Evi is a vi emulator aimed at those who are either just plain accustomed to vi, or who just simply happen to like its style of editing better than emacs' default. Evi's first goal is vi compatibility. Its second goal is to be an extension of vi, taking advantage of features supplied by the emacs environment, without simply becoming emacs with vi'ish key bindings.
Ideally, you shouldn't need any special manual to start using Evi, other than your vi manual. By default evi is set-up to emulate vi as closely as possible, including using your EXINIT environment variable and/or .exrc startup files. Of course, you'll need to know how to get started with Evi, and you'll probably want to take advantage of some of Evi's extensions, and be aware of what's different between vi and Evi. That's what this manual covers.
A note on conventions used in this manual: M-x means the escape key followed by the character x (where x is any character), C-x means control-x, RET means the return key, ESC means the escape key.
To just test Evi out, type:
M-x load-file RET evi.el RET M-x evi RET(don't type the spaces) You will now be in the emulator.
If you have any trouble at this point due to something in your `.exrc' or
EXINIT, and wish to suppress sourcing of
`.exrc' and EXINIT, place the
following in your `.emacs':
(setq evi-supress-ex-startup t)Then, send me a note with the contents of your `.exrc' and
EXINIT
so we can fix the problem.
If you decide to continue using Evi, I would recommend first you byte compile it to speed things up, using the following shell command:
emacs -batch -l evi.el -f batch-byte-compile evi.elNext, if you want to use Evi all the time, put the following at the end of your `.emacs' file:
(load "<wherever-evi-is>/evi") (setq term-setup-hook 'evi)this will make emacs go into Evi every time you fire it up. Of course, you may wish to have Evi properly installed for all to use - consult your local emacs guru.
Vi's file management commands have long tried to mimic having multiple buffers, and as such, work well enough to use in an emacs setting. They of course have to take on slightly different meanings, since it makes little sense to emulate the limitations of the vi/ex model that presumably you are trying to avoid by using Evi!
:e
:e! will override any complaints about
the current buffer being modified, and discards all
modifications. With a filename argument, edits that file
in the current window (using the copy already in the editor
if it was previously read in). There is no difference
between :e! filename and :e filename, because in Evi we
don't need to worry about the disposition of the current
file before editing the next one. Use :e# as a shorthand for
editing the most recently accessed buffer not visible in a window. :E
:e, but edits the file in another window, creating
that window if necessary. If used with no filename, this command splits the
current buffer into two windows. :n
:n will cycle thru all buffers. :N
:n, but switches to another window, or creates another
window and puts the next file into it. :b
:b foo would switch
to the buffer named foo (if it exists). By default :b
switches to the next buffer not displayed. :b! will create
the buffer if it doesn't exist. :B
:k
:k! will kill the buffer
even if it is modified. :wk
:W
:W! will
write all buffers with no questions asked. :Wq
The ex commands which accept filenames as arguments can be file
completed using space or tab. Similarly,
:b and :k will buffer-name complete.
One thing you might find handy is
:b TAB to see a list of buffers.
Use C-c or backspace to escape this.
Vi had commands for multiple file management, but it never tried to pretend it handled multiple windows. However, it did have the z command for simple window management, like making the window bigger or smaller. In Evi, the z command is extended to handle most Emacs window management needs. Additional suggestions are welcome.
By default, no emacs commands are accessible from Evi. However, the three main classes of emacs commands can easily be made accessible by adding selected elisp commands suggested below to your .evirc file.
The C-x prefix commands integrate most easily with Evi since it would only override kill-line in input modes. To access C-x prefix commands in all modes, use:
(evi-define-key evi-all-keymaps "\C-x" 'evi-emacs-command)The Meta (or ESC prefix) commands are no problem if you just want to access them from the top-level, where in vi ESC only serves the purpose of ringing the bell, which in vi is a handy way of verifying that you're in command mode, but is not vital considering the emacs mode line. Use the following to access Meta (or ESC prefix) commands only from the top-level:
(evi-define-key '(top-level) "\e" 'evi-emacs-command)Accessing Meta (or ESC prefix) commands from other modes is problematic because ESC is the command to exit that mode. However, if you wish, you can redefine the meta prefix to something else - I recommend C-a. Thus to use Meta commands in all modes use:
(setq evi-meta-prefix-char ?\C-a) (evi-define-key evi-all-keymaps "\C-a" (cons (current-global-map) ?\e))The C-c prefix commands are also problematic because C-c is vi's interrupt character. At the top-level, this is again not much of a problem, and you could use:
(evi-define-key '(top-level) "\C-c" 'evi-emacs-command)C-c will still function as an interrupt character at all other places (including in the middle of a command, where it's most useful). In the various input modes you could forego C-c as your interrupt character and access the C-c prefix commands by using the following (presumably in addition to the above):
(evi-define-key '(input replace ex) "\C-c" 'evi-emacs-command)Note that setting evi-insert-mode-local-bindings (described in the next section) will have the same effect because all C-c commands are local bindings.
A number of emacs editing modes have handy local key bindings other than Meta, C-x and C-c prefix bindings. For example, in C mode, RET does intelligent indenting, and } is bound to a command which automatically exdents. By default, these aren't accessible of course, but you can have Evi enable these local bindings in insert mode by setting:
(setq evi-insert-mode-local-bindings t)As current policy, however, Evi will not allow local mode bindings to override TAB, BS, or DEL, as well as, for obvious reasons, ESC. ESC prefix commands, however, can be accessed as described in the previous section.
Like vi, Evi will source your `.exrc' or `~/.exrc' file,
and/or your EXINIT
environment variable. If your startup runs into problems, let me know -
you shouldn't have to change your vi initialization stuff to make Evi
happy.
If you wish to use some Evi extensions in your startup, but still need
to use vi, place these in `.exrc.evi', `~/.exrc.evi'
or EVIINIT so that vi won't gag on them.
Emacs lisp startup code for evi, such at that suggested in the previous sections, can be placed in either `.evirc' or `~/.evirc'.
And you can, of course, hack away at the the Evi source code if you want something not easily addressed by the above methods. If you feel what you've done would be generally useful, please email it to me, or post it.
One particular customization, not covered elsewhere, is how Evi handles
the current directory. By default it behaves like vi - you have one
global current directory, which you change using :cd (also see
:pushd, and friends described below). Alternately, you may like the
emacs behaviour better, which is that each buffer has its own idea of
the current directory, and by default that directory is the directory
that the file for that buffer resides in. In this mode, you can also
change the current directory using :cd, but that will only affect the
current buffer. To get this behaviour, place the following in your
`.evirc':
(setq evi-global-directory nil)
Another customization you might like to make is to alter the behaviour
of ZZ. By default it is bound to :Wq!, which quietly writes all
modified files and exits. If, however, you would like to be asked about
each modified buffer before it is saved in order to avoid accidently
saving a file you didn't want saved, map ZZ to :Wq:
map ZZ :Wq\n
If you like your arrow keys, there's one incompatibility that's not
easily fixed under version 18 of emacs. Vi recognizes arrow keys that
send ESC prefixed sequences by using a very short timeout to determine
the difference between an ESC typed by a person,
and an ESC sequence
sent by the terminal. A compromise, if you happen to have such a
terminal is to :set notimeout,
which makes ESC ESC behave like a
single ESC and maps your arrow keys to h, j, k,
and l. As a
side effect in Evi, the normal emacs ESC prefix commands will be in
effect (except for ESC ESC of course).
This could be undesirable if
you're in the habit of hitting the escape key a lot to verify you're in
command mode, and follow that immediately with a command. This should
be fixable in version 19 of emacs.
_ is a new version of the repeat command . that prompts you with the keystrokes to repeat, allowing you to edit them before executing. This is particularly useful for the abovementioned complex operators. If you don't wish to re-execute the command, just hit C-c.
If you put
(setq ex-input-escapes t)in your `.evirc', then Ex (:) commands will accept the following escapes:
\e for ESC,
\n for newline, \r for RET,
and \C-x for control-x (for all valid control characters).
\ otherwise works like C-v. Thus: map \ \C-fwould make the space character be page forward, and
map g \|would make g be goto-column. Note that | is normally a command separator and thus must be escaped.
The command [u continues the previous undo, by undoing one more change previous to the last change undone. Thus, a long enough sequence of [us will take you back to the unmodified state. If you went back too far, a u will reverse this process, and subsequent [us will move forward through the changes. Note that vip does this using repeat command (.); however, that conflicts with the meaning of u. in vi, which is: `undo, then do again'. This is quite handy for reapplying a change that you initially did in the wrong place, so evi leaves that meaning alone, and defines a new command for `extended undo'. Personally, I use `extended undo' more than `undo line', so I swap the two definitions in my .evirc:
(evi-define-key '(vi) "[u" 'evi-undo-line) (evi-define-key '(vi) "U" 'evi-undo-more)
In Ex (:) commands, you can use the TAB to perform completion on the following: commands, variable names (:set), filenames, buffernames, maps, or abbreviations. Which completion to perform is determined solely by where you are in the partial command you are typing. For example:
:set erTABwould complete to:
:set errorbells(leaving the cursor after the
s in errorbells.
You can define exactly what Evi treats as words for the w, b, e, W, B
and E commands. They are defined by setting either or both of the new
options word (for w, b and e) or
Word (for W, B and E) to a regular
expression describing what words look like. For example, here's a
definition of words that only considers alphanumeric words:
set word=[a-zA-Z0-9]+Contrast this with the default definition:
[a-zA-Z0-9_]+\\\|[^a-zA-Z0-9_\ \t\n]+\\\|^[\ \t]*\nSee the emacs documentation on regular expressions for details.
My interpretation of sentence, paragraph, and section motion differs somewhat from vi's in that they behave more analogously to how word motion behaves - e.g. a forward paragraph takes you to the beginning of the next paragraph - not the blank line after the previous paragraph. However, when doing a delete using one of these motions, unless you are at the beginning of the sentence, paragraph or section, the delete will only happen to the end of the sentence, paragraph or section, not to the beginning of the next. I find this *much* more useful than the vi behaviour - if you disagree, please let me know.
:pushd, :popd and :dirs commands exist,
similar to those found in
csh and bash. Note these only make sense in conjunction with
evi-global-directory = t (which is the default).
The :! command now takes an optional & (as in :!&),
which causes
the shell command to be executed asynchronously, with the output going
to the window `*Shell Command Output*'.
The unnamed register (where deleted text goes) is preserved across excursions into insert mode, etc. This means you can delete something, insert something, then `put' the deleted text. In vi, for no apparent reason, you can't do this, even though insert mode doesn't use the unnamed register.
The unnamed register is also preserved between buffers, so you can yank text in one buffer and put it into another.
Several commands that didn't take counts in vi take counts in Evi. p and P take a prefix count and will put the text that many times, regardless of the size of the text - vi will apparently only do the prefix count for less than line sized text. / takes a prefix count to find the nth occurence of a string. D takes a count (I could never figure out why it didn't, since C takes a count).
In vi, you do most of your editing based on regions defined by vi's motion commands. For example, dw deletes the region starting at the current cursor location and extending up to where the cursor would be if you'd typed w. Those edit commands that don't explicitly use a motion command are just shorthand for ones that do: e.g. x is shorthand for dl.
However, this doesn't give you a convenient way of operating on arbitrary rectangular regions, or regions that are inconvenient to describe by a single motion command. Evi allows you to operate on such regions by a special form of the m (mark) command, and several new commands that are only understood when they are operands to edit operators (such as c, d, >, etc).
The basic idea is that you will define a region by two points. For a rectangle region, for example, you will define two opposite corners. You define the first point by using m.. Then you move to the second point and execute the operator you want, specifying the motion to be one of: r for rectangle, R for rows, C for columns, and a for arbitrary (meaning everything between the marked point and where the cursor currently is).
m.3j5wdawould delete the text from where the cursor started to 3 lines down and 5 words over. R is often handy for operating on large arbitrary sections of text, for example say you needed to shift some text that ran on for several pages and you weren't sure just how long it was at the start:
m.C-fC-fjjj>R
:shell starts up an emacs shell in the current window (instead of
suspending emacs, and starting a subshell). The shell to run comes from
the vi variable ishell, and defaults to the value of the environment
variable SHELL.
:gdb program starts up gdb in the current window on
the specified program. For both of these, you are automatically placed
in insert mode, where you should be able to interact as you would expect
to, except that ESC will take you into command mode. While in command
mode, hitting return will send the current line as input to the
shell/gdb, similar to command-line editing vi-style in bash, but will
leave you in command mode.
The marks used by the mark command m are emacs markers, thus they mark a position in a buffer, not necessarily the current one. This affects the goto mark commands ` and '. For example, if mark a is placed in one buffer, then later in another buffer, the command `a is typed, evi will first switch to that buffer, then go to the location in that buffer. ' and ` also accept . and , for pop context, and unpop context respectively. Thus, '. will take you to the previous context (defined as in vi by a region of relative motion, with an `absolute' motion pushing a new context. quotes surround `absolute' because a search is considered an absolute motion for this purpose), and '. will take you to the context before that. There is a ring of 10 contexts so after 10 '. commands you'll end up at the original previous context. `Unpop context' means move forward thru the ring. " and " are defined as exchange current location with the location of the previous context. The context ring is buffer local, so use of it will always keep you in the same buffer.
Two changes involving registers. First, "', and "", are new commands which allow you to insert literal text directly into a register. "' inserts a single character, and "" inserts a string. E.g. ""helloESC inserts the string `hello' into the unnamed register, and "a"'/ inserts a slash into register a. The choice of command names are intuitive (they suggest quotes around a literal char/string), albeit unfortunate because " is supposed to be the prefix for a register specification, not a command. Second, the register specification "^ specifies appending to the unnamed register (the one that gets used when no register is specified). E.g., "^""ickESC appends `ick' to the unnamed register.
% exhibits language sensitivity in that it ignores parentheses embedded in quotes. What defines quotes is based on what minor mode emacs is in (such as c-mode or lisp-mode), or you can roll your own (see emacs command modify-syntax-entry).
= is no longer specific to :set lisp. It indents according to the
mode. See emacs command indent-according-to-mode.
The * operator can be used to send text to emacs processes. *
prompts for the name of a process buffer, and the region specified is
sent to that process. Subsequent invocations of * will use the same
process buffer as last specified as a default. E.g., to send the
current line of text as a command to the emacs shell (see :shell),
type ***shell*RET, or if the shell is already the default, just
**RET. Paragraph motion or parenthesis match is often perfect for
sending function definitions to an interpreter, e.g. place the cursor at
the beginning of the function,
and type *}RET or *%RET. If the
function def is less easily described you can use m. and yR
described above. In specifiying the process buffer, you can use buffer
completion using space or tab.
I'm experimenting with some new complex operators. I'm particularly interested in your thoughts on these:
[{ operates over lines in a region. It takes a motion, and a sequence of operations to perform on each line in the region defined by the motion. The sequence of operations is prompted for on the bottom line. Double the { to operate on whole lines. The point starts in the first column for each line operated on. For example:
[{}i> C-vESCRET
would prefix every line in the rest of the current paragraph with `> '.
The C-v ESC sequence inserts an ESC into the string you are
entering so that it will terminate input when the loop body is executed,
not as you are entering the command. For example:
10[{{i/* C-vESCA */C-vESCRET
would place C-style comments around the next 10 lines.
[( defines a parameterized macro body. A parameterized macro is diff- erent from standard macro text in that it is parameterized by prefix count and register specification. In the body of such a macro, there are two special commands: # and &. # is replaced by the prefix count applied to this macro, and & is replaced by the register specification applied to this macro. For example:
"a8[(j#w&dwRETwould go down one line, move over 8 words, then delete the next word into register `a'. This is rather contrived, but it gives you the idea. Param- eterized macro bodies are obviously not very useful typed out each time, and are intended to be the body of a map macro. For example:
:map M [(j#w&dw\eRET "a8Mwould be a much more likely scenario for the use of such a macro.
The following vi commands behave differently in Evi or not implemented:
Digit registers don't work entirely correctly - there are circumstances in which separate lines of a change/deletion are supposed to go into separate registers
:set lisp has no effect, however, emacs does largely take care of
any lisp'ish behaviour you'd want automatically if the file you're
editing is suffixed with `.l' or `.el'.
One particular loss, however,
is that ) and ( don't work on s-expressions
like they would in vi with lisp set.
The following ex commands are supported in Evi:
abbrev, cd, chdir, copy, delete, edit, file, global, map, move, next, preserve, print, put, quit, read, recover, set, shell, source, substitute, tag, unabbrev, unmap, write, wq, yank, !, <, >, &
The following ex variable settings are supported in Evi:
autoindent, errorbells, ignorecase, magic, notimeout, shell, shiftwidth, showmatch, tabstop, wrapmargin, wrapscan
Undo does not continue via .. This is incompatible with vi - the sequence u. in vi means `undo, then do again', whereas in vip it means `undo, then undo some more.' For the vip functionality use evi-undo-more, described in the section on enhancements.
The vip commands for editing a file (v and V) and switching buffers
(s, and S) are not supported.
Use :e, :E, :b, and :B
instead. See previous section on file and buffer management, or try
these cute macros which are (mostly) functional replacements, including
doing file and buffer completion (note the space on the end of the
line):
:map v :edit |map V :Edit |map K :kill |map s :buffer |map S :Buffer
:q exits emacs. I believe the default behaviour in vip is to simply
kill the current buffer (a concept vi doesn't really have) - either that
or quit vi emulation. Any of these choices is reasonable, however,
given :k for killing buffers (a new command for a new concept), and
Q for exiting vi emulation, I chose to have :q do exactly what it
does in vi.
There are still some glitches in evi, of course, and if you encounter one, please email me what went wrong, in as much detail as you can. My address is jlewis@cse.ogi.edu. In the meantime, if Evi doesn't seem to be responding, or you're having difficulty getting out of some mode, first try C-c (this is vi's interrupt character), then if you're still in a pickle, try C-g (this is emacs' interrupt character - there are some situations that you might get into where it works instead of C-c).
There is a mailing list for discussion of evi - usage, bugs, new features, etc. To join the list, send a note to evi-list-request@brandx.rain.com including your e-mail address in the body of the message. To submit to the list, send mail to evi-list@brandx.rain.com.
You can obtain a copy of Evi via FTP from an elisp archive site, or from the mail-server on my machine (which will assure the most up-to-date version). The main elisp-archive is at archive.cis.ohio-state.edu (in the directory gnu/emacs/elisp-archive) - Evi is in modes/evi.el.Z. To get it from my mail-server, send a message to mail-server@brandx.rain.com containing:
send evi.el-- or --
send evi.el.Zfollowed by:
end
evi.el.Z will be uuencoded for delivery. Also, add `send index'
if you want the file sizes.
Masahiko Sato - for having the audacity to write vip in the first place ;-) Eric Bensen - for being the first real advocate and those helpful folks who gave feedback on the first release and the following immensely helpful souls: James Montebello, Steven Dick, Mike Sangrey, Brian Divine, Bill Reynolds, Roger E. Benz, Richard Ryan, Volker Englisch, John Haugen, Calvin Clark, Sam Falkner, Mark Pulver and Cameron Gregory