File: README.html

package info (click to toggle)
elvis 2.1.4-1
links: PTS
area: main
in suites: potato
size: 4,528 kB
ctags: 6,177
sloc: ansic: 57,188; sh: 1,026; makefile: 299
file content (1026 lines) | stat: -rw-r--r-- 40,974 bytes
parent folder | download | duplicates (2)
<html><head>
<title>README.html file for elvis 2.1</title>
<meta name="description" content="Quick intro to elvis 2.1, with links to source code and binaries">
<meta name="keywords" content="vi, elvis, unix, linux, msdos, win32, windows95 windowsNT, nt, os2, os/2, editor, source">
</head><body>

<center>
<h1>Quick intro to elvis 2.1, with links to source code and binaries</h1>
</center>

<strong>CONTENTS</strong>
<menu>
<li><a HREF="#thisfile">1. About this file</a>
<li><a HREF="#vi">2. Differences between vi and elvis 2.1</a>
<li><a HREF="#new">3. Differences between 2.0 and 2.1</a>
	<menu>
	<li><a HREF="#ex">3.1 New ex commands</a>
	<li><a HREF="#options">3.2 New options commands</a>
	<li><a HREF="#functions">3.3 New functions</a>
	<li><a HREF="#gui">3.4 New ports and GUI features</a>
	<li><a HREF="#misc">3.5 Miscellany</a>
	<li><a HREF="#revision3">3.6 Differences between 2.1_3 and the original 2.1 release</a>
	<li><a HREF="#revision4">3.7 Differences between 2.1_4 and 2.1_3</a>
	</menu>
<li><a HREF="#future">4. The future of elvis</a>
<li><a HREF="#links">5. Links to related files</a>
</menu>
<pre graphic>
.---------------------------------------------------------------.
| <strong>This is not elvis' user manual!</strong>  The <em>real</em> documentation for   |
| elvis is located in its online help facility.  While running  |
| elvis, enter the command "<code>:help</code>" to see the table of contents.|
^---------------------------------------------------------------^

<h1><a NAME="thisfile"></a>1. About this file</h1>

This file is written in the HTML markup language.
You can view it with any WWW viewer, such as Netscape.
You can also use elvis 2.1 to view it; this version of elvis
has the ability to view HTML documents, and print them.

<p>This file has many hypertext links.
Use them!
If you're using elvis 2.1 to browse this file, then hypertextual
references will appear as underlined text.
(Except on color PCs;
since color video cards don't support underlining,
hypertextual references will be colored -- white on red, by default.)
To follow the hypertext link, move the cursor onto the underlined
text and press (Enter).
To go back, press (Control-T).
The (Tab) key moves the cursor forward to the next hypertext reference.

<p>If elvis 2.1 doesn't automatically start up in HTML mode when you view this file,
then you'll need to force it into HTML mode by giving the command
"<code>:display html</code>".

<h1><a name="vi"></a>2. Differences between vi and elvis 2.1</h1>
Elvis is a superset of vi.
It runs on more operating systems than vi,
it is free,
and you can obtain the source code.
Elvis also has many new features.
These new features are described in the first chapter of the online manual,
which hypertext links to the other parts of the manual where those features
are described in detail.
Here's a just brief list:

<p>
<ul>

<li>Multiple edit buffers, so you can edit several files at once.
<li>Multiple windows, so you can see multiple edit buffers, or different
	parts of the same edit buffer.
<li>Multiple user interfaces, including graphical interfaces under
	Windows95/98/NT and X11.
<li>A variety of display modes, including syntax coloring and HTML.
<li>Online help, with hypertext links.
<li>Enhanced tags, to support overloading in C++.
<li>Network support, so you can load/save files via FTP, or even use elvis 		as a light-weight Web browser.
<li>Aliases, which allow you to define new ex commands.
<li>Built-in calculator

</ul>

<h1><a name="new"></a>3. Differences between 2.0 and 2.1</h1>

The following is a summary of changes made since the release of elvis 2.0.
These are in addition to bug changes.

<h2>3.1 <a name="ex">New ex commands</a></h2>
<dl>

<dt>:only
<dd>This closes all windows except the current window.
<pre>
	:only
</pre>

<dt>:browse
<br>:sbrowse
<dd>These build a list of all tags which match given criteria (e.g., all tags
in a particular file), and make an HTML document with links to all the
definition points in your source code.
The <code>:browse</code> command saves the cursor position on the stack and
then displays the table in the current window.
The <code>:sbrowse</code> command creates a new window showing that table.
<pre>
	:browse foo.c
	:sbrowse class:/Ball
</pre>

<dt>:bbrowse
<br>:sbbrowse
<dd>These build an HTML document with links to all the edit buffers.
The <code>:bbrowse</code> command saves the cursor position on the stack and
then displays the table in the current window.
The <code>:sbbrowse</code> command creates a new window showing that table.
<pre>
	:bbrowse
</pre>

<dt>:alias
<br>:unalias
<dd>The <code>:alias</code> command creates an alias -- a new ex command which
you can implement via a series of existing ex commands.
Aliases can accept arguments, and perform complex operations.
Many sample aliases are included in the distribution.
<pre>
	:alias lower !% s/.*/\L&amp;/
</pre>

<dt>:message
<br>:warning
<br>:error
<dd>These output messages of various types.
In particular, the <code>:error</code> message type has the side-effect of
aborting any pending macros or aliases.
<pre>
	:if !&lt; == !&gt;
	:then error Address range required
</pre>

<dt>:local
<dd>This is handy in aliases.
It is like <code>:set</code>, except that the options will be restored
to their previous value when the alias exits.
<pre>
	:alias total {
		"Sum the integers in a range of lines
		local t=0 report=0 nosaveregexp
		!% s/[0-9]\+/let t=t+\1/x
		eval !% c (t)
	}
</pre>

<dt>:try
<dd>This executes a command, and then checks the success/failure indicator
returned by that command.
The then/else flag is set accordingly.
Regardless of whether the command succeeded or not, <code>:try</code> itself
always succeeds;
this is significant because when a command fails
(anywhere except as the argument of a <code>:try</code> command),
any pending macros or aliases are aborted.
<pre>
	:alias togglecase {
		try !% s/.*[a-z].*/\U&amp;
		else !% s/.*/\L&amp;
	}
</pre>

<dt>:while
<br>:do
<dd>This executes a command in a loop, as long as a given condition holds true.
<pre>
	:let i = 1
	:while i &lt;= 10
	:do {
		calc i
		let i = i + 1
	}
</pre>

<dt>:switch
<br>:case
<br>:default
<dd>This is a multi-way branch.
The <code>:switch</code> command evaluates an expression, and stores the
result in an internal variable.
Each <code>:case</code> command compares the switch value to a literal string;
if it matches then the remainder of the line is executed an an ex command.
If none of the cases match, then <code>:default</code> will run its ex command.
<pre>
	:switch os
	:case unix echo LF
	:case mac echo CR
	:default echo CR-LF
</pre>

</dl>

<h2><a name="options">3.2 Options</a></h2>
Many new options have been added.
Some exist only in a particular user interface, though.
These are all described below.

<dl>

<dt>bufid, bufferid
<dd>Each user buffer is automatically assigned a numeric identifier,
which is stored in this option.
These can be used later as a shorthand for either the buffer's filename or
buffername.
<pre>
	:e #1
	:1,20 copy (1)$
</pre>

<dt>locked, lock
<dd>When set, this option prevents any changes from being made to the buffer.
This is a slightly stronger version of the "readonly" option.

<dt>readeol, reol
<br>writeeol, weol
<dd>Together, these two options replace the old "binary" option.
<code>readeol</code> can be set to one of <strong>dos</strong>,
<strong>unix</strong>, <strong>mac</strong>, <strong>text</strong>, or
<strong>binary</strong> to indicate the newline translations that were
used to read a file into a buffer.
The <code>writeeol</code> option can be set to any of those values, or
the special value <strong>same</strong> to use the same translation as
each buffer was read with.

<dt>partiallastline, pll
<dd>Traditionally, when vi is invoked on a file which doesn't end with a
newline, vi adds one.
Elvis does that too now, but since elvis can also be used on binary files
it needs to remember whether it has added a newline, so it can omit that
bogus newline when doing a binary write.
That's what this option is for.

<dt>putstyle, ps
<dd>This is only meaningful for cut buffers, not normal edit buffers.
It indicates whether the cut buffer's contents should be pasted as
a series of characters, as a rectangular block, or as whole lines.
(Previously this information was stored in the first line of a cut buffer,
but that made cut buffers hard to edit.)

<dt>matchchar, mc
<dd>The visual <code>%</code> command has been extended to match any
pairs of characers.
This option stores the list of character pairs.

<dt>completebinary, cob
<dd>Elvis performs file name completion -- you can type a partial file name
and hit &lt;Tab&gt; to have elvis complete the name for you (or as much of
the name as possible before ambiguities are encountered).
Normally elvis ignores binary files when looking for possible completions;
it only considers the names of text files.
Setting this option makes it check the names of binary files as well as
text files.

<dt>showname, snm
<br>showtag, st
<dd>These cause elvis to display extra information at the bottom of the
window, whenever it has nothing better to show there.
<code>showname</code> causes it to display the name of the current file.
<code>showtag</code> causes it to display the name of the tag which is
defined on the current line, or the nearest line above it which has a tag.
The <code>showtag</code> option can make elvis take more time to load files,
since it must find the location of every tag for that file.

<dt>true, True
<br>false, False
<dd>These options mostly exist so you can use the symbols <code>true</code>
and <code>false</code> in the expressions used by <code>:if</code> and the like.
The value of the <code>false</code> option is used as a logical false value,
just like "", "0", or "false"; anything else is considered to be a logical
true value.

<p>Any boolean option returns the value of the <code>true</code> option if it
is set, or the <code>false</code> option if it is reset.
Non-english speakers may wish to set <code>true</code> and <code>false</code>
to the local words for those logical states.
You can do that by setting those options explicitly, or by adding lines to
the "elvis.msg" file (e.g., "true:verdad" and "false:falso").

<dt>submit, Submit
<br>cancel, Cancel
<br>help, Help
<dd>The "x11" user interface uses these like the <code>true</code> and
<code>false</code> options, to translate a few words that it uses.
Actually, <code>help</code> doesn't do anything yet, but the values of
<code>submit</code> and <code>cancel</code> are used as the labels of
buttons on dialog windows.

<dt>scrollbar, sb
<br>toolbar, tb
<br>statusbar, stb
<br>menubar, mb
<dd>These enable or disable certain parts of the a GUI window.
They are supported by the "windows" GUI (for Windows95/98/NT) and the
"x11" GUI (for Unix).
Actually <code>menubar</code> isn't support by "x11" because "x11" doesn't
have a menu bar yet.

<dt>font, fnt
<br>normalstyle, nfn
<br>boldstyle, bfn
<br>italicstyle, ifn
<br>fixedstyle, ffn
<br>emphasizedstyle, efn
<br>underlinedstyle, ufn
<dd>These control the appearance of text in the "windows" interface (for
Windows95/98/NT).
<code>font</code> stores the base font, and the other options tell elvis
how to derive a display font from the base font.

<dt>underline, uln
<br>scrollbarleft, xsl
<br>borderwidth, xbw
<br>textcursor, tc
<br>outlinemono, om
<dd>These are all "x11" options.

<dt>saveregexp, sre
<dd>Many scripts and aliases reset this.
Normally when you perform a search or substitution, elvis remembers
the regular expression you used so you can perform the same search/substitution
again later without retyping the whole regular expression.
That's handy.
But many aliases and scripts perform their own internal searches and
substitutions, and those shouldn't cause elvis to forget the your
regular expression.
So aliases often do <code>:local nosre</code> to protect your regular
expression.

<dt>tagprg, tp
<dd>This allows you to use an external program to perform tag searches.

<dt>blkhit, bh
<br>blkmiss, bm
<br>blkwrite, bw
<dd>These provide extra information about elvis' internal block cache.

<dt>program, argv0
<dd>This stores the name by which elvis was invoked.
The default "elvis.ini" file uses this information to distinguish between
"ex" behavior and "view" behavior from the normal "vi" behavior.

<dt>lpnumber, lpn
<br>lpheader, lph
<dd>These affect the way text is printed.
Setting <code>lpnumber</code> causes line numbers to be printed, just as
<code>number</code> causes line number to be shown on the screen.
Setting the <code>lpheader</code> option
causes pages to have a header showing the file name, page number, and date.

<dt>previousdir, pdir
<dd>This stores the previous directory.
You can use "~-" (tilde minus) at the start of a file name as a shorthand
for the previous directory name.

<dt>smarttab, sta
<dd>If set, this causes the <kbd>&lt;Tab&gt;</kbd> key to act like
<kbd>^T</kbd> when in the indentation area of a line.
Elsewhere in a line, <kbd>&lt;Tab&gt;</kbd> will act normally.

</dl>

<h2><a name="functions">3.3 New functions</a></h2>
The following functions have been added to the built-in calculator.
This is used by <code>:if</code>, <code>:let</code>, and others.

<dl>

<dt>htmlsafe(<var>string</var>)
<dd>This replaces the &amp;, &lt;, &gt;, and &quot; characters with
<strong>&amp;amp;</strong>,
<strong>&amp;lt;</strong>,
<strong>&amp;gt;</strong>, and
<strong>&amp;quot;</strong>, respectively.

<dt>quote(<var>chars</var>, <var>str</var>)
<br>unquote(<var>chars</var>, <var>str</var>)
<dd>These add or remove backslashes.
<code>quote()</code> adds a backslash before each existing backslash, and
before each character listed in the <var>chars</var> string.
<code>unquote()</code> does the opposite.

<dt>current(<var>what</var>)
<dd>This examines elvis' internal variables.
It can be any of the following:
<pre>
	<strong>line</strong>	cursor's current line number
	<strong>column</strong>	cursor's current column number
	<strong>word</strong>	word at the cursor position
	<strong>mode</strong>	mode name as reported by showmode option
	<strong>next</strong>	next file in args list
	<strong>prev</strong>	previous file in args list
	<strong>tag</strong>	current tag as reported by showtag option
</pre>

<dt>fileeol(<var>file</var>)
<dd>Examine the contents of a file, and make a guess about what value
the <code>readeol</code> option should be set to.

<dt>getcwd()
<dd>Return the name of the current directory.

<dt>absolute(<var>file</var>)
<dd>Return the fully-specified name of a given file.
In other words, if it is in the current directory then combine the current
directory name with the file name.

<dt>alias(<var>name</var>)
<dd>Test for the existence of a given alias.

<dt>shell(<var>command</var>)
<dd>Executes the <var>command</var> via a shell, and returns the text that
the program wrote to stdout.

<dt>line(<var>buffer</var>, <var>line</var>)
<dd>Returns a single line from a buffer.
If <var>buffer</var> is omitted, then the current buffer is assumed.
If both <var>buffer</var> and <var>line</var> are omitted, then the current
line of the current buffer is returned.

</dl>

<h2><a name="gui">3.4 New ports, and new GUI features</a></h2>
<dl>

<dt>Graphical Win32 port
<dd>In addition to a text-mode version, there is now a fully graphical version
of elvis for Windows95/98/NT, named WinElvis.
It is included in the Win32 binary distribution.

<dt>OS/2 port
<dd>A text-mode port of elvis for OS/2 has been completed.

<dt>X11 improvements
<dd>The "x11" user interface has been improved through the addition of a
status bar and a configurable tool bar.
Tool bar buttons can even have configurable dialog windows.

</dl>

<h2><a name="misc">3.5 Miscellany</a></h2>

<dl>

<dt>Name completion
<dd>The &lt;Tab&gt; key is used for name completion on ex command lines.
It can complete file names, ex command names, option names, and tag names.
You can also make it perform tag name completion in your own files by
running "<code>:set&nbsp;inputtab=identifier</code>".

<dt>Tilde in a file name
<dd>Elvis 2.0 allowed you to use "~" at the start of a file name to access
files in your home directory.
2.1 adds the ability to use "~+" for the current directory's absolute path name,
and "~-" for the previous directory.
Also, under Unix you can use "~<var>user</var>" for the home directory of
<var>user</var>.

<dt>Network protocols
<dd>Elvis can read via HTTP, and read or write via FTP.
Just use a Web URL in place of a file name, like this:
<pre>
	:e ftp://ftp.cs.pdx.edu/pub/elvis/README.html</pre>
<p>Together with elvis' "html" display mode, this allows elvis to be used
as a light-weight Web browser.

<dt>"tex" display mode
<dd>A "tex" display mode has been implemented.
It handles the simplest TeX documents moderately well.
Unlike real TeX formatters, though, elvis' "tex" mode is not programmable
so it doesn't handle custom commands.

<dt>Enhanced tags
<dd>Elvis' tags implementation has been enhanced to support overloaded tags
as required for C++.
There is a whole chapter in the manual describing all the wonderful things
you can do with tags now.

<dt>Multi-line ex clauses
<dd>Some ex commands accept another ex command as an argument.
The <code>:g</code> command is an example of this.
Traditionally, vi has allowed multiple commands by placing a backslash before
the end of each line except the last, but this didn't always work.
Elvis supports that, and adds a better alternative:
If you give a '{' in place of the command, then elvis will read the following
lines up to the next matching '}', and use the intervening lines as the ex
command.
This is both more reliable and more readable than the backslash notation.

<dt>Confirming substitutions
<dd>Elvis 2.0's implementation of the <code>:s///</code> command lacked
support for the <code>c</code> flag.
Elvis 2.1 supports <code>:s///c</code>, at long last!

</dl>

<h2><a name="revision3"></a>3.6 Differences between 2.1_3 and the original 2.1 release</h2>

Many bugs and annoyances have been fixed since the first 2.1 release
a couple months ago.
The highlights are:
<dl>

<dt>Portability
<dd>The "configure" script has been tweaked to support some Unix variants better.
In particular, you should now be able to compile elvis under Linux systems that
use libc-6, such as Red Hat, without doing anything except "configure;make".
The same goes for SCO Unix.

<p>Elvis' version of ctags will now be installed as "<strong>elvtags</strong>"
on all Unix systems, including Linux.
Previously for Linux it was installed as "ctags", which made sense because
originally most versions of Linux used elvis 1.8's "ctags", but now most
Linux systems ship with Exuberant Ctags as their default ctags.
Previously, other versions of Unix skipped elvis' ctags
because they already had their own.
The new name for elvis' ctags should allow you to keep the old ctags and
still have access to elvis' ctags when you want it.
(Elvis' version can generate the "ln" attribute, which makes the
<code>showtag</code> option load files faster, but Exuberant Ctags is
better in just about every other way.)

<p>Elvis' "fmt" program is no longer included in the default program list
for any version of Unix.
Previously it was installed under Linux, and skipped for other versions of Unix.

<dt>Vi commands
<dd>The [[ and ]] commands were acting like line-mode commands; i.e.,
even with the cursor in the middle of a line, an operator such as d]] would
delete whole lines.
The real vi doesn't behave that way, so this was a bug.
It has been fixed.

<p>In fixing that bug, I noticed a problem in the way elvis was handling
character-mode operations which happen to end at the start of a line.
In certain circumstances, elvis was including the final newline character
as part of the affected text, where the real vi would exclude it.
This has been fixed.

<p>Another minor bug: After a command such as "<kbd>25i*^[</kbd>" (to insert
25 asterisks), attempting to repeat the command via "<kbd>.</kbd>" would only
insert 24 asterisks.  This has been fixed.

<dt>Name completion
<dd>The behavior of name completion is now slightly different.  In 2.1_3,
even if you've left the <code>completebinary</code> option turned off, elvis
will still complete binary files' names when you're entering a shell command.
This is handy when you're trying to enter an ex command such as...
<pre>
	:r !uuencode elvis-2.1_3.tar.gz &lt;elvis-2.1_3.tar.gz
</pre>
... in which you're running a program which takes a binary file as input.  In any
other context, the <code>completebinary</code> option will continue to work
as before.

<p>If you normally have <code>inputtab=identifier</code> set,
elvis will ignore that (treat it like <code>inputtab=tab</code>)
unless there is a "tags" file in the current directory.
Without this rule, there could be a long delay if you're editing a
non-source file and you hit the <kbd>&lt;Tab&gt;</kbd> key, because
elvis would search through the whole tag path for the word before the tab.
(A better solution is to only set <code>inputtab=identifier</code> if the
file is in a known programming language, as detected by the
<code>knownsyntax(</code><var>file</var><code>)</code> function.
You can add that to your ~/.exrc file.)

<dt>Tags (other than name completion)
<dd>The <code>tagprgonce</code> option has been removed because the
<code>:local</code> command offers a cleaner way to achieve the same effect.
In any alias or script, you can simply replace ":set tagprgonce=..." with
":local tagprg=...".

<p>Also, there was a bug which prevented elvis from recognizing tags
whose address contained a C++ comment.
(In other words, "//" confused the tag file parser.)
This bug is now fixed.

<dt>Error reporting
<dd>When there's an error in a script file or an alias, elvis 2.1_3 will
now (usually) tell you the name of the script and the line number within it,
where the error occurred.
It only does this if the verbose option is set to 1 or higher
(i.e., you give at least one uppercase -V flag on the command line).

<p>I added this mostly because users on some Unix platforms have reported
that elvis says "# is illegal during initialization".  It doesn't happen on
any system I use, and none of the standard scripts contain a "#" command,
so I'm hoping this will allow users to give me more information.

<p>Also, the <code>:cc</code> and <code>:make</code> commands used to output
the first error message before the user hit <kbd>&lt;Enter&gt;</kbd>, but
now it will show that message afterward.

<dt>Networking
<dd>The notation for FTP URLs has been extended.
The original release of elvis 2.1 supported the standard
<strong>ftp://</strong><var>host</var><strong>/</strong><var>resource</var>
notation for anonymous FTP, and elvis' own
<strong>ftp://</strong><var>host</var><strong>/~/</strong><var>resource</var>
notation for accessing your own FTP account.
Elvis 2.1_3 now adds
<strong>ftp://</strong><var>host</var><strong>/~</strong><var>user</var><strong>/</strong><var>resource</var>
to access some other user's account.

<p>The format of the <code>~/.netrc</code> file or <code>elvis.ftp</code> file
remains the same, but the way elvis interprets it has changed.
It now allows multiple entries for the same machine, with different user names.
When you use the
<strong>ftp://</strong><var>host</var><strong>/~</strong><var>user</var><strong>/</strong><var>resource</var>
notation to access a given user's account,
elvis will search for an entry with a matching <var>host</var> and <var>user</var>, and use the password from that entry.

<p>Elvis' implementation of HTTP was having trouble talking to some sites that
run older server software.
This should be fixed now.

<dt>Windows95/98/NT
<dd>The font selection dialog was listing all fonts, even though elvis can
only use fixed-pitch fonts correctly.  This has been changed, so the dialog
will now list only the fixed-pitch fonts.

<p>Many fixed-pitch fonts were being displayed with a large gap between the
characters.  For example, although "Courier" looked good, "Courier New" was
unusable.  This has been corrected by making elvis use the average character
width instead of the maximum width.
(Shouldn't those two values be the same for fixed-pitch fonts?
Apparently they aren't.)

<p>You can now define your own colors, in addition to the named colors, by
using the notation #RRGGBB anywhere a color name is expected.  The "RR", "GG",
and "BB" are any pairs of hex digits, giving the amount of of red, green, and
blue in the color.  For example, "#c0d0ff" is a nice sky-blue color.
(This is the same notation that X-windows uses for non-standard colors.)
Presently this feature is limited to defining one extra foreground color plus
one extra background color, and it only works correctly on "high color" or
"true color" displays (<em>not 16-color or 256-color</em>); these limitations
should be lifted in later versions of elvis.

<p>Sometimes, for some programs, elvis was unable to redirect a program's stdin.
This messed up attempts to filter text through an external program.
Typically, the <code>fmt</code> program included with elvis worked correctly,
but the standard MS-DOS <code>sort</code> program didn't.
I've spent probably 10 whole days trying to get the Win32 I/O redirection
functions to work consistently, and failed.
I suspect that Serge has had the same bad luck.
Elvis 2.1_3 has a totally rewritten module for running programs, which
simply creates a batch file and runs it.
If there is filtering to be done, then the batch file redirects stdin and/or
stdout to temporary files.
This is a kludge, but it seems to work reliably.

<p>There were a couple of bugs in the "Options-&gt;Syntax" menu item.
It wasn't always enabled for all syntaxes, and if it was enabled then
selecting it would occasionally cause elvis to crash.
Both of these should be fixed now.

<dt>Miscellaneous Annoyances

<dd>In filenames, you can now use the backtick notation to run an external
program and have its output be parsed as part or all of the filename list.
E.g., <code>:n `grep -l sometext *.c`</code>.
While I was at it, I added a new <code>shell("command")</code> function which
runs an external program and returns its output, so you can do things like
<code>:let d=shell("date")</code>.
(These features are omitted from the MS-DOS version to help elvis fit in
the lower 640k.)

<p>The <code>:mark</code> command, when invoked without any addresses, will
now set the mark to the exact cursor position; previously it was just setting
it to the start of the cursor's line.

<p>I added a <code>:kwic</code> alias, which greps for a given word anywhere
in elvis' documentation and builds a table table of all matches, with links
to their locations.
This gives you yet another way to search through elvis' documentation
(in addition to the <code>:howto</code> alias and the <code>:help</code>
built-in command).
For non-Unix systems, you must find a grep utility somewhere for this to work.

<p>If your ~/.exrc file contained <code>:set wrapmargin=0</code> then
wrapmargin would be listed by <code>:set</code>, but it shouldn't be.
And now it won't be.

<p>It has been reported that "dead keys" (for entering accented letters)
don't work under X11.
It appears that my big mistake here was failing to #include
&lt;X11/Intrinsics.h&gt;, so elvis' source code didn't know that
the multi-character input methods were available.
That's fixed, but now the multi-byte input code is being compiled for
the first time, and new bugs may turn up.

<p>If the TERM environment variable is unset and there is no reasonable default,
then elvis will use the "open" user interface instead of "termcap".
Previously you had to give an explicit "-Gopen" flag to make this happen.
The new behavior is more like the traditional vi.

<p>The markup display modes (html, man, and tex) can now be configured in or
out separately, via DISPLAY_HTML, DISPLAY_MAN, and DISPLAY_TEX macros in the
config.h file.
Previously, there was a single DISPLAY_MARKUP macro which controlled all of
them.
This change was made mostly so that "tex" could be omitted from the MS-DOS
version.

</dl>

<h2><a name="revision4"></a>3.7 Differences between 2.1_4 and 2.1_3</h2>

This version of elvis is 2.1_4 (pronounced "two point one patch-level 4").
It may also be written as 2.1.4;
however, the "2.1_4" notation was chosen for a reason:
the directory name "elvis-2.1_4" is a valid MS-DOS directory name, but
"elvis-2.1.4" is not.

<p>This is primarily a bug-fix update.
There are some big changes that I'd like to make to elvis, but
they'll all have to wait until version 2.2.
This bug-fix version is simply intended to make waiting for 2.2 a little
easier.

<p>Some of the main areas of bug fixes are:

<dl>

<dt>Buffer bugs
<dd>Several people reported that elvis complained about the '#' being used
in initialization scripts (even though '#' doesn't appear anywhere in those
scripts), or that sometimes a buffer would get scrambled.
These problems have now been fixed, thanks mostly to William Plant.
He tracked the problem down to the use of FEATURE_LITRE in scan.c.

<p>This was a significant bug!
The '#' thing was only one manifestation of a dangerous bug.
Other manifestations could mangle buffers, cause core dumps, etc.
This bug fix alone is probably enough to justify downloading 2.1_4.

<dt>Cygwin support
<dd>The Win32 version of elvis 2.1_3 (or earlier) didn't work very well
under Cygwin's <code>bash</code> shell.
elvis 2.1_4 should fix that.  Specifically...
<ul>
<li>Cygwin sets the TERM environment variable to "cygwin",
which elvis will now recognize as a synonym for "console".
<li>Elvis now supports the mount table produced by the Cygwin
<code>mount</code> utility.

<dt>Windows98 chokes on vi.exe, etc.
<dd>An apparent bug in recent versions of Windows causes the
<code>vi.exe</code>, <code>view.exe</code>, and <code>ex.exe</code> programs
to run in the background.
(Those programs worked in Windows95, but choked in Windows98 and NT.)
To work around this, I replaced the execvp() call with a spawnvp() call
followed by an exit() call.
It'll waste a little RAM now, but at least it works.

<p>This bug also hit Cygwin's <code>bash</code> shell, when run under Windows98
or NT.
I've tested 2.1_4 under <code>bash</code>, and it works now.

<dt>Running external programs under Win32
<dd>There were two bugs which, together, caused elvis to act weird after
running an external command such as ":!dir".
This was particularly acute in the text-mode version, but now both versions
seem to work as intended.

<dt>Screen glitch in MS-DOS with NANSI.SYS
<dd>The screen wasn't always drawn correctly when using one of the better
ANSI drivers, and setting TERM=nansi (or whatever).
This was due to a bug in the nansi termcap entry in tinytcap.c;
it should be fixed now.

<dt>Scrolling after :split
<dd>Previously there was a bug which showed up when you have two windows
open on the same buffer.
Inserting text in one window could cause the other window to scroll after
each keystroke.
This has been fixed.

<dt>Name completion with spaces
<dd>There were some problems when attempting filename completion on
files whose names contain spaces.
The bug responsible for this has been fixed.

<dt>Ctags/elvtags
<dd>Some bugs in the ctags/elvtags program have been fixed.
Mostly these only occured when you're using long lines.

<dt>Display modes
<dd>The "tex" and "syntax" display modes should work a little better now.
I added some TeX markups, and enhanced the "syntax" display mode to handle
PERL slightly better.
Also, some users have contributed syntax descriptions for new languages.

<dt>:make
<dd>The <code>:make</code>, <code>:cc</code>, and <code>:errlist</code>
commands were treating any alphanumeric word as a filename, without
checking to see if it even exists.
It was <em>supposed</em> to verify that the name was an existing,
readable/writable (or just readable if <code>noanyerror</code>),
non-binary file... and now it will.

<dt>POSIX compatibility tweaks
<dd>The <code>:=</code> command, when invoked without line number, should
display the total number of lines in the buffer.  Previous versions of elvis
didn't do that correctly, but 2.1_3 did.
Unfortunately, that fix broke the <code>:.=</code> command, which is supposed
to display the current line number.
Version 2.1_4 should both of those right, finally.

<p>The <kbd>^F</kbd> and <kbd>^B</kbd> paging commands are supposed to leave
two lines of overlap, but previous versions of elvis have only kept one line
of overlap.
(NOTE: I considered adding an option to control the overlap, because if you
have a 24-line screen and 66-line pages, then then it would be nice to leave
1 line of overlap because three <kbd>^F</kbd>s or <kbd>^B</kbd>s would equal
one page.
If you believe an option would be useful, please let me know.)

<p>The <kbd>$</kbd> command accepts a <var>count</var> argument, and moves
forward (<var>count</var>-1) lines before moving to the end.

</dl>

<hr>
<p>The only new features (since 2.1_3) are...
<dl>

<dt>:switch/:case/:default commands
<dd>Elvis now has a :switch command for doing a multi-way branch in a readable
way.
I believe this is justified, since the current :if/:then/:else commands would
require you to use deeper and deeper nested braces to achieve the same effect.
The old way:
<pre>
	if os=="unix"
	then echo Unix
	else {
	 if os=="win32"
	 then echo Windows95/98/NT/2000
	 else {
	  if os=="msdos"
	  then echo MS-DOS
	  else {
	   if os=="os2"
	   then echo OS/2
	   else echo Who?
	  }
	 }
	}
</pre>
The new way:
<pre>
	switch os
	case unix echo Unix
	case win32 echo Windows95/98/NT/2000
	case msdos echo MS-DOS
	case os2 echo OS/2
	default echo Who?
</pre>

<dt>"html" printer type
<dd>Setting the <code>lptype</code> option to "html" causes the
<code>:lpr</code> command to write HTML output, instead of normal printer
control sequences.
This give a good way to generate WYSIWYG HTML code from any of elvis'
display modes.
It required minimal changes to the code, too -- basically, I just created
a new printer type that uses "&lt;b&gt;" to start bold font, "&lt;i&gt;" to
start italics, and so on, instead of some printer-dependent escape sequence.

<dt>"makehtml" alias
<dd>There is a new <code>:makehtml</code> alias which attempts to convert plain text to HTML.
It works pretty well, though you'll still need to hand-edit the results in
many cases.  You might want to compare the output of <code>:makehtml</code> to
that of the "html" lptype, to see which one fits your needs better.

<dt>"smarttab" option
<dd>I added VIM's "smarttab" alias.
When set, this causes the <kbd>&lt;Tab&gt;</kbd> key to act like <kbd>^T</kbd>
when pressed in the indentation area of a line.
Elsewhere in the line, <kbd>&lt;Tab&gt;</kbd> is treated normally.
Now you can do things like "<code>:set&nbsp;smarttab&nbsp;shiftwidth=4</code>"
to change your indentation levels while leaving the <code>tabstop</code>
option unchanged.

<p>The current implementation is a little half-assed though --
you can't backspace over the indentation.
Instead, you must hit <kbd>^D</kbd> to reduce the indentation level.

</dl>

<h1><a name="future"></a>4. The future of elvis</h1>

One of the biggest tasks on my list is to rewrite the ctags program
so that it will be able to parse C++ code better.
It should read the language descriptions from "elvis.syn", and do at least
a half-assed job of generating tags for any language described there.
I'm also toying with the idea of a statistical tags generator, but that's
still pretty nebulous at this point.

<p>Windowing will be more versatile.
Currently the GUI versions of elvis always split detached windows;
sometimes it would be nice if they could be attached to the existing window,
as a horizontal or vertical pane.

<p>More colors will be supported.
The syntax of the <code>:color</code> command will change to allow you
to specify colors and other attributes by the type of text, instead of
defining colors by attributes.
For example, "<code>:color link=underlined green</code>" will define the
appearance of HTML links.

<p>I intend to add a true extension language to elvis.
The language interface will be general enough to support a variety of
languages.
The first language supported will probably be PERL,
followed rapidly by Python and TCL.

<p>I'm thinking of modifying the markup display modes (man, tex, and html)
to store the list of supported markups, and their effects, in a separate file.
That way, you could create your own markups to display nroff -ms, RTF, SGML,
and MIME "rich text" documents.

<p>I'd like to redesign the way elvis assigns assigns buffer names.
I'd like to make it always use the full pathname of a file as the buffer name.
In addition, I'd like for elvis to store a "current working directory" for
each window, and use that to convert a relative buffer name into a the absolute
pathname.
A similar trick would be used for filenames.
When running an external program, elvis would change the real current
directory to that window's directory.
The benefits of all this:
<ul>
<li>If you're editing "foo" and "./foo", they'll both use the same buffer.
This sounds contrived, but a similar thing happens with tag files fairly often.
<li>If you do a :cd command, then the old buffers will still write out to
their old directories.
<li>Different windows could have different directories.
<li>The X-windows "-client" flag would behave in a more reasonable manner.
<li>The "current directory" could be something other than a local directory.
For example, there's no reason you shouldn't be able to ":cd ftp://ftp.sunsite.unc.edu/pub/incoming".
</ul>

<h1><a name="links"></a>5. Links to related files</H1>

If the main site (<a href="ftp://ftp.cs.pdx.edu/pub/elvis/README.html">ftp.cs.pdx.edu</a>) is too slow, try the mirror site at
<a href="ftp://ftp.false.com/pub/elvis/README.html">
ftp.false.com</a>.

<p>Most of the following are binary files, not text or HTML files,
so you can't view then with your Web browser.
But you can use your browser to download the files.
For Netscape, use &lt;Shift-Click&gt;;
for MSIE, use &lt;RightClick&gt; and "download".

<dl>

<dt><a href="/pub/elvis/untar.c">untar.c</a>
<dd>This is the complete source code for "untar",
a little program which extracts files from a gzipped tar archive.
Comments near the top of "untar.c" describe how to compile and use it.
If you already have the gzip and tar utilities, then you don't need this.

<dt><a href="ftp://ftp.cs.pdx.edu/pub/elvis/untardos.exe">untardos.exe</a>
<dd>This is an MS-DOS executable, produced from the above "untar.c" file.
It can also be run under Windows 3.1, in a Dos-prompt window.
For brief instructions on how to use <code>untardos,</code>
run it with no arguments.

<dt><a href="ftp://ftp.cs.pdx.edu/pub/elvis/untarw32.exe">untarw32.exe</a>
<dd>This is a Win32 executable, produced from the above "untar.c" file.
It runs under WindowsNT and Windows95.
It runs somewhat faster than the MS-DOS version.
It also supports long file names.
For brief instructions on how to use <code>untarw32,</code>
run it with no arguments, in a text-mode window.

<p><strong>NOTE:</strong>
MS-Windows95 and MS-DOS use incompatible methods for mapping long file names
to short ones.
So if you extract the files under Windows95, DOS programs won't be able to
find them with their expected names, and vice versa.
Consequently, you must use <code>untardos.exe</code> to unpack
<code>elvis-2.1_4-msdos.tar.gz</code>, and <code>untarw32.exe</code> to unpack
<code>elvis-2.1_4-win32.tar.gz</code>.

<dt><a href="ftp://ftp.cs.pdx.edu/pub/elvis/untaros2.exe">untaros2.exe</a>
<dd>This is an OS/2 executable, produced from the above "untar.c" file.
For brief instructions on how to use <code>untaros2,</code>
run it with no arguments.

<dt><a href="elvis-2.1_4.tar.gz">elvis-2.1_4.tar.gz</a>
<dd>This is a gzipped tar archive of the source code and documentation for
Elvis 2.1 and its related programs.

<dt><a href="elvis-2.1_4-msdos.tar.gz">elvis-2.1_4-msdos.tar.gz</a>
<dd>This archive contains the documentation and MS-DOS executables
for Elvis 2.1.

<dt><a href="elvis-2.1_4-win32.tar.gz">elvis-2.1_4-win32.tar.gz</a>
<dd>This archive contains the documentation and Win32 executables
for Elvis 2.1.
These were compiled and tested under Windows95, but should work under
WindowsNT 3.51 (or later) as well.

<dt><a href="elvis-2.1_4-os2.tar.gz">elvis-2.1_4-os2.tar.gz</a>
<dd>This archive contains the documentation and OS/2 executables
for Elvis 2.1.

<dt>
<a href="ftp://ftp.fh-wedel.de/pub/fh-wedel/staff/herbert/elvis/00-index.html">
ftp://ftp.fh-wedel.de/pub/fh-wedel/staff/herbert/elvis/00-index.html</a>
<dd>This is where the OS/2 maintainer stores his most up-to-date versions.
It may be better than the <code>elvis-2.1_4-os2.tar.gz</code> file, above.
</dl>

</body></html>