tt></span><span style="font-size: 10pt">
2031 
matches three times in the string
2032 
</span><span style="font-size: 10pt"><tt>ab</tt></span><span style="font-size: 10pt">
2033 
(the character
2034 
</span><span style="font-size: 10pt"><tt>a</tt></span><span style="font-size: 10pt">,
2035 
the null string between the
2036 
</span><span style="font-size: 10pt"><tt>a</tt></span><span style="font-size: 10pt">
2037 
and
2038 
</span><span style="font-size: 10pt"><tt>b</tt></span><span style="font-size: 10pt">,
2039 
and the final null string).
2040 
After returning a match for the
2041 
</span><span style="font-size: 10pt"><tt>a</tt></span><span style="font-size: 10pt">,
2042 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2043 
must not match the null string before the
2044 
</span><span style="font-size: 10pt"><tt>b</tt></span><span style="font-size: 10pt">.
2045 
The algorithm starts
2046 
</span><span style="font-size: 10pt"><tt>execute</tt></span><span style="font-size: 10pt">
2047 
at the end of its previous match, and
2048 
if the match it returns
2049 
is null and abuts the previous match, rejects the match and advances
2050 
the initial position one character.
2051 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
2052 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2053 
<span style="font-size: 10pt"><b>Memory allocation
2054 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2055 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2056 
<span style="font-size: 10pt">The C language has no memory allocation primitives, although a standard
2057 
library routine,
2058 
</span><span style="font-size: 10pt"><tt>malloc</tt></span><span style="font-size: 10pt">,
2059 
provides adequate service for simple programs.
2060 
For specific uses, however,
2061 
it can be better to write a custom allocator.
2062 
The allocator (or rather, pair of allocators) described here
2063 
work in both the terminal and host parts of
2064 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">.
2065 
They are designed for efficient manipulation of strings,
2066 
which are allocated and freed frequently and vary in length from essentially
2067 
zero to 32 Kbytes (very large strings are written to disc).
2068 
More important, strings may be large and change size often,
2069 
so to minimize memory usage it is helpful to reclaim and to coalesce the
2070 
unused portions of strings when they are truncated.
2071 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2072 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2073 
<span style="font-size: 10pt">Objects to be allocated in
2074 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2075 
are of two flavors:
2076 
the first is C
2077 
</span><span style="font-size: 10pt"><tt>structs</tt></span><span style="font-size: 10pt">,
2078 
which are small and often addressed by pointer variables;
2079 
the second is variable-sized arrays of characters
2080 
or integers whose
2081 
base pointer is always used to access them.
2082 
The memory allocator in
2083 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2084 
is therefore in two parts:
2085 
first, a traditional first-fit allocator that provides fixed storage for
2086 
</span><span style="font-size: 10pt"><tt>structs</tt></span><span style="font-size: 10pt">;
2087 
and second, a garbage-compacting allocator that reduces storage
2088 
overhead for variable-sized objects, at the cost of some bookkeeping.
2089 
The two types of objects are allocated from adjoining arenas, with
2090 
the garbage-compacting allocator controlling the arena with higher addresses.
2091 
Separating into two arenas simplifies compaction and prevents fragmentation due
2092 
to immovable objects.
2093 
The access rules for garbage-compactable objects
2094 
(discussed in the next paragraph) allow them to be relocated, so when
2095 
the first-fit arena needs space, it moves the garbage-compacted arena
2096 
to higher addresses to make room.  Storage is therefore created only
2097 
at successively higher addresses, either when more garbage-compacted
2098 
space is needed or when the first-fit arena pushes up the other arena.
2099 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2100 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2101 
<span style="font-size: 10pt">Objects that may be compacted declare to the
2102 
allocator a cell that is guaranteed to be the sole repository of the
2103 
address of the object whenever a compaction can occur.
2104 
The compactor can then update the address when the object is moved.
2105 
For example, the implementation of type
2106 
</span><span style="font-size: 10pt"><tt>List</tt></span><span style="font-size: 10pt">
2107 
(really a variable-length array)
2108 
is:
2109 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
2110 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2111 
<span style="font-size: 9pt"><tt>typedef struct List{</tt></span></p>
2112 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2113 
<span style="font-size: 9pt"><tt>        int     nused;</tt></span></p>
2114 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2115 
<span style="font-size: 9pt"><tt>        long    *ptr;</tt></span></p>
2116 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2117 
<span style="font-size: 9pt"><tt>}List;</tt></span></p>
2118 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
2119 
 
2120 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
2121 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2122 
<span style="font-size: 10pt">The
2123 
</span><span style="font-size: 10pt"><tt>ptr</tt></span><span style="font-size: 10pt">
2124 
cell must always be used directly, and never copied.  When a
2125 
</span><span style="font-size: 10pt"><tt>List</tt></span><span style="font-size: 10pt">
2126 
is to be created the
2127 
</span><span style="font-size: 10pt"><tt>List</tt></span><span style="font-size: 10pt">
2128 
structure is allocated in the ordinary first-fit arena
2129 
and its
2130 
</span><span style="font-size: 10pt"><tt>ptr</tt></span><span style="font-size: 10pt">
2131 
is allocated in the garbage-compacted arena.
2132 
A similar data type for strings, called
2133 
</span><span style="font-size: 10pt"><tt>String</tt></span><span style="font-size: 10pt">,
2134 
stores variable-length character arrays of up to 32767 elements.
2135 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2136 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2137 
<span style="font-size: 10pt">A related matter of programming style:
2138 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2139 
frequently passes structures by value, which
2140 
simplifies the code.
2141 
Traditionally, C programs have
2142 
passed structures by reference, but implicit allocation on
2143 
the stack is easier to use.
2144 
Structure passing is a relatively new feature of C
2145 
(it is not in the
2146 
standard reference manual for C<sup></sup></span><sup><span style="font-size: 6pt">14</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">), and is poorly supported in most
2147 
commercial C compilers.
2148 
It&rsquo;s convenient and expressive, though,
2149 
and simplifies memory management by
2150 
avoiding the allocator altogether
2151 
and eliminating pointer aliases.
2152 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
2153 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2154 
<span style="font-size: 10pt"><b>Data structures for manipulating files
2155 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2156 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2157 
<span style="font-size: 10pt">Experience with
2158 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
2159 
showed that the requirements
2160 
of the file data structure were few, but strict.
2161 
First, files need to be read and written quickly;
2162 
adding a fresh file must be painless.
2163 
Second, the implementation must place no arbitrary upper limit on
2164 
the number or sizes of files.  (It should be practical to edit many files,
2165 
and files up to megabytes in length should be handled gracefully.)
2166 
This implies that files be stored on disc, not in main memory.
2167 
(Aficionados of virtual memory may argue otherwise, but the
2168 
implementation of virtual
2169 
memory in our system is not something to depend on
2170 
for good performance.)
2171 
Third, changes to files need be made by only two primitives:
2172 
deletion and insertion.
2173 
These are inverses of each other,
2174 
which simplifies the implementation of the undo operation.
2175 
Finally,
2176 
it must be easy and efficient to access the file, either
2177 
forwards or backwards, a byte at a time.
2178 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2179 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2180 
<span style="font-size: 10pt">The
2181 
</span><span style="font-size: 10pt"><tt>File</tt></span><span style="font-size: 10pt">
2182 
data type is constructed from three simpler data structures that hold arrays
2183 
of characters.
2184 
Each of these types has an insertion and deletion operator, and the
2185 
insertion and deletion operators of the
2186 
</span><span style="font-size: 10pt"><tt>File</tt></span><span style="font-size: 10pt">
2187 
type itself are constructed from them.
2188 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2189 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2190 
<span style="font-size: 10pt">The simplest type is the
2191 
</span><span style="font-size: 10pt"><tt>String</tt></span><span style="font-size: 10pt">,
2192 
which is used to hold strings in main memory.
2193 
The code that manages
2194 
</span><span style="font-size: 10pt"><tt>Strings</tt></span><span style="font-size: 10pt">
2195 
guarantees that they will never be longer
2196 
than some moderate size, and in practice they are rarely larger than 8 Kbytes.
2197 
</span><span style="font-size: 10pt"><tt>Strings</tt></span><span style="font-size: 10pt">
2198 
have two purposes: they hold short strings like file names with little overhead,
2199 
and because they are deliberately small, they are efficient to modify.
2200 
They are therefore used as the data structure for in-memory caches.
2201 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2202 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2203 
<span style="font-size: 10pt">The disc copy of the file is managed by a data structure called a
2204 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">,
2205 
which corresponds to a temporary file.  A
2206 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">
2207 
has no storage in main memory other than bookkeeping information;
2208 
the actual data being held is all on the disc.
2209 
To reduce the number of open files needed,
2210 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2211 
opens a dozen temporary UNIX files and multiplexes the
2212 
</span><span style="font-size: 10pt"><tt>Discs</tt></span><span style="font-size: 10pt">
2213 
upon them.
2214 
This permits many files to
2215 
be edited; the entire
2216 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2217 
source (48 files) may be edited comfortably with a single
2218 
instance of
2219 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">.
2220 
Allocating one temporary file per
2221 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">
2222 
would strain the operating system&rsquo;s limit on the number of open files.
2223 
Also, spreading the traffic among temporary files keeps the files shorter,
2224 
and shorter files are more efficiently implemented by the UNIX
2225 
I/O subsystem.
2226 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2227 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2228 
<span style="font-size: 10pt">A
2229 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">
2230 
is an array of fixed-length blocks, each of which contains
2231 
between 1 and 4096 characters of active data.
2232 
(The block size of our UNIX file system is 4096 bytes.)
2233 
The block addresses within the temporary file and the length of each
2234 
block are stored in a
2235 
</span><span style="font-size: 10pt"><tt>List</tt></span><span style="font-size: 10pt">.
2236 
When changes are made the live part of blocks may change size.
2237 
Blocks are created and coalesced when necessary to try to keep the sizes
2238 
between 2048 and 4096 bytes.
2239 
An actively changing part of the
2240 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">
2241 
therefore typically has about a kilobyte of slop that can be
2242 
inserted or deleted
2243 
without changing more than one block or affecting the block order.
2244 
When an insertion would overflow a block, the block is split, a new one
2245 
is allocated to receive the overflow, and the memory-resident list of blocks
2246 
is rearranged to reflect the insertion of the new block.
2247 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2248 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2249 
<span style="font-size: 10pt">Obviously, going to the disc for every modification to the file is
2250 
prohibitively expensive.
2251 
The data type
2252 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">
2253 
consists of a
2254 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">
2255 
to hold the data and a
2256 
</span><span style="font-size: 10pt"><tt>String</tt></span><span style="font-size: 10pt">
2257 
that acts as a cache.
2258 
This is the first of a series of caches throughout the data structures in
2259 
</span><span style="font-size: 10pt"><tt>sam.</tt></span><span style="font-size: 10pt">
2260 
The caches not only improve performance, they provide a way to organize
2261 
the flow of data, particularly in the communication between the host
2262 
and terminal.
2263 
This idea is developed below, in the section on communications.
2264 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2265 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2266 
<span style="font-size: 10pt">To reduce disc traffic, changes to a
2267 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">
2268 
are mediated by a variable-length string, in memory, that acts as a cache.
2269 
When an insertion or deletion is made to a
2270 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">,
2271 
if the change can be accommodated by the cache, it is done there.
2272 
If the cache becomes bigger than a block because of an insertion,
2273 
some of it is written to the
2274 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">
2275 
and deleted from the cache.
2276 
If the change does not intersect the cache, the cache is flushed.
2277 
The cache is only loaded at the new position if the change is smaller than a block;
2278 
otherwise, it is sent directly to the
2279 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">.
2280 
This is because
2281 
large changes are typically sequential,
2282 
whereupon the next change is unlikely to overlap the current one.
2283 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2284 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2285 
<span style="font-size: 10pt">A
2286 
</span><span style="font-size: 10pt"><tt>File</tt></span><span style="font-size: 10pt">
2287 
comprises a
2288 
</span><span style="font-size: 10pt"><tt>String</tt></span><span style="font-size: 10pt">
2289 
to hold the file name and some ancillary data such as dot and the modified bit.
2290 
The most important components, though, are a pair of
2291 
</span><span style="font-size: 10pt"><tt>Buffers</tt></span><span style="font-size: 10pt">,
2292 
one called the transcript and the other the contents.
2293 
Their use is described in the next section.
2294 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2295 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2296 
<span style="font-size: 10pt">The overall structure is shown in Figure 5.
2297 
Although it may seem that the data is touched many times on its
2298 
way from the
2299 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">,
2300 
it is read (by one UNIX system call) directly into the cache of the
2301 
associated
2302 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">;
2303 
no extra copy is done.
2304 
Similarly, when flushing the cache, the text is written
2305 
directly from the cache to disc.
2306 
Most operations act directly on the text in the cache.
2307 
A principle applied throughout
2308 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2309 
is that the fewer times the data is copied, the faster the program will run
2310 
(see also the paper by Waite<sup></sup></span><sup><span style="font-size: 6pt">15</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">).
2311 
</span><span style="font-size: 10pt"></span></p><center><img src="sam2.png"></center>
2312 
</center>
2313 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2314 
<span style="font-size: 8pt"><i>Figure 5. File data structures.
2315 
The temporary files are stored in the standard repository for such files
2316 
on the host system.
2317 
</i></span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
2318 
<p style="margin-top: 0; margin-bottom: 0.02in"></p>
2319 
 
2320 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
2321 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2322 
<span style="font-size: 10pt">The contents of a
2323 
</span><span style="font-size: 10pt"><tt>File</tt></span><span style="font-size: 10pt">
2324 
are accessed by a routine that
2325 
copies to a buffer a substring of a file starting at a specified offset.
2326 
To read a byte at a time, a
2327 
per-</span><span style="font-size: 10pt"><tt>File</tt></span><span style="font-size: 10pt">
2328 
array is loaded starting from a specified initial position,
2329 
and bytes may then be read from the array.
2330 
The implementation is done by a macro similar to the C standard I/O
2331 
</span><span style="font-size: 10pt"><tt>getc</tt></span><span style="font-size: 10pt">
2332 
macro.<sup></sup></span><sup><span style="font-size: 6pt">14</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
2333 
Because the reading may be done at any address, a minor change to the
2334 
macro allows the file to be read backwards.
2335 
This array is read-only; there is no
2336 
</span><span style="font-size: 10pt"><tt>putc</tt></span><span style="font-size: 10pt">.
2337 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
2338 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2339 
<span style="font-size: 10pt"><b>Doing and undoing
2340 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2341 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2342 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
2343 
has an unusual method for managing changes to files.
2344 
The command language makes it easy to specify multiple variable-length changes
2345 
to a file millions of bytes long, and such changes
2346 
must be made efficiently if the editor is to be practical.
2347 
The usual techniques for inserting and deleting strings
2348 
are inadequate under these conditions.
2349 
The
2350 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">
2351 
and
2352 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">
2353 
data structures are designed for efficient random access to long strings,
2354 
but care must be taken to avoid super-linear behavior when making
2355 
many changes simultaneously.
2356 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2357 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2358 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
2359 
uses a two-pass algorithm for making changes, and treats each file as a database
2360 
against which transactions are registered.
2361 
Changes are not made directly to the contents.
2362 
Instead, when a command is started, a &lsquo;mark&rsquo; containing
2363 
a sequence number is placed in the transcript
2364 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">,
2365 
and each change made to the file, either an insertion or deletion
2366 
or a change to the file name,
2367 
is appended to the end of the transcript.
2368 
When the command is complete, the transcript is rewound to the
2369 
mark and applied to the contents.
2370 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2371 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2372 
<span style="font-size: 10pt">One reason for separating evaluation from
2373 
application in this way is to simplify tracking the addresses of changes
2374 
made in the middle of a long sequence.
2375 
The two-pass algorithm also allows all changes to apply to the
2376 
</span><span style="font-size: 10pt"><i>original</i></span><span style="font-size: 10pt">
2377 
data: no change can affect another change made in the same command.
2378 
This is particularly important when evaluating an
2379 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
2380 
command because it prevents regular expression matches
2381 
from stumbling over changes made earlier in the execution.
2382 
Also, the two-pass
2383 
algorithm is cleaner than the way other UNIX editors allow changes to
2384 
affect each other;
2385 
for example,
2386 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">&rsquo;s
2387 
idioms to do things like delete every other line
2388 
depend critically on the implementation.
2389 
Instead,
2390 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
2391 
simple model, in which all changes in a command occur effectively
2392 
simultaneously, is easy to explain and to understand.
2393 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2394 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2395 
<span style="font-size: 10pt">The records in the transcript are of the form &lsquo;&lsquo;delete substring from
2396 
locations
2397 
123 to 456&rsquo;&rsquo; and &lsquo;&lsquo;insert 11 characters &lsquo;hello there&rsquo; at location 789.&rsquo;&rsquo;
2398 
(It is an error if the changes are not at monotonically greater
2399 
positions through the file.)
2400 
While the update is occurring, these numbers must be
2401 
offset by earlier changes, but that is straightforward and
2402 
local to the update routine;
2403 
moreover, all the numbers have been computed
2404 
before the first is examined.
2405 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2406 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2407 
<span style="font-size: 10pt">Treating the file as a transaction system has another advantage:
2408 
undo is trivial.
2409 
All it takes is to invert the transcript after it has been
2410 
implemented, converting insertions
2411 
into deletions and vice versa, and saving them in a holding
2412 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">.
2413 
The &lsquo;do&rsquo; transcript can then be deleted from
2414 
the transcript
2415 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">
2416 
and replaced by the &lsquo;undo&rsquo; transcript.
2417 
If an undo is requested, the transcript is rewound and the undo transcript
2418 
executed.
2419 
Because the transcript
2420 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">
2421 
is not truncated after each command, it accumulates
2422 
successive changes.
2423 
A sequence of undo commands
2424 
can therefore back up the file arbitrarily,
2425 
which is more helpful than the more commonly implemented self-inverse form of undo.
2426 
(</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
2427 
provides no way to undo an undo, but if it were desired,
2428 
it would be easy to provide by re-interpreting the &lsquo;do&rsquo; transcript.)
2429 
Each mark in the transcript contains a sequence number and the offset into
2430 
the transcript of the previous mark, to aid in unwinding the transcript.
2431 
Marks also contain the value of dot and the modified bit so these can be
2432 
restored easily.
2433 
Undoing multiple files is easy; it merely demands undoing all files whose
2434 
latest change has the same sequence number as the current file.
2435 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2436 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2437 
<span style="font-size: 10pt">Another benefit of having a transcript is that errors encountered in the middle
2438 
of a complicated command need not leave the files in an intermediate state.
2439 
By rewinding the transcript to the mark beginning the command,
2440 
the partial command can be trivially undone.
2441 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2442 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2443 
<span style="font-size: 10pt">When the update algorithm was first implemented, it was unacceptably slow,
2444 
so a cache was added to coalesce nearby changes,
2445 
replacing multiple small changes by a single larger one.
2446 
This reduced the number
2447 
of insertions into the transaction
2448 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">,
2449 
and made a dramatic improvement in performance,
2450 
but made it impossible
2451 
to handle changes in non-monotonic order in the file; the caching method
2452 
only works if changes don&rsquo;t overlap.
2453 
Before the cache was added, the transaction could in principle be sorted
2454 
if the changes were out of order, although
2455 
this was never done.
2456 
The current status is therefore acceptable performance with a minor
2457 
restriction on global changes, which is sometimes, but rarely, an annoyance.
2458 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2459 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2460 
<span style="font-size: 10pt">The update algorithm obviously paws the data more than simpler
2461 
algorithms, but it is not prohibitively expensive;
2462 
the caches help.
2463 
(The principle of avoiding copying the data is still honored here,
2464 
although not as piously:
2465 
the data is moved from contents&rsquo; cache to
2466 
the transcript&rsquo;s all at once and through only one internal buffer.)
2467 
Performance figures confirm the efficiency.
2468 
To read from a dead start a hundred kilobyte file on a VAX-11/750
2469 
takes 1.4 seconds of user time, 2.5 seconds of system time,
2470 
and 5 seconds of real time.
2471 
Reading the same file in
2472 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">
2473 
takes 6.0 seconds of user time, 1.7 seconds of system time,
2474 
and 8 seconds of real time.
2475 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
2476 
uses about half the CPU time.
2477 
A more interesting example is the one stated above:
2478 
inserting a character between every pair of characters in the file.
2479 
The
2480 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2481 
command is
2482 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
2483 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2484 
<span style="font-size: 9pt"><tt>,y/@/ a/x/</tt></span></p>
2485 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
2486 
 
2487 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
2488 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2489 
<span style="font-size: 10pt">and takes 3 CPU seconds per kilobyte of input file, of which
2490 
about a third is spent in the regular expression code.
2491 
This translates to about 500 changes per second.
2492 
</span><span style="font-size: 10pt"><tt>Ed</tt></span><span style="font-size: 10pt">
2493 
takes 1.5 seconds per kilobyte to make a similar change (ignoring newlines),
2494 
but cannot undo it.
2495 
The same example in
2496 
</span><span style="font-size: 10pt"><tt>ex</tt></span><span style="font-size: 10pt">,<sup></sup></span><sup><span style="font-size: 6pt">9</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
2497 
a variant of
2498 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">
2499 
done at the University of California at Berkeley,
2500 
which allows one level of undoing, again takes 3 seconds.
2501 
In summary,
2502 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
2503 
performance is comparable to that of other UNIX editors, although it solves
2504 
a harder problem.
2505 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
2506 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2507 
<span style="font-size: 10pt"><b>Communications
2508 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2509 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2510 
<span style="font-size: 10pt">The discussion so far has described the implementation of the host part of
2511 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">;
2512 
the next few sections explain how a machine with mouse and bitmap display
2513 
can be engaged to improve interaction.
2514 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
2515 
is not the first editor to be written as two processes,<sup></sup></span><sup><span style="font-size: 6pt">16</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
2516 
but its implementation
2517 
has some unusual aspects.
2518 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2519 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2520 
<span style="font-size: 10pt">There are several ways
2521 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
2522 
host and terminal parts may be connected.
2523 
The first and simplest is to forgo the terminal part and use the host
2524 
part&rsquo;s command language to edit text on an ordinary terminal.
2525 
This mode is invoked by starting
2526 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2527 
with the
2528 
</span><span style="font-size: 10pt"><tt>-d</tt></span><span style="font-size: 10pt">
2529 
option.
2530 
With no options,
2531 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2532 
runs separate host and terminal programs,
2533 
communicating with a message protocol over the physical
2534 
connection that joins them.
2535 
Typically, the connection is an RS-232 link between a Blit
2536 
(the prototypical display for
2537 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">)
2538 
and a host running
2539 
the Ninth Edition of the UNIX operating system.<sup></sup></span><sup><span style="font-size: 6pt">8</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
2540 
(This is the version of the system used in the Computing Sciences Research
2541 
Center at AT&amp;T Bell Laboratories [now Lucent Technologies, Bell Labs], where I work.  Its relevant
2542 
aspects are discussed in the Blit paper.<sup></sup></span><sup><span style="font-size: 6pt">1</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">)
2543 
The implementation of
2544 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2545 
for the SUN computer runs both processes on the same machine and
2546 
connects them by a pipe.
2547 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2548 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2549 
<span style="font-size: 10pt">The low bandwidth of an RS-232 link
2550 
necessitated the split between
2551 
the two programs.
2552 
The division is a mixed blessing:
2553 
a program in two parts is much harder to write and to debug
2554 
than a self-contained one,
2555 
but the split makes several unusual configurations possible.
2556 
The terminal may be physically separated from the host, allowing the conveniences
2557 
of a mouse and bitmap display to be taken home while leaving the files at work.
2558 
It is also possible to run the host part on a remote machine:
2559 
</span></p><p style="margin-top: 0; margin-bottom: 0.08in"></p>
2560 
<p style="line-height: 1.1em; margin-left: 1.28in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2561 
<span style="font-size: 9pt"><tt>sam -r host</tt></span></p>
2562 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
2563 
 
2564 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
2565 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2566 
<span style="font-size: 10pt">connects to the terminal in the usual way, and then makes a call
2567 
across the network to establish the host part of
2568 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2569 
on the named machine.
2570 
Finally, it cross-connects the I/O to join the two parts.
2571 
This allows
2572 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2573 
to be run on machines that do not support bitmap displays;
2574 
for example,
2575 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2576 
is the editor of choice on our Cray X-MP/24.
2577 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
2578 
</span><span style="font-size: 10pt"><tt>-r</tt></span><span style="font-size: 10pt">
2579 
involves
2580 
</span><span style="font-size: 10pt"><i>three</i></span><span style="font-size: 10pt">
2581 
machines: the remote host, the terminal, and the local host.
2582 
The local host&rsquo;s job is simple but vital: it passes the data
2583 
between the remote host and terminal.
2584 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2585 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2586 
<span style="font-size: 10pt">The host and terminal exchange messages asynchronously
2587 
(rather than, say, as remote procedure calls) but there is no
2588 
error detection or correction
2589 
because, whatever the configuration, the connection is reliable.
2590 
Because the terminal handles mundane interaction tasks such as
2591 
popping up menus and interpreting the responses, the messages are about
2592 
data, not actions.
2593 
For example, the host knows nothing about what is displayed on the screen,
2594 
and when the user types a character, the message sent to the host says
2595 
&lsquo;&lsquo;insert a one-byte string at location 123 in file 7,&rsquo;&rsquo; not &lsquo;&lsquo;a character
2596 
was typed at the current position in the current file.&rsquo;&rsquo;
2597 
In other words, the messages look very much like the transaction records
2598 
in the transcripts.
2599 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2600 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2601 
<span style="font-size: 10pt">Either the host or terminal part of
2602 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2603 
may initiate a change to a file.
2604 
The command language operates on the host, while typing and some
2605 
mouse operations are executed directly in the terminal to optimize response.
2606 
Changes initiated by the host program must be transmitted to the terminal,
2607 
and
2608 
vice versa.
2609 
(A token is exchanged to determine which end is in control,
2610 
which means that characters typed while a time-consuming command runs
2611 
must be buffered and do not appear until the command is complete.)
2612 
To maintain consistent information,
2613 
the host and terminal track changes through a per-file
2614 
data structure that records what portions of the file
2615 
the terminal has received.
2616 
The data structure, called a
2617 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
2618 
(a weak pun: it&rsquo;s a file with holes)
2619 
is held and updated by both the host and terminal.
2620 
A
2621 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
2622 
is a list of
2623 
</span><span style="font-size: 10pt"><tt>Strings</tt></span><span style="font-size: 10pt">
2624 
holding those parts of the file known to the terminal,
2625 
separated by counts of the number of bytes in the interstices.
2626 
Of course, the host doesn&rsquo;t keep a separate copy of the data (it only needs
2627 
the lengths of the various pieces),
2628 
but the structure is the same on both ends.
2629 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2630 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2631 
<span style="font-size: 10pt">The
2632 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
2633 
in the terminal doubles as a cache.
2634 
Since the terminal keeps the text for portions of the file it has displayed,
2635 
it need not request data from the host when revisiting old parts of the file
2636 
or redrawing obscured windows, which speeds things up considerably
2637 
over low-speed links.
2638 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2639 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2640 
<span style="font-size: 10pt">It&rsquo;s trivial for the terminal to maintain its
2641 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">,
2642 
because all changes made on the terminal apply to parts of the file
2643 
already loaded there.
2644 
Changes made by the host are compared against the
2645 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
2646 
during the update sequence after each command.
2647 
Small changes to pieces of the file loaded in the terminal
2648 
are sent in their entirety.
2649 
Larger changes, and changes that fall entirely in the holes,
2650 
are transmitted as messages without literal data:
2651 
only the lengths of the deleted and inserted strings are transmitted.
2652 
When a command is completed, the terminal examines its visible
2653 
windows to see if any holes in their
2654 
</span><span style="font-size: 10pt"><tt>Rasps</tt></span><span style="font-size: 10pt">
2655 
intersect the visible portion of the file.
2656 
It then requests the missing data from the host,
2657 
along with up to 512 bytes of surrounding data, to minimize
2658 
the number of messages when visiting a new portion of the file.
2659 
This technique provides a kind of two-level lazy evaluation for the terminal.
2660 
The first level sends a minimum of information about
2661 
parts of the file not being edited interactively;
2662 
the second level waits until a change is displayed before
2663 
transmitting the new data.
2664 
Of course,
2665 
performance is also helped by having the terminal respond immediately to typing
2666 
and simple mouse requests.
2667 
Except for small changes to active pieces of the file, which are
2668 
transmitted to the terminal without negotiation,
2669 
the terminal is wholly responsible for deciding what is displayed;
2670 
the host uses the
2671 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
2672 
only to tell the terminal what might be relevant.
2673 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2674 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2675 
<span style="font-size: 10pt">When a change is initiated by the host,
2676 
the messages to the terminal describing the change
2677 
are generated by the routine that applies the transcript of the changes
2678 
to the contents of the
2679 
</span><span style="font-size: 10pt"><tt>File</tt></span><span style="font-size: 10pt">.
2680 
Since changes are undone by the same update routine,
2681 
undoing requires
2682 
no extra code in the communications;
2683 
the usual messages describing changes to the file are sufficient
2684 
to back up the screen image.
2685 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2686 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2687 
<span style="font-size: 10pt">The
2688 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
2689 
is a particularly good example of the way caches are used in
2690 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">.
2691 
First, it facilitates access to the active portion of the text by placing
2692 
the busy text in main memory.
2693 
In so doing, it provides efficient access
2694 
to a large data structure that does not fit in memory.
2695 
Since the form of data is to be imposed by the user, not by the program,
2696 
and because characters will frequently be scanned sequentially,
2697 
files are stored as flat objects.
2698 
Caches help keep performance good and linear when working with such
2699 
data.
2700 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2701 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2702 
<span style="font-size: 10pt">Second, the
2703 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
2704 
and several of the other caches have some
2705 
</span><span style="font-size: 10pt"><i>read-ahead;</i></span><span style="font-size: 10pt">
2706 
that is, the cache is loaded with more information than is needed for
2707 
the job immediately at hand.
2708 
When manipulating linear structures, the accesses are usually sequential,
2709 
and read-ahead can significantly reduce the average time to access the
2710 
next element of the object.
2711 
Sequential access is a common mode for people as well as programs;
2712 
consider scrolling through a document while looking for something.
2713 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2714 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2715 
<span style="font-size: 10pt">Finally, like any good data structure,
2716 
the cache guides the algorithm, or at least the implementation.
2717 
The
2718 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
2719 
was actually invented to control the communications between the host and
2720 
terminal parts, but I realized very early that it was also a form of
2721 
cache.  Other caches were more explicitly intended to serve a double
2722 
purpose: for example, the caches in
2723 
</span><span style="font-size: 10pt"><tt>Files</tt></span><span style="font-size: 10pt">
2724 
that coalesce updates not only reduce traffic to the
2725 
transcript and contents
2726 
</span><span style="font-size: 10pt"><tt>Buffers</tt></span><span style="font-size: 10pt">,
2727 
they also clump screen updates so that complicated changes to the
2728 
screen are achieved in
2729 
just a few messages to the terminal.
2730 
This saved me considerable work: I did not need to write special
2731 
code to optimize the message traffic to the
2732 
terminal.
2733 
Caches pay off in surprising ways.
2734 
Also, they tend to be independent, so their performance improvements
2735 
are multiplicative.
2736 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
2737 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2738 
<span style="font-size: 10pt"><b>Data structures in the terminal
2739 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2740 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2741 
<span style="font-size: 10pt">The terminal&rsquo;s job is to display and to maintain a consistent image of
2742 
pieces of the files being edited.
2743 
Because the text is always in memory, the data structures are
2744 
considerably simpler than those in the host part.
2745 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2746 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2747 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
2748 
typically has far more windows than does
2749 
</span><span style="font-size: 10pt"><tt>mux</tt></span><span style="font-size: 10pt">,
2750 
the window system within which its Blit implementation runs.
2751 
</span><span style="font-size: 10pt"><tt>Mux</tt></span><span style="font-size: 10pt">
2752 
has a fairly small number of asynchronously updated windows;
2753 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2754 
needs a large number of synchronously updated windows that are
2755 
usually static and often fully obscured.
2756 
The different tradeoffs guided
2757 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2758 
away from the memory-intensive implementation of windows, called
2759 
</span><span style="font-size: 10pt"><tt>Layers</tt></span><span style="font-size: 10pt">,<sup></sup></span><sup><span style="font-size: 6pt">17</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
2760 
used in
2761 
</span><span style="font-size: 10pt"><tt>mux.</tt></span><span style="font-size: 10pt">
2762 
Rather than depending on a complete bitmap image of the display for each window,
2763 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2764 
regenerates the image from its in-memory text
2765 
(stored in the
2766 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">)
2767 
when necessary, although it will use such an image if it is available.
2768 
Like
2769 
</span><span style="font-size: 10pt"><tt>Layers</tt></span><span style="font-size: 10pt">,
2770 
though,
2771 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2772 
uses the screen bitmap as active storage in which to update the image using
2773 
</span><span style="font-size: 10pt"><tt>bitblt</tt></span><span style="font-size: 10pt">.<sup></sup></span><sup><span style="font-size: 6pt">18,19</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
2774 
The resulting organization, pictured in Figure 6,
2775 
has a global array of windows, called
2776 
</span><span style="font-size: 10pt"><tt>Flayers</tt></span><span style="font-size: 10pt">,
2777 
each of which holds an image of a piece of text held in a data structure
2778 
called a
2779 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">,
2780 
which in turn represents
2781 
a rectangular window full of text displayed in some
2782 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">.
2783 
Each
2784 
</span><span style="font-size: 10pt"><tt>Flayer</tt></span><span style="font-size: 10pt">
2785 
appears in a global list that orders them all front-to-back
2786 
on the display, and simultaneously as an element of a per-file array
2787 
that holds all the open windows for that file.
2788 
The complement in the terminal of the
2789 
</span><span style="font-size: 10pt"><tt>File</tt></span><span style="font-size: 10pt">
2790 
on the host is called a
2791 
</span><span style="font-size: 10pt"><tt>Text</tt></span><span style="font-size: 10pt">;
2792 
each connects its
2793 
</span><span style="font-size: 10pt"><tt>Flayers</tt></span><span style="font-size: 10pt">
2794 
to the associated
2795 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">.
2796 
</span><span style="font-size: 10pt"></span></p><center><img src="sam3.png"></center>
2797 
</center>
2798 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2799 
<span style="font-size: 8pt"><i>Figure 6. Data structures in the terminal.
2800 
</i></span><span style="font-size: 8pt"><tt>Flayers</tt></span><span style="font-size: 8pt"><i>
2801 
are also linked together into a front-to-back list.
2802 
</i></span><span style="font-size: 8pt"><tt>Boxes</tt></span><span style="font-size: 8pt"><i>
2803 
are discussed in the next section.
2804 
</i></span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
2805 
<p style="margin-top: 0; margin-bottom: 0.02in"></p>
2806 
 
2807 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
2808 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2809 
<span style="font-size: 10pt">The
2810 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
2811 
for a
2812 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">
2813 
contains the image of the text.
2814 
For a fully visible window, the
2815 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
2816 
will be the screen (or at least the
2817 
</span><span style="font-size: 10pt"><tt>Layer</tt></span><span style="font-size: 10pt">
2818 
in which
2819 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2820 
is being run),
2821 
while for partially obscured windows the
2822 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
2823 
will be off-screen.
2824 
If the window is fully obscured, the
2825 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
2826 
will be null.
2827 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2828 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2829 
<span style="font-size: 10pt">The
2830 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
2831 
is a kind of cache.
2832 
When making changes to the display, most of the original image will
2833 
look the same in the final image, and the update algorithms exploit this.
2834 
The
2835 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">
2836 
software updates the image in the
2837 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
2838 
incrementally; the
2839 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
2840 
is not just an image, it is a data structure.<sup></sup></span><sup><span style="font-size: 6pt">18,19</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
2841 
The job of the software that updates the display is therefore
2842 
to use as much as possible of the existing image (converting the
2843 
text from ASCII characters to pixels is expensive) in a sort of two-dimensional
2844 
string insertion algorithm.
2845 
The details of this process are described in the next section.
2846 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2847 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2848 
<span style="font-size: 10pt">The
2849 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">
2850 
software has no code to support overlapping windows;
2851 
its job is to keep a single
2852 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
2853 
up to date.
2854 
It falls to the
2855 
</span><span style="font-size: 10pt"><tt>Flayer</tt></span><span style="font-size: 10pt">
2856 
software to multiplex the various
2857 
</span><span style="font-size: 10pt"><tt>Bitmaps</tt></span><span style="font-size: 10pt">
2858 
onto the screen.
2859 
The problem of maintaining overlapping
2860 
</span><span style="font-size: 10pt"><tt>Flayers</tt></span><span style="font-size: 10pt">
2861 
is easier than for
2862 
</span><span style="font-size: 10pt"><tt>Layers</tt></span><span style="font-size: 10pt"><sup></sup></span><sup><span style="font-size: 6pt">17</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt">
2863 
because changes are made synchronously and because the contents of the window
2864 
can be reconstructed from the data stored in the
2865 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">;
2866 
the
2867 
</span><span style="font-size: 10pt"><tt>Layers</tt></span><span style="font-size: 10pt">
2868 
software
2869 
makes no such assumptions.
2870 
In
2871 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">,
2872 
the window being changed is almost always fully visible, because the current
2873 
window is always fully visible, by construction.
2874 
However, when multi-file changes are being made, or when
2875 
more than one window is open on a file,
2876 
it may be necessary to update partially obscured windows.
2877 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2878 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2879 
<span style="font-size: 10pt">There are three cases: the window is
2880 
fully visible, invisible (fully obscured), or partially visible.
2881 
If fully visible, the
2882 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
2883 
is part of the screen, so when the
2884 
</span><span style="font-size: 10pt"><tt>Flayer</tt></span><span style="font-size: 10pt">
2885 
update routine calls the
2886 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">
2887 
update routine, the screen will be updated directly.
2888 
If the window is invisible,
2889 
there is no associated
2890 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">,
2891 
and all that is necessary is to update the
2892 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">
2893 
data structure, not the image.
2894 
If the window is partially visible, the
2895 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">
2896 
routine is called to update the image in the off-screen
2897 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">,
2898 
which may require regenerating it from the text of the window.
2899 
The
2900 
</span><span style="font-size: 10pt"><tt>Flayer</tt></span><span style="font-size: 10pt">
2901 
code then clips this
2902 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
2903 
against the
2904 
</span><span style="font-size: 10pt"><tt>Bitmaps</tt></span><span style="font-size: 10pt">
2905 
of all
2906 
</span><span style="font-size: 10pt"><tt>Frames</tt></span><span style="font-size: 10pt">
2907 
in front of the
2908 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">
2909 
being modified, and the remainder is copied to the display.
2910 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2911 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2912 
<span style="font-size: 10pt">This is much faster than recreating the image off-screen
2913 
for every change, or clipping all the changes made to the image
2914 
during its update.
2915 
Unfortunately, these caches can also consume prohibitive amounts of
2916 
memory, so they are freed fairly liberally &mdash; after every change to the
2917 
front-to-back order of the
2918 
</span><span style="font-size: 10pt"><tt>Flayers</tt></span><span style="font-size: 10pt">.
2919 
The result is that
2920 
the off-screen
2921 
</span><span style="font-size: 10pt"><tt>Bitmaps</tt></span><span style="font-size: 10pt">
2922 
exist only while multi-window changes are occurring,
2923 
which is the only time the performance improvement they provide is needed.
2924 
Also, the user interface causes fully-obscured windows to be the
2925 
easiest to make &mdash;
2926 
creating a canonically sized and placed window requires only a button click
2927 
&mdash; which reduces the need for caching still further.
2928 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2929 
<p style="margin-top: 0; margin-bottom: 0.17in"></p>
2930 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2931 
<span style="font-size: 10pt"><b>Screen update
2932 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2933 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2934 
<span style="font-size: 10pt">Only two low-level primitives are needed for incremental update:
2935 
</span><span style="font-size: 10pt"><tt>bitblt</tt></span><span style="font-size: 10pt">,
2936 
which copies rectangles of pixels, and
2937 
</span><span style="font-size: 10pt"><tt>string</tt></span><span style="font-size: 10pt">
2938 
(which in turn calls
2939 
</span><span style="font-size: 10pt"><tt>bitblt</tt></span><span style="font-size: 10pt">),
2940 
which draws a null-terminated character string in a
2941 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">.
2942 
A
2943 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">
2944 
contains a list of
2945 
</span><span style="font-size: 10pt"><tt>Boxes</tt></span><span style="font-size: 10pt">,
2946 
each of which defines a horizontal strip of text in the window
2947 
(see Figure 7).
2948 
A
2949 
</span><span style="font-size: 10pt"><tt>Box</tt></span><span style="font-size: 10pt">
2950 
has a character string
2951 
</span><span style="font-size: 10pt"><tt>str</tt></span><span style="font-size: 10pt">,
2952 
and a
2953 
</span><span style="font-size: 10pt"><tt>Rectangle</tt></span><span style="font-size: 10pt">
2954 
</span><span style="font-size: 10pt"><tt>rect</tt></span><span style="font-size: 10pt">
2955 
that defines the location of the strip in the window.
2956 
(The text in
2957 
</span><span style="font-size: 10pt"><tt>str</tt></span><span style="font-size: 10pt">
2958 
is stored in the
2959 
</span><span style="font-size: 10pt"><tt>Box</tt></span><span style="font-size: 10pt">
2960 
separately from the
2961 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
2962 
associated with the window&rsquo;s file, so
2963 
</span><span style="font-size: 10pt"><tt>Boxes</tt></span><span style="font-size: 10pt">
2964 
are self-contained.)
2965 
The invariant is that
2966 
the image of the
2967 
</span><span style="font-size: 10pt"><tt>Box</tt></span><span style="font-size: 10pt">
2968 
can be reproduced by calling
2969 
</span><span style="font-size: 10pt"><tt>string</tt></span><span style="font-size: 10pt">
2970 
with argument
2971 
</span><span style="font-size: 10pt"><tt>str</tt></span><span style="font-size: 10pt">
2972 
to draw the string in
2973 
</span><span style="font-size: 10pt"><tt>rect</tt></span><span style="font-size: 10pt">,
2974 
and the resulting picture fits perfectly within
2975 
</span><span style="font-size: 10pt"><tt>rect</tt></span><span style="font-size: 10pt">.
2976 
In other words, the
2977 
</span><span style="font-size: 10pt"><tt>Boxes</tt></span><span style="font-size: 10pt">
2978 
define the tiling of the window.
2979 
The tiling may be complicated by long lines of text, which
2980 
are folded onto the next line.
2981 
Some editors use horizontal scrolling to avoid this complication,
2982 
but to be comfortable this technique requires that lines not be
2983 
</span><span style="font-size: 10pt"><i>too</i></span><span style="font-size: 10pt">
2984 
long;
2985 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
2986 
has no such restriction.
2987 
Also, and perhaps more importantly, UNIX programs and terminals traditionally fold
2988 
long lines to make their contents fully visible.
2989 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
2990 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
2991 
<span style="font-size: 10pt">Two special kinds of
2992 
</span><span style="font-size: 10pt"><tt>Boxes</tt></span><span style="font-size: 10pt">
2993 
contain a single
2994 
character: either a newline or a tab.
2995 
Newlines and tabs are white space.
2996 
A newline
2997 
</span><span style="font-size: 10pt"><tt>Box</tt></span><span style="font-size: 10pt">
2998 
always extends to the right edge of the window,
2999 
forcing the following
3000 
</span><span style="font-size: 10pt"><tt>Box</tt></span><span style="font-size: 10pt">
3001 
to the next line.
3002 
The width of a tab depends on where it is located:
3003 
it forces the next
3004 
</span><span style="font-size: 10pt"><tt>Box</tt></span><span style="font-size: 10pt">
3005 
to begin at a tab location.
3006 
Tabs also
3007 
have a minimum width equivalent to a blank (blanks are
3008 
drawn by
3009 
</span><span style="font-size: 10pt"><tt>string</tt></span><span style="font-size: 10pt">
3010 
and are not treated specially); newlines have a minimum width of zero.
3011 
</span><span style="font-size: 10pt"></span></p><center><img src="sam4.png"></center>
3012 
</center>
3013 
<p style="margin-top: 0; margin-bottom: 0.08in"></p>
3014 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3015 
<span style="font-size: 8pt"><i>Figure 7. A line of text showing its
3016 
</i></span><span style="font-size: 8pt"><tt>Boxes</tt></span><span style="font-size: 8pt"><i>.
3017 
The first two blank
3018 
</i></span><span style="font-size: 8pt"><tt>Boxes</tt></span><span style="font-size: 8pt"><i>
3019 
contain tabs; the last contains a newline.
3020 
Spaces are handled as ordinary characters.
3021 
</i></span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
3022 
<p style="margin-top: 0; margin-bottom: 0.02in"></p>
3023 
 
3024 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
3025 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3026 
<span style="font-size: 10pt">The update algorithms always use the
3027 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">
3028 
image of the text (either the display or cache
3029 
</span><span style="font-size: 10pt"><tt>Bitmap</tt></span><span style="font-size: 10pt">);
3030 
they never examine the characters within a
3031 
</span><span style="font-size: 10pt"><tt>Box</tt></span><span style="font-size: 10pt">
3032 
except when the
3033 
</span><span style="font-size: 10pt"><tt>Box</tt></span><span style="font-size: 10pt">
3034 
needs to be split in two.
3035 
Before a change, the window consists of a tiling of
3036 
</span><span style="font-size: 10pt"><tt>Boxes</tt></span><span style="font-size: 10pt">;
3037 
after the change the window is tiled differently.
3038 
The update algorithms rearrange the tiles in place, without
3039 
backup storage.
3040 
The algorithms are not strictly optimal &mdash; for example, they can
3041 
clear a pixel that is later going to be written upon &mdash;
3042 
but they never move a tile that doesn&rsquo;t need to be moved,
3043 
and they move each tile at most once.
3044 
</span><span style="font-size: 10pt"><tt>Frinsert</tt></span><span style="font-size: 10pt">
3045 
on a Blit can absorb over a thousand characters a second if the strings
3046 
being inserted are a few tens of characters long.
3047 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3048 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3049 
<span style="font-size: 10pt">Consider
3050 
</span><span style="font-size: 10pt"><tt>frdelete</tt></span><span style="font-size: 10pt">.
3051 
Its job is to delete a substring from a
3052 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">
3053 
and restore the image of the
3054 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">.
3055 
The image of a substring has a peculiar shape (see Figure 2) comprising
3056 
possibly a partial line,
3057 
zero or more full lines,
3058 
and possibly a final partial line.
3059 
For reference, call this the
3060 
</span><span style="font-size: 10pt"><i>Z-shape.
3061 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Frdelete</tt></span><span style="font-size: 10pt">
3062 
begins by splitting, if necessary, the
3063 
</span><span style="font-size: 10pt"><tt>Boxes</tt></span><span style="font-size: 10pt">
3064 
containing the ends of
3065 
the substring so the substring begins and ends on
3066 
</span><span style="font-size: 10pt"><tt>Box</tt></span><span style="font-size: 10pt">
3067 
boundaries.
3068 
Because the substring is being deleted, its image is not needed,
3069 
so the Z-shape is then cleared.
3070 
Then, tiles (that is, the images of
3071 
</span><span style="font-size: 10pt"><tt>Boxes</tt></span><span style="font-size: 10pt">)
3072 
are copied, using
3073 
</span><span style="font-size: 10pt"><tt>bitblt</tt></span><span style="font-size: 10pt">,
3074 
from immediately after the Z-shape to
3075 
the beginning of the Z-shape,
3076 
resulting in a new Z-shape.
3077 
(</span><span style="font-size: 10pt"><tt>Boxes</tt></span><span style="font-size: 10pt">
3078 
whose contents would span two lines in the new position must first be split.)
3079 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3080 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3081 
<span style="font-size: 10pt">Copying the remainder of the
3082 
</span><span style="font-size: 10pt"><tt>Frame</tt></span><span style="font-size: 10pt">
3083 
tile by tile
3084 
this way will clearly accomplish the deletion but eventually,
3085 
typically when the copying algorithm encounters a tab or newline,
3086 
the old and new
3087 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
3088 
coordinates of the tile
3089 
to be copied are the same.
3090 
This correspondence implies
3091 
that the Z-shape has its beginning and ending edges aligned
3092 
vertically, and a sequence of at most two
3093 
</span><span style="font-size: 10pt"><tt>bitblts</tt></span><span style="font-size: 10pt">
3094 
can be used to copy the remaining tiles.
3095 
The last step is to clear out the resulting empty space at the bottom
3096 
of the window;
3097 
the number of lines to be cleared is the number of complete lines in the
3098 
Z-shape closed by the final
3099 
</span><span style="font-size: 10pt"><tt>bitblts.</tt></span><span style="font-size: 10pt">
3100 
The final step is to merge horizontally adjacent
3101 
</span><span style="font-size: 10pt"><tt>Boxes</tt></span><span style="font-size: 10pt">
3102 
of plain text.
3103 
The complete source to
3104 
</span><span style="font-size: 10pt"><tt>frdelete</tt></span><span style="font-size: 10pt">
3105 
is less than 100 lines of C.
3106 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3107 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3108 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>frinsert</tt></span><span style="font-size: 10pt">
3109 
is more complicated because it must do four passes:
3110 
one to construct the
3111 
</span><span style="font-size: 10pt"><tt>Box</tt></span><span style="font-size: 10pt">
3112 
list for the inserted string,
3113 
one to reconnoitre,
3114 
one to copy (in opposite order to
3115 
</span><span style="font-size: 10pt"><tt>frdelete</tt></span><span style="font-size: 10pt">)
3116 
the
3117 
</span><span style="font-size: 10pt"><tt>Boxes</tt></span><span style="font-size: 10pt">
3118 
to make the hole for the new text,
3119 
and finally one to copy the new text into place.
3120 
Overall, though,
3121 
</span><span style="font-size: 10pt"><tt>frinsert</tt></span><span style="font-size: 10pt">
3122 
has a similar flavor to
3123 
</span><span style="font-size: 10pt"><tt>frdelete</tt></span><span style="font-size: 10pt">,
3124 
and needn&rsquo;t be described further.
3125 
</span><span style="font-size: 10pt"><tt>Frinsert</tt></span><span style="font-size: 10pt">
3126 
and its subsidiary routines comprise 211 lines of C.
3127 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3128 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3129 
<span style="font-size: 10pt">The terminal source code is 3024 lines of C,
3130 
and the host source is 5797 lines.
3131 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
3132 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3133 
<span style="font-size: 10pt"><b>Discussion
3134 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
3135 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3136 
<span style="font-size: 10pt"><b>History
3137 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3138 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3139 
<span style="font-size: 10pt">The immediate ancestor of
3140 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3141 
was the original text editor for the Blit, called
3142 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">.
3143 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
3144 
inherited
3145 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">&rsquo;s
3146 
two-process structure and mouse language almost unchanged, but
3147 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
3148 
suffered from several drawbacks that were addressed in the design of
3149 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">.
3150 
The most important of these was the lack of a command language.
3151 
Although
3152 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
3153 
was easy to use for simple editing, it provided no direct help with
3154 
large or repetitive editing tasks.  Instead, it provided a command to pass
3155 
selected text through a shell pipeline,
3156 
but this was no more satisfactory than could be expected of a stopgap measure.
3157 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3158 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3159 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Jim</tt></span><span style="font-size: 10pt">
3160 
was written primarily as a vehicle for experimenting with a mouse-based
3161 
interface to text, and the experiment was successful.
3162 
</span><span style="font-size: 10pt"><tt>Jim</tt></span><span style="font-size: 10pt">
3163 
had some spin-offs:
3164 
</span><span style="font-size: 10pt"><tt>mux</tt></span><span style="font-size: 10pt">,
3165 
the second window system for the Blit, is essentially a multiplexed
3166 
version of the terminal part of
3167 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">;
3168 
and the debugger
3169 
</span><span style="font-size: 10pt"><tt>pi</tt></span><span style="font-size: 10pt">&rsquo;s
3170 
user interface<sup></sup></span><sup><span style="font-size: 6pt">20</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt"> was closely modeled on
3171 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">&rsquo;s.
3172 
But after a couple of years,
3173 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
3174 
had become difficult to maintain and limiting to use,
3175 
and its replacement was overdue.
3176 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3177 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3178 
<span style="font-size: 10pt">I began the design of
3179 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3180 
by asking
3181 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
3182 
customers what they wanted.
3183 
This was probably a mistake; the answers were essentially a list of features
3184 
to be found in other editors, which did not provide any of the
3185 
guiding principles I was seeking.
3186 
For instance, one common request was for a &lsquo;&lsquo;global substitute,&rsquo;&rsquo;
3187 
but no one suggested how to provide it within a cut-and-paste editor.
3188 
I was looking for a scheme that would
3189 
support such specialized features comfortably in the context of some
3190 
general command language.
3191 
Ideas were not forthcoming, though, particularly given my insistence
3192 
on removing all limits on file sizes, line lengths and so on.
3193 
Even worse, I recognized that, since the mouse could easily
3194 
indicate a region of the screen that was not an integral number of lines,
3195 
the command language would best forget about newlines altogether,
3196 
and that meant the command language had to treat the file as a single
3197 
string, not an array of lines.
3198 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3199 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3200 
<span style="font-size: 10pt">Eventually, I decided that thinking was not getting me very far and it was
3201 
time to try building.
3202 
I knew that the terminal part could be built easily &mdash;
3203 
that part of
3204 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
3205 
behaved acceptably well &mdash; and that most of the hard work was going
3206 
to be in the host part: the file interface, command interpreter and so on.
3207 
Moreover, I had some ideas about how the architecture of
3208 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
3209 
could be improved without destroying its basic structure, which I liked
3210 
in principle but which hadn&rsquo;t worked out as well as I had hoped.
3211 
So I began by designing the file data structure,
3212 
starting with the way
3213 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
3214 
worked &mdash; comparable to a single structure merging
3215 
</span><span style="font-size: 10pt"><tt>Disc</tt></span><span style="font-size: 10pt">
3216 
and
3217 
</span><span style="font-size: 10pt"><tt>Buffer</tt></span><span style="font-size: 10pt">,
3218 
which I split to make the cache more general
3219 
&mdash; and thinking about how global substitute could be implemented.
3220 
The answer was clearly that it had to be done in two passes,
3221 
and the transcript-oriented implementation fell out naturally.
3222 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3223 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3224 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
3225 
was written bottom-up,
3226 
starting from the data structures and algorithms for manipulating text,
3227 
through the command language and up to the code for maintaining
3228 
the display.
3229 
In retrospect, it turned out well, but this implementation method is
3230 
not recommended in general.
3231 
There were several times when I had a large body of interesting code
3232 
assembled and no clue how to proceed with it.
3233 
The command language, in particular, took almost a year to figure out,
3234 
but can be implemented (given what was there at the beginning of that year)
3235 
in a day or two.  Similarly, inventing the
3236 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
3237 
data structure delayed the
3238 
connection of the host and terminal pieces by another few months.
3239 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
3240 
took about two years to write, although only about four months were
3241 
spent actually working on it.
3242 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3243 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3244 
<span style="font-size: 10pt">Part of the design process was unusual:
3245 
the subset of the protocol that maintains the
3246 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
3247 
was simulated, debugged
3248 
and verified by an automatic protocol analyzer,<sup></sup></span><sup><span style="font-size: 6pt">21</span><span style="font-size: 10pt"></span></sup><span style="font-size: 10pt"> and was bug-free
3249 
from the start.
3250 
The rest of the protocol, concerned mostly
3251 
with keeping menus up to date,
3252 
was unfortunately too unwieldy for such analysis,
3253 
and was debugged by more traditional methods, primarily
3254 
by logging in a file all messages in and out of the host.
3255 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
3256 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3257 
<span style="font-size: 10pt"><b>Reflections
3258 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3259 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3260 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
3261 
is essentially the only interactive editor used by the sixty or so members of
3262 
the computing science research center in which I work.
3263 
The same could not be said of
3264 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">;
3265 
the lack of a command language kept some people from adopting it.
3266 
The union of a user interface as comfortable as
3267 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">&rsquo;s
3268 
with a command language as powerful as
3269 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">&rsquo;s&dagger;
3270 
</span></p><p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.50in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3271 
<span style="font-size: 10pt">is essential to
3272 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
3273 
success.
3274 
When
3275 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3276 
was first made available to the
3277 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
3278 
community,
3279 
almost everyone switched to it within two or three days.
3280 
In the months that followed, even people who had never adopted
3281 
</span><span style="font-size: 10pt"><tt>jim</tt></span><span style="font-size: 10pt">
3282 
started using
3283 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3284 
exclusively.
3285 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3286 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3287 
<span style="font-size: 10pt">To be honest,
3288 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">
3289 
still gets occasional use, but usually when
3290 
something quick needs to be done and the overhead of
3291 
downloading the terminal part of
3292 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3293 
isn&rsquo;t worth the trouble.
3294 
Also, as a &lsquo;line&rsquo; editor,
3295 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3296 
</span><span style="font-size: 10pt"><tt>-d</tt></span><span style="font-size: 10pt">
3297 
is a bit odd;
3298 
when using a good old ASCII terminal, it&rsquo;s comforting to have
3299 
a true line editor.
3300 
But it is fair to say that
3301 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
3302 
command language has displaced
3303 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">&rsquo;s
3304 
for most of the complicated editing that has kept line editors
3305 
(that is, command-driven editors) with us.
3306 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3307 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3308 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">&rsquo;s
3309 
command language is even fancier than
3310 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">&rsquo;s,
3311 
and most
3312 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3313 
customers don&rsquo;t come near to using all its capabilities.
3314 
Does it need to be so sophisticated?
3315 
I think the answer is yes, for two reasons.
3316 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3317 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3318 
<span style="font-size: 10pt">First, the
3319 
</span><span style="font-size: 10pt"><i>model</i></span><span style="font-size: 10pt">
3320 
for
3321 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
3322 
command language is really relatively simple, and certainly simpler than that of
3323 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">.
3324 
For instance, there is only one kind of textual loop in
3325 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3326 
&mdash; the
3327 
</span><span style="font-size: 10pt"><tt>x</tt></span><span style="font-size: 10pt">
3328 
command &mdash;
3329 
while
3330 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">
3331 
has three (the
3332 
</span><span style="font-size: 10pt"><tt>g</tt></span><span style="font-size: 10pt">
3333 
command, the global flag on substitutions, and the implicit loop over
3334 
lines in multi-line substitutions).
3335 
Also,
3336 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">&rsquo;s
3337 
substitute command is necessary to make changes within lines, but in
3338 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3339 
the
3340 
</span><span style="font-size: 10pt"><tt>s</tt></span><span style="font-size: 10pt">
3341 
command is more of a familiar convenience than a necessity;
3342 
</span><span style="font-size: 10pt"><tt>c</tt></span><span style="font-size: 10pt">
3343 
and
3344 
</span><span style="font-size: 10pt"><tt>t</tt></span><span style="font-size: 10pt">
3345 
can do all the work.
3346 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3347 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3348 
<span style="font-size: 10pt">Second,
3349 
given a community that expects an editor to be about as powerful as
3350 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">,
3351 
it&rsquo;s hard to see how
3352 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3353 
could really be much simpler and still satisfy that expectation.
3354 
People want to do &lsquo;&lsquo;global substitutes,&rsquo;&rsquo; and most are content
3355 
to have the recipe for that and a few other fancy changes.
3356 
The sophistication of the command language is really just a veneer
3357 
over a design that makes it possible to do global substitutes
3358 
in a screen editor.
3359 
Some people will always want something more, however, and it&rsquo;s gratifying to
3360 
be able to provide it.
3361 
The real power of
3362 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
3363 
command language comes from composability of the operators, which is by
3364 
nature orthogonal to the underlying model.
3365 
In other words,
3366 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3367 
is not itself complex, but it makes complex things possible.
3368 
If you don&rsquo;t want to do anything complex, you can ignore the
3369 
complexity altogether, and many people do so.
3370 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3371 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3372 
<span style="font-size: 10pt">Sometimes I am asked the opposite question: why didn&rsquo;t I just make
3373 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3374 
a real programmable editor, with macros and variables and so on?
3375 
The main reason is a matter of taste: I like the editor
3376 
to be the same every time I use it.
3377 
There is one technical reason, though:
3378 
programmability in editors is largely a workaround for insufficient
3379 
interactivity.
3380 
Programmable editors are used to make particular, usually short-term,
3381 
things easy to do, such as by providing shorthands for common actions.
3382 
If things are generally easy to do in the first place,
3383 
shorthands are not as helpful.
3384 
</span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
3385 
makes common editing operations very easy, and the solutions to
3386 
complex editing problems seem commensurate with the problems themselves.
3387 
Also, the ability to edit the
3388 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3389 
window makes it easy to repeat commands &mdash; it only takes a mouse button click
3390 
to execute a command again.
3391 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
3392 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3393 
<span style="font-size: 10pt"><b>Pros and cons
3394 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3395 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3396 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
3397 
has several other good points,
3398 
and its share of problems.
3399 
Among the good things is the idea of
3400 
structural regular expressions,
3401 
whose usefulness has only begun to be explored.
3402 
They were arrived at serendipitously when I attempted to distill the essence of
3403 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">&rsquo;s
3404 
way of doing global substitution and recognized that the looping command in
3405 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">
3406 
was implicitly imposing a structure (an array of lines) on the file.
3407 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3408 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3409 
<span style="font-size: 10pt">Another of
3410 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
3411 
good things is its undo capability.
3412 
I had never before used an editor with a true undo,
3413 
but I would never go back now.
3414 
Undo
3415 
</span><span style="font-size: 10pt"><i>must</i></span><span style="font-size: 10pt">
3416 
be done well, but if it is, it can be relied on.
3417 
For example,
3418 
it&rsquo;s safe to experiment if you&rsquo;re not sure how to write some intricate command,
3419 
because if you make a mistake, it can be fixed simply and reliably.
3420 
I learned two things about undo from writing
3421 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">:
3422 
first, it&rsquo;s easy to provide if you design it in from the beginning, and
3423 
second, it&rsquo;s necessary, particularly if the system has some subtle
3424 
properties that may be unfamiliar or error-prone for users.
3425 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3426 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3427 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">&rsquo;s
3428 
lack of internal limits and sizes is a virtue.
3429 
Because it avoids all fixed-size tables and data structures,
3430 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3431 
is able to make global changes to files that some of our other
3432 
tools cannot even read.
3433 
Moreover, the design keeps the performance linear when doing such
3434 
operations, although I must admit
3435 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3436 
does get slow when editing a huge file.
3437 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3438 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3439 
<span style="font-size: 10pt">Now, the problems.
3440 
Externally, the most obvious is that it is poorly integrated into the
3441 
surrounding window system.
3442 
By design, the user interface in
3443 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3444 
feels almost identical to that of
3445 
</span><span style="font-size: 10pt"><tt>mux</tt></span><span style="font-size: 10pt">,
3446 
but a thick wall separates text in
3447 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3448 
from the programs running in
3449 
</span><span style="font-size: 10pt"><tt>mux</tt></span><span style="font-size: 10pt">.
3450 
For instance, the &lsquo;snarf buffer&rsquo; in
3451 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3452 
must be maintained separately from that in
3453 
</span><span style="font-size: 10pt"><tt>mux</tt></span><span style="font-size: 10pt">.
3454 
This is regrettable, but probably necessary given the unusual configuration
3455 
of the system, with a programmable terminal on the far end of an RS-232 link.
3456 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3457 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3458 
<span style="font-size: 10pt"></span><span style="font-size: 10pt"><tt>Sam</tt></span><span style="font-size: 10pt">
3459 
is reliable; otherwise, people wouldn&rsquo;t use it.
3460 
But it was written over such a long time, and has so many new (to me)
3461 
ideas in it, that I would like to see it done over again to clean
3462 
up the code and remove many of the lingering problems in the implementation.
3463 
The worst part is in the interconnection of the host and terminal parts,
3464 
which might even be able to go away in a redesign for a more
3465 
conventional window system.
3466 
The program must be split in two to use the terminal effectively,
3467 
but the low bandwidth of the connection forces the separation to
3468 
occur in an inconvenient part of the design if performance is to be acceptable.
3469 
A simple remote procedure call
3470 
protocol driven by the host, emitting only graphics
3471 
commands, would be easy to write but wouldn&rsquo;t have nearly the
3472 
necessary responsiveness.  On the other hand, if the terminal were in control
3473 
and requested much simpler file services from the host, regular expression
3474 
searches would require that the terminal read the entire file over its RS-232
3475 
link, which would be unreasonably slow.
3476 
A compromise in which either end can take control is necessary.
3477 
In retrospect, the communications protocol should have been
3478 
designed and verified formally, although I do not know of any tool
3479 
that can adequately relate the protocol to
3480 
its implementation.
3481 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3482 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3483 
<span style="font-size: 10pt">Not all of
3484 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
3485 
users are comfortable with its command language, and few are adept.
3486 
Some (venerable) people use a sort of
3487 
&lsquo;&lsquo;</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">
3488 
subset&rsquo;&rsquo; of
3489 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
3490 
command language,
3491 
and even ask why
3492 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
3493 
command language is not exactly
3494 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">&rsquo;s.
3495 
(The reason, of course, is that
3496 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
3497 
model for text does not include newlines, which are central to
3498 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">.
3499 
Making the text an array of newlines to the command language would
3500 
be too much of a break from the seamless model provided by the mouse.
3501 
Some editors, such as
3502 
</span><span style="font-size: 10pt"><tt>vi</tt></span><span style="font-size: 10pt">,
3503 
are willing to make this break, though.)
3504 
The difficulty is that
3505 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">&rsquo;s
3506 
syntax is so close to
3507 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">&rsquo;s
3508 
that people believe it
3509 
</span><span style="font-size: 10pt"><i>should</i></span><span style="font-size: 10pt">
3510 
be the same.
3511 
I thought, with some justification in hindsight,
3512 
that making
3513 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3514 
similar to
3515 
</span><span style="font-size: 10pt"><tt>ed</tt></span><span style="font-size: 10pt">
3516 
would make it easier to learn and to accept.
3517 
But I may have overstepped and raised the users&rsquo;
3518 
expectations too much.
3519 
It&rsquo;s hard to decide which way to resolve this problem.
3520 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3521 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.35in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3522 
<span style="font-size: 10pt">Finally, there is a tradeoff in
3523 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3524 
that was decided by the environment in which it runs:
3525 
</span><span style="font-size: 10pt"><tt>sam</tt></span><span style="font-size: 10pt">
3526 
is a multi-file editor, although in a different system there might instead be
3527 
multiple single-file editors.
3528 
The decision was made primarily because starting a new program in a Blit is
3529 
time-consuming.
3530 
If the choice could be made freely, however, I would
3531 
still choose the multi-file architecture, because it allows
3532 
groups of files to be handled as a unit;
3533 
the usefulness of the multi-file commands is incontrovertible.
3534 
It is delightful to have the source to an entire program
3535 
available at your fingertips.
3536 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
3537 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3538 
<span style="font-size: 10pt"><b>Acknowledgements
3539 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3540 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3541 
<span style="font-size: 10pt">Tom Cargill suggested the idea behind the
3542 
</span><span style="font-size: 10pt"><tt>Rasp</tt></span><span style="font-size: 10pt">
3543 
data structure.
3544 
Norman Wilson and Ken Thompson influenced the command language.
3545 
This paper was improved by comments from
3546 
Al Aho,
3547 
Jon Bentley,
3548 
Chris Fraser,
3549 
Gerard Holzmann,
3550 
Brian Kernighan,
3551 
Ted Kowalski,
3552 
Doug McIlroy
3553 
and
3554 
Dennis Ritchie.
3555 
</span></p><p style="margin-top: 0; margin-bottom: 0.17in"></p>
3556 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3557 
<span style="font-size: 10pt"><b>REFERENCES
3558 
</b></span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3559 
<p style="margin-top: 0; margin-bottom: 0.05in"></p>
3560 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3561 
<span style="font-size: 10pt"> 1.    R. Pike,
3562 
&lsquo;The Blit: a multiplexed graphics terminal,&rsquo;
3563 
</span><span style="font-size: 10pt"><i>AT&amp;T Bell Labs. Tech. J.,
3564 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><b>63</b></span><span style="font-size: 10pt">,
3565 
(8),
3566 
1607-1631 (1984).
3567 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3568 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3569 
<span style="font-size: 10pt"> 2.    L. Johnson,
3570 
</span><span style="font-size: 10pt"><i>MacWrite,</i></span><span style="font-size: 10pt">
3571 
Apple Computer Inc., Cupertino, Calif. 1983.
3572 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3573 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3574 
<span style="font-size: 10pt"> 3.    B. Lampson,
3575 
&lsquo;Bravo Manual,&rsquo;
3576 
in
3577 
</span><span style="font-size: 10pt"><i>Alto User&rsquo;s Handbook,
3578 
</i></span><span style="font-size: 10pt">pp. 31-62,
3579 
Xerox Palo Alto Research Center,
3580 
Palo Alto, Calif.
3581 
1979.
3582 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3583 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3584 
<span style="font-size: 10pt"> 4.    W. Teitelman,
3585 
&lsquo;A tour through Cedar,&rsquo;
3586 
</span><span style="font-size: 10pt"><i>IEEE Software,
3587 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><b>1</b></span><span style="font-size: 10pt">
3588 
(2), 44-73 (1984).
3589 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3590 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3591 
<span style="font-size: 10pt"> 5.    J. Gutknecht,
3592 
&lsquo;Concepts of the text editor Lara,&rsquo;
3593 
</span><span style="font-size: 10pt"><i>Comm. ACM,
3594 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><b>28</b></span><span style="font-size: 10pt">,
3595 
(9),
3596 
942-960 (1985).
3597 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3598 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3599 
<span style="font-size: 10pt"> 6.    Bell Telephone Laboratories,
3600 
</span><span style="font-size: 10pt"><i>UNIX Programmer&rsquo;s Manual,
3601 
</i></span><span style="font-size: 10pt">Holt, Rinehart and Winston, New York 1983.
3602 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3603 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3604 
<span style="font-size: 10pt"> 7.    B. W. Kernighan and R. Pike,
3605 
</span><span style="font-size: 10pt"><i>The Unix Programming Environment,
3606 
</i></span><span style="font-size: 10pt">Prentice-Hall, Englewood Cliffs, New Jersey 1984.
3607 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3608 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3609 
<span style="font-size: 10pt"> 8.    </span><span style="font-size: 10pt"><i>Unix Time-Sharing System Programmer&rsquo;s Manual, Research Version, Ninth Edition,
3610 
Volume 1,
3611 
</i></span><span style="font-size: 10pt">AT&amp;T Bell Laboratories, Murray Hill, New Jersey 1986.
3612 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3613 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3614 
<span style="font-size: 10pt"> 9.    </span><span style="font-size: 10pt"><i>Unix Time-Sharing System Programmer&rsquo;s Manual, 4.1 Berkeley Software Distribution,
3615 
Volumes 1 and 2C,
3616 
</i></span><span style="font-size: 10pt">University of California, Berkeley, Calif. 1981.
3617 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3618 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3619 
<span style="font-size: 10pt">10.    R. Pike,
3620 
&lsquo;Structural Regular Expressions,&rsquo;
3621 
</span><span style="font-size: 10pt"><i>Proc. EUUG Spring Conf., Helsinki 1987,
3622 
</i></span><span style="font-size: 10pt">Eur. Unix User&rsquo;s Group, Buntingford, Herts, UK 1987.
3623 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3624 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3625 
<span style="font-size: 10pt">11.    A. Goldberg,
3626 
</span><span style="font-size: 10pt"><i>Smalltalk-80 &ndash; The Interactive Programming Environment,
3627 
</i></span><span style="font-size: 10pt">Addison-Wesley, Reading, Mass. 1984.
3628 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3629 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3630 
<span style="font-size: 10pt">12.    K. Thompson,
3631 
&lsquo;Regular expression search algorithm,&rsquo;
3632 
</span><span style="font-size: 10pt"><i>Comm. ACM,
3633 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><b>11</b></span><span style="font-size: 10pt">,
3634 
(6),
3635 
419-422 (1968).
3636 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3637 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3638 
<span style="font-size: 10pt">13.    A. V. Aho, J. E. Hopcroft and J. D. Ullman,
3639 
</span><span style="font-size: 10pt"><i>The Design and Analysis of Computer Algorithms,
3640 
</i></span><span style="font-size: 10pt">Addison-Wesley, Reading, Mass. 1974.
3641 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3642 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3643 
<span style="font-size: 10pt">14.    B. W. Kernighan and D. M. Ritchie,
3644 
</span><span style="font-size: 10pt"><i>The C Programming Language,
3645 
</i></span><span style="font-size: 10pt">Prentice-Hall, Englewood Cliffs, New Jersey 1978.
3646 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3647 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3648 
<span style="font-size: 10pt">15.    W. M. Waite,
3649 
&lsquo;The cost of lexical analysis,&rsquo;
3650 
</span><span style="font-size: 10pt"><i>Softw. Pract. Exp.,
3651 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><b>16</b></span><span style="font-size: 10pt">,
3652 
(5),
3653 
473-488 (1986).
3654 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3655 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3656 
<span style="font-size: 10pt">16.    C. W. Fraser,
3657 
&lsquo;A generalized text editor,&rsquo;
3658 
</span><span style="font-size: 10pt"><i>Comm. ACM,
3659 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><b>23</b></span><span style="font-size: 10pt">,
3660 
(3),
3661 
154-158 (1980).
3662 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3663 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3664 
<span style="font-size: 10pt">17.    R. Pike,
3665 
&lsquo;Graphics in overlapping bitmap layers,&rsquo;
3666 
</span><span style="font-size: 10pt"><i>ACM Trans. on Graph.,
3667 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><b>2</b></span><span style="font-size: 10pt">,
3668 
(2)
3669 
135-160 (1983).
3670 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3671 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3672 
<span style="font-size: 10pt">18.    L. J. Guibas and J. Stolfi,
3673 
&lsquo;A language for bitmap manipulation,&rsquo;
3674 
</span><span style="font-size: 10pt"><i>ACM Trans. on Graph.,
3675 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><b>1</b></span><span style="font-size: 10pt">,
3676 
(3),
3677 
191-214 (1982).
3678 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3679 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3680 
<span style="font-size: 10pt">19.    R. Pike, B. Locanthi and J. Reiser,
3681 
&lsquo;Hardware/software trade-offs for bitmap graphics on the Blit,&rsquo;
3682 
</span><span style="font-size: 10pt"><i>Softw. Pract. Exp.,
3683 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><b>15</b></span><span style="font-size: 10pt">,
3684 
(2),
3685 
131-151 (1985).
3686 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3687 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3688 
<span style="font-size: 10pt">20.    T. A. Cargill,
3689 
&lsquo;The feel of Pi,&rsquo;
3690 
</span><span style="font-size: 10pt"><i>Winter USENIX Conference Proceedings,
3691 
Denver 1986,
3692 
</i></span><span style="font-size: 10pt">62-71,
3693 
USENIX Assoc., El Cerrito, CA.
3694 
</span></p><p style="margin-top: 0; margin-bottom: 0.05in"></p>
3695 
<p style="line-height: 1.2em; margin-left: 1.00in; text-indent: 0.00in; margin-right: 1.00in; margin-top: 0; margin-bottom: 0; text-align: justify;">
3696 
<span style="font-size: 10pt">21.    G. J. Holzmann,
3697 
&lsquo;Tracing protocols,&rsquo;
3698 
</span><span style="font-size: 10pt"><i>AT&amp;T Tech. J.,
3699 
</i></span><span style="font-size: 10pt"></span><span style="font-size: 10pt"><b>64</b></span><span style="font-size: 10pt">,
3700 
(10),
3701 
2413-2434 (1985).
3702 
</span></p><p style="margin-top: 0; margin-bottom: 0.50in"></p>
3703 
</body>
3704 
</html>
3705