vi reference (by maart@cs.vu.nl)<![CDATA[
Warning: some vi versions don't support the more esoteric features
described in this document. You can edit/redistribute this document
freely, as long as you don't make false claims on original authorship.
default values : 1
^X : <ctrl>x
[*] : * is optional
<*> : * must not be taken literally
<sp> : space
<cr> : carriage return
<lf> : linefeed
<ht> : horizontal tab
<esc> : escape
<erase> : your erase character
<kill> : your kill character
<intr> : your interrupt character
<a-z> : an element in the range
N : number (* = allowed, - = not used)
CHAR : char unequal to <ht>|<sp>
WORD : word followed by <ht>|<sp>|<lf>
]]>move commands<![CDATA[
N | Command | Meaning
---+--------------------+-----------------------------------------------
* | h | ^H | <*> chars to the left
* | j | <lf> | ^N | <*> lines downward
* | l | <sp> | <*> chars to the right
* | k | ^P | <*> lines upward
* | $ | to the end of line <*> from the cursor
- | ^ | to the first CHAR of the line
* | _ | to the first CHAR <*> - 1 lines lower
* | - | to the first CHAR <*> lines higher
* | + | <cr> | to the first CHAR <*> lines lower
- | 0 | to the first char of the line
* | | | to column <*> (<ht>: only to the endpoint!)
* | f<char> | <*> <char>s to the right (find)
* | t<char> | till before <*> <char>s to the right
* | F<char> | <*> <char>s to the left
* | T<char> | till after <*> <char>s to the left
* | ; | repeat latest "f"|"t"|"F"|"T" <*> times
* | , | idem in opposite direction
* | w | <*> words forward
* | W | <*> WORDS forward
* | b | <*> words backward
* | B | <*> WORDS backward
* | e | to the end of word <*> forward
* | E | to the end of WORD <*> forward
* | G | go to line <*> (default EOF)
* | H | to line <*> from top of the screen (home)
* | L | to line <*> from bottom of the screen (last)
- | M | to the middle line of the screen
* | ) | <*> sentences forward
* | ( | <*> sentences backward
* | } | <*> paragraphs forward
* | { | <*> paragraphs backward
- | ]] | to the next section (default EOF)
- | [[ | to the previous section (default begin of file)
- | `<a-z> | to the mark
- | '<a-z> | to the first CHAR of the line with the mark
- | `` | to the cursor position before the latest absolute
| jump (of which are examples "/" and "G")
- | '' | to the first CHAR of the line on which the cursor
| was placed before the latest absolute jump
- | /<string> | to the next occurrence of <string>
- | ?<string> | to the previous occurrence of <string>
- | n | repeat latest "/"|"?" (next)
- | N | idem in opposite direction
- | % | find the next bracket and go to its match
| (also { } and [ ])
]]>searching (see above)<![CDATA[
^] | search in the tags file where the function
| under the cursor is defined (file, line) and
| go to it
:[x,y]g/<string>/<cmd> | search globally [from line x to y] for <string>
| and execute the "ex" <cmd> on each occurrence
]]>undoing changes<![CDATA[
u | undo the latest change
U | undo all changes on a line, while not having
| moved off it (unfortunately)
:q! | quit vi without writing
:e! | re-edit a messed-up file
]]>appending text (end with <esc>)<![CDATA[
* | a | <*> times after the cursor
* | A | <*> times at the end of line
* | i | <*> times before the cursor (insert)
* | I | <*> times before the first CHAR of the line
* | o | on a new line below the current (open)
| the count is only useful on a slow terminal
* | O | on a new line above the current
| the count is only useful on a slow terminal
* | ><move> | shift the lines described by <*><move> one
| shiftwidth to the right (layout)
* | >> | shift <*> lines one shiftwidth to the right
* | ["<a-z1-9>]p | put the contents of the (default undo)
| buffer <*> times after the cursor;
| a buffer containing lines is put only once,
| below the current line
* | ["<a-z1-9>]P | put the contents of the (default undo) buffer <*>
| times before the cursor
| a buffer containing lines is put only once, above
| the current line
* | . | repeat latest command <*> times
| if the last command before a "." command
| references a numbered buffer, the buffer number
| is incremented first (and the count is ignored):
|
| "1pu.u.u.u.u - `walk through' buffers 1
| through 5
| "1P.... - restore them
]]>deleting text<![CDATA[
Everything deleted can be stored into a buffer. This is achieved by
putting a " and a letter <a-z> before the delete command. The
deleted text will be in the buffer with the used letter. If <A-Z>
is used as buffer name, the adjugate buffer <a-z> will be augmented
instead of overwritten with the text. The undo buffer always
contains the latest change. Buffers "<1-9> contain the latest 9
LINE deletions ("1 is most recent).
* | x | delete <*> chars under and after the cursor
* | X | <*> chars before the cursor
* | d<move> | from begin to endpoint of <*><move>
* | dd | <*> lines
- | D | the rest of the line
* | <<move> | shift the lines described by <*><move> one
| shiftwidth to the left (layout)
* | << | shift <*> lines one shiftwidth to the left
* | . | repeat latest command <*> times
]]>changing text (end with <esc>)<![CDATA[
* | r<char> | replace <*> chars by <char> - no <esc>
* | R | overwrite the rest of the line,
| appending change <*> - 1 times
* | s | substitute <*> chars
* | S | <*> lines
* | c<move> | change from begin to endpoint of <*><move>
* | cc | <*> lines
* | C | the rest of the line and <*> - 1 next lines
* | =<move> | if the option 'lisp' is set, this command
| will realign the lines described by <*><move>
| as though they had been typed with the option
| 'ai' set too
- | ~ | switch lower and upper cases
* | J | join <*> lines (default 2)
* | . | repeat latest command <*> times ("J" only once)
- | & | repeat latest "ex" substitute command, e.g.
| ":s/wrong/good"
]]>substitute replacement patterns<![CDATA[
The basic metacharacters for the replacement pattern are `&' and `~';
these are given as `\&' and `\~' when nomagic is set. Each instance
of `&' is replaced by the characters which the regular expression
matched. The meta- character `~' stands, in the replacement
pattern, for the defining text of the previous replacement
pattern. Other metasequences possible in the replacement pattern
are always introduced by the escaping character `\'. The sequence
`\n' (with `n' in [1-9]) is replaced by the text matched by the
n-th regular subexpression enclosed between `\(' and `\)'. The
sequences `\u' and `\l' cause the immediately following character
in the replacement to be converted to upper- or lower-case
respectively if this character is a letter. The sequences `\U' and
`\L' turn such conversion on, either until `\E' or `\e' is
encountered, or until the end of the replacement pattern.
]]>remembering text (yanking)<![CDATA[
With yank commands you can put "<a-z> before the command, just as
with delete commands. Otherwise you only copy to the undo buffer.
The use of buffers <a-z> is THE way of copying text to another
file; see the ":e <file>" command.
* | y<move> | yank from begin to endpoint of <*><move>
* | yy | <*> lines
* | Y | idem (should be equivalent to "y$" though)
- | m<a-z> | mark the cursor position with a letter
]]>commands while in append|change mode<![CDATA[
^@ | if typed as the first character of the
| insertion, it is replaced with the previous
| text inserted (max. 128 chars), after which
| the insertion is terminated
^V | deprive the next char of its special meaning
| (e.g. <esc>)
^D | one shiftwidth to the left
0^D | remove all indentation on the current line
| (there must be no other chars on the line)
^^D | idem, except that it is restored on the next line
^T | one shiftwidth to the right
^H | <erase> | one char back
^W | one word back
<kill> | back to the begin of the change on the
| current line
<intr> | like <esc>
]]>writing, editing other files, and quitting vi<![CDATA[
:q | quit vi, unless the buffer has been changed
:q! | quit vi without writing
:w | write the file
:w <name> | write to the file <name>
:w >> <name> | append the buffer to the file <name>
:w! <name> | overwrite the file <name>
:x,y w <name> | write lines x through y to the file <name>
:wq | write the file and quit vi, WITHOUT
| checking if the write has been
| successful (use "ZZ" instead)
ZZ | write if the buffer has been changed, and
| quit vi; if you have invoked vi with the "-r"
| option, you'd better write the file
| explicitly ("w" or "w!"), or quit the
| editor explicitly ("q!") if you don't want
| to overwrite the file - some versions of vi
| don't handle the "recover" option very well
:x | idem
:x! | ":w!" and ":q"
:e <file> | edit another file without quitting vi - the
| buffers are not changed (except the undo
| buffer), so text can be copied from one
| file to another this way
:e! <file> | idem, without writing the current buffer
:e# | edit the previous file
^^ | idem
:rew | edit the first file (when "vi file1 file2 ...")
:rew! | idem, without writing the current buffer
:n [<file>] | edit the next file
:n! [<file>] | idem, without writing the current buffer
]]>display commands<![CDATA[
^G | give current line number and relative position
^L | refresh the screen (sometimes "^P" or "^R")
^R | sometimes vi replaces a deleted line by
| a '@', to be deleted by "^R" (also with
| option 'noredraw')
[*]^E | expose <*> more lines at bottom, cursor
| stays put (if possible)
[*]^Y | expose <*> more lines at top, cursor
| stays put (if possible)
[*]^D | scroll <*> lines downward
| (default the number of the previous scroll;
| initialization: half a page)
[*]^U | scroll <*> lines upward
| (default the number of the previous scroll;
| initialization: half a page)
[*]^F | <*> pages forward
[*]^B | <*> pages backward (in older versions "^B" only
| works without count)
If in the next commands the field <wi> is present, the windowsize
will change to <wi>. The window will always be displayed at the
bottom of the screen.
[*]z[wi]<cr> | put line <*> at the top of the window
| (default the current line)
[*]z[wi]+ | put line <*> at the top of the window
| (default the first line of the next page)
[*]z[wi]- | put line <*> at the bottom of the window
| (default the current line)
[*]z[wi]. | put line <*> in the centre of the window
| (default the current line)
]]>mapping and abbreviation<![CDATA[
When mapping take a look at the options 'to' and 'remap' (below).
:map <string> <seq> | <string> is interpreted as <seq>, e.g.
| ":map ^C :!cc %^V<cr>" to compile from within vi
| (vi replaces % with the current file name)
:map | show all mappings
:unmap <string> | deprive <string> of its mapping
| when vi complains about non-mapped macros
| (whereas no typos have been made), first do
| something like ":map <string> Z", followed
| by ":unmap <string>" ('Z' must not be a
| macro itself), or switch to "ex" mode first
:map! <string> <seq> | mapping in append mode, e.g.
| ":map! \be begin^V<cr>end;^V<esc>O<ht>"
| when <string> is preceded by ^V, no
| mapping is done
:map! | show all append mode mappings
:unmap! <string> | deprive <string> of its mapping (see ":unmap")
:ab <string> <seq> | whenever in append mode <string> is followed
| by a breakpoint (e.g. <sp> or ','), it is
| interpreted as <seq>, e.g. ":ab p procedure"
| a char preceded by ^V is not considered
| a breakpoint
:ab | show all abbreviations
:unab <string> | do not consider <string> an abbreviation
| anymore (see ":unmap")
@<a-z> | consider the contents of the named register a
| command, e.g.:
| o0^D:s/wrong/good/<esc>"zdd
| explanation:
| o - open a new line
| 0^D - remove indentation
| :s/wrong/good/ - this input text is an
| "ex" substitute command
| <esc> - finish the input
| "zdd - delete the line just
| created into register 'z'
| now you can type @z to substitute `wrong'
| by `good' on the current line
@@ | repeat last register command
]]>switch and shell commands<![CDATA[
Q | ^\ | <intr><intr> | switch from vi to "ex"
: | an "ex" command can be given
:vi | switch from "ex" to vi
:sh | execute a subshell, back to vi by ^D
:!<cmd> | execute a shell <cmd>
:!! | repeat the last shell command
[*]!<move><cmd> | the shell executes <cmd>, with as standard
| input the lines described by <*><move>,
| next the standard output replaces those lines
| (think of cb, sort, nroff, etc.)
[*]!<move>!<args> | append <args> to the last <cmd> and execute it,
| using the lines described by the current
| <*><move>
[*]!!<cmd> | give <*> lines as standard input to the
| shell <cmd>, next let the standard output
| replace those lines
[*]!!! [<args>] | use the previous <cmd> (and append <args> to it)
:x,y w !<cmd> | let lines x to y be standard input for <cmd>
| (notice the space between 'w' and '!')
:r!<cmd> | put the output of <cmd> onto a new line
:r <name> | read the file <name> into the buffer
]]>vi startup<![CDATA[
vi [file] : edit the file and display the first page
The editor can be initialized by the shell variable EXINIT, which
looks like:
EXINIT='<cmd>|<cmd>|...'
<cmd>: set options
map ...
ab ...
export EXINIT (in the Bourne shell)
However, the list of initializations can also be put into a file.
If this file is located in your home directory, and is named
".exrc" AND the variable EXINIT is NOT set, the list will be
executed automatically at startup time. If the 3 conditions are not
met, you have to give the execute command yourself:
:source file
or
:so file
On-line initializations can be given with "vi +<cmd> file", e.g.:
vi +x file | the cursor will immediately jump to line x
| (default last line)
vi +/<string> file | ~ to the first occurrence of <string>
You can start at a particular tag with:
vi -t <tag> | starts in the right file in the right place
Sometimes (e.g. if the system crashed while you were editing) it is
possible to recover files lost in the editor by "vi -r file". If
you just want to view a file by using vi, and you want to avoid any
change, instead of vi you can use the "view" or "vi -R" command:
the option 'readonly' will be set automatically (with ":w!" you can
override this option).
]]>the most important options<![CDATA[
ai | autoindent - in append mode after a <cr> the
| cursor will move directly below the first
| CHAR on the previous line. However, if the
| option 'lisp' is set, the cursor will align
| at the first argument to the last open list.
aw | autowrite - write at every shell escape
| (useful when compiling from within vi)
dir=<string> | directory - the directory for vi to make
| temporary files (default /tmp)
eb | errorbells - beeps when you goof
| (not on every terminal)
ic | ignorecase - no distinction between upper and
| lower cases when searching
lisp | redefine the following commands:
| "(", ")" - move backward (forward) over
| S-expressions
| "{", "}" - idem, but don't stop at atoms
| "[[", "]]" - go to previous (next) line
| beginning with a '('
| see option 'ai'
list | <lf> is shown as '$', <ht> as '^I'
magic | some metachars can be used when searching:
| ^<string> - <string> must begin the line
| <string>$ - <string> must end the line
| . - matches any char
| [a-z] - matches any char in the range
| [<string>] - matches any char in <string>
| [^<string>] - matches any char not in <string>
| <char>* - 0 or more <char>s
| \<<string>\> - <string> must be a word
nu | number - numbers before the lines
para=<string> | paragraphs - every pair of chars in <string> is
| considered a paragraph delimiter nroff macro
| (for "{" and "}").
| a <sp> preceded by a '\' indicates that
| the previous char is a single letter
| macro. ":set para=P\ bp" introduces '.P'
| and '.bp' as paragraph delimiters.
| completely empty lines and section
| boundaries are paragraph boundaries too.
redraw | the screen remains up to date
remap | if on (default), macros are repeatedly
| expanded until they are unchanged.
| Example: if 'o' is mapped to 'O', and 'O'
| is mapped to 'I', then 'o' will map to 'I'
| if "remap" is set, else it will map to 'O'.
report=<*> | vi reports whenever e.g. a delete
| or yank command affects <*> or more lines
ro | readonly - the file is not to be changed
| however, ":w!" will override this option
sect=<string> | sections - gives the section delimiters
| (for "[[" and "]]"); see option 'para',
| however a '{' as first char on a line also
| starts a section (as in C functions)
sh=<string> | shell - which program is to be used for
| shell escapes
sw=<*> | shiftwidth - gives the shiftwidth (default sw=8)
sm | showmatch - whenever you append a ')', vi
| tries to show its match by putting the cursor
| for a moment onto it (also with { })
terse | short error messages
to | timeout - if this option is set, append mode
| mappings will be interpreted only if they're
| typed fast enough
ts=<*> | tabstop - the length of a <ht>;
| warning: this is only IN the editor,
| outside of it <ht>s have their normal
| length (default ts=8)
wa | writeany - no checks when writing (dangerous)
warn | warn you when you try to quit without writing
wi=<*> | window - the default number of lines vi shows
wm=<*> | wrapmargin - when in append mode vi
| automatically puts a <lf> whenever there is
| a <sp> or <ht> within <wm> columns from the
| right margin
ws | wrapscan - when searching, the end is
| considered `stuck' to the begin of the file
:set <option> | turn <option> on
:set no<option> | turn <option> off
:set <option>=<value> | set <option> to <value>
:set | show all non-default options and their values
:set <option>? | show <option>'s value
:set all | show all options and their values
]]>contributions<![CDATA[
Maarten Litmaath <maart@cs.vu.nl>
Rich Salz <rsalz@bbn.com>
Eamonn McManus <emcmanus@cs.tcd.ie>
Diomidis Spinellis <diomidis%ecrcvax.uucp@pyramid.pyramid.com>
Blair P. Houghton <bph@buengc.bu.edu>
HTML conversion by CvL <lynx@visual.pages.de>
Patchlevel: 4
]]>