Getting along with vile ----------------------- Use Ctrl-D and Ctrl-U to scroll through this help information. Type Ctrl-O to make this the only window on the screen. Type Ctrl-K to get rid of this window. The only vile commands described herein are those not present in vi, or differing greatly from those in vi. There is a section at the bottom describing other differences between vile and vi. First, to leave vile, use any of the following: :q :quit :exit :wq (writes current buffer before quitting) :q! (quits without writing any changes!) Q :wwq or ZZ (will write all changed buffers) ^X-^C (don't know why. _They_ don't put in ":q" for _us_! Actually, if ^C is your interrupt character, this won't get you out of vile.) To get help (probably just this text), use any of: :h :help ^X-^H ^A-^H To get a complete list of all commands, type ":describe-bindings". To get a list of all commands that contain the strings "name", type ":apropos name". These lists will show all command synonyms and key sequences that are bound to the same function. You needn't type full command names -- type a few characters and hit the space bar to perform command completion. General Remarks --------------- Vile holds text in "buffers". Usually, these correspond to a file that you are editing, but not always. For instance, a buffer might be used to display this help text, or to hold the output of a shell command that you have run. Buffers have names, and these usually match the names of the files they contain. Buffers are sometimes displayed in windows. A buffer may be displayed in more than one window at the same time. There is no such thing as a hidden window. All existing windows are on the screen. There may, however, be hidden buffers, which are not currently associated with any window. All yank/delete registers (the default unnamed register, the numbered registers ("1-"9) that hold line-deletes, and the named registers ("a-"z)) are global to the editor, and not attached to any single buffer. Thus you can delete text in one buffer and put it back in another. Undo operations are attached to a buffer, not a window. Thus if you have two windows to the same buffer, and make a change in one, you can undo it in the other. Buffer manipulation commands: ----------------------------- _ Show a history list of the up to 9 most recently visited buffers. Follow this command with a digit to select that buffer, or simply repeat it ("__") as a synonym for "_1". Modified buffers are preceded by a '*' in the history list. There are many different ways to get the previous file: __ _1 1_ :p :e# ^^ (ctrl-^) (but many keyboards can't produce this) :e Edit a file. If the file is already in a buffer, that buffer will be recalled into the current window. This occurs as follows: If the name given contains no path delimiters (i.e. slashes), then it will be compared to the existing vile buffer names. Failing that comparison (or if there are any slashes in the name), the name will be stretched into an absolute path, and will be compared to the existing buffers' filenames. In either case, the matching buffer will be chosen. If there are no such matches, the file will be fetched from the filesystem. This matching technique introduces an ambiguity, since buffer names are created from the last path component of filenames. To force vile to edit a file from the current directory whose basename matches that of one that was edited elsewhere, simply preface the name with "./". For example, if you edit "../Makefile" and later attempt to edit "Makefile", vile will assume you are referring to the _buffer_ named "Makefile". To get the file in the current directory, specify "./Makefile". :e! Re-edit a file. A different filename can be given, and the buffer name will change to match it. This command is not as necessary in vile as it is in vi, since multiple buffers may have outstanding unwritten changes. ^X-e Edit the file whose pathname appears under the cursor. For example, if you are editing a makefile, you could edit one of the source files by placing the cursor at the start of its name and using this command. :n Go to the next buffer. "next" means least recently used. (The ":n file ..." version of the command is not supported.) :rename Rename the current buffer. Prompts for a new name. Does not affect the filename associated with the buffer. Use ":f" to do that. This command is useful for renaming the "[Output]" buffer, if you wish to preserve its contents, but run a new command. :b Edit a buffer. Recalls the named buffer. Does not look for a file of that name. Will find "invisible" buffers. :ki Kill a buffer. Remove the buffer and its contents from the editor. Will ask if changes to the buffer should be discarded. Currently, a buffer that is being displayed cannot be killed. * Display a list of all buffers, or make that display go away if it's already present. Leave your finger on the key, and it's easy to create and destroy the list. The buffers are numbered; the numbers correspond to the history numbers shown and used by the '_' command, described above. (If the buffer number is greater than 9, then the "nn_" form of the '_' command must be used, since '_' will only accept a single following digit.) Note that since the buffer names are displayed in order of use, the list can quickly grow "stale" -- the numbers may be meaningless if buffer switching has been done since the last display of the list. The program version is also displayed with this command. ^A-* Always display a list of all buffers. Useful for updating the list if it's already on the screen but may be out of date. Any argument will cause the list to include _all_ buffers, even those normally considered "invisible". (For example, macros are stored in "invisible" buffers.) Window manipulation commands: ----------------------------- ^X-2 Make Two windows. Splits the current window in half. This is the usual way to create a new window. ^K or ^X-0 Get rid of (Kill) this window. ^O or ^X-1 Make this the Only window on the screen. ^X-o Move to the next window. ^X-O Move to the previous window. v Make the current window smaller. V Make the current window larger. ^A-^D Scroll the next window down half a screen. ^A-^U Scroll the next window up half a screen. ^A-^E Scroll the next window up one line. ^A-^Y Scroll the next window down one line. (The previous four commands are useful when comparing two buffers. Mnemonic -- think of them as affecting the "A"lternate window.) zH zM zL These are synonyms for vi's 'z+', 'z.', and 'z-', which position the line holding the cursor at the top, middle, or bottom of the screen, respectively. ^X-^R Scroll the window right by 1/3 of a screen, or by the number of lines specified. Changes the "sideways" value. ^X-^L Scroll the window left by 1/3 of a screen, or by the number of lines specified. Changes the "sideways" value. If for some reason you can't get your screen set right via a TERM variable, try the ":screen-rows" or ":screen-columns" commands (which take their args (number of rows or columns respectively) before you type the ":"). X Window System commands: ------------------------- If you are using xvile, the following additions are available: Mouse buttons: 1 - Sets cursor position and the start of the selection. With any modifier, it will center the buffer at the cursor (this results in a primitive form of scrolling.) Motion with the button down will select text. Multiple clicks will expand the selection to cover words or the current line. 2 - Paste the current PRIMARY selection. With a modifier, it pastes at the mouse position, otherwise it pastes at the text cursor position. 3 - Extends the current selection. Additional command line arguments: -f fontname Font to use. -r Use reverse video. -d display Use an alternate display. X Resources: font Font to use. charClass Character classes for multiple click selections. The format is identical to that of xterm(1). multiClickTime How long between clicks (in milliseconds) to be accepted as a multi-click. Changing the font on the fly: Use the ':setv' command to set the $font variable: : setv $font File manipulation commands: --------------------------- The usual ":e", ":r", ":f", ":w" commands are available, though only ":e!" is available of the "!" options. The ":r" command reads the named file in after the current line. To read a file before the first line, use ":0r". The commands ":ww" and ":wwq" correspond roughly to ":w" and ":wq", but they each write all modified buffers. As in vi, ranges of lines specified by line numbers (including '.', '$', and '%' shorthands) or marks may precede these commands. Unlike vi, search patterns cannot be used as line specifiers. In addition, two non-"colon" commands have been added: ^R Prompts for a filename, and then reads it in _above_ the current line. If a register is specified (e.g. "a^R ), the file is read into that named register, but not inserted into the current buffer. ^W is a writing operator, which prompts for a filename, and writes the specified region to that file. Like all operators, if the command is repeated, as in ^W^W, then lines are affected. Use 10^W^W to write 10 lines. If a register is specified (e.g. "a^W ) then the command is _not_ an operator, but writes the specified register to the named file. Shell Access ------------ Anywhere a filename is valid, a command name is also valid, entered in the form "!shell-command". The whole line is handed to the shell, and the read or write operation is done on the commands standard input or output, as appropriate. Thus you can type ":e !date" to edit a copy of today's date. The ": !cmd" shell escape works pretty much as it does in vi. The command ":!!" will rerun the previous such shell command. The '!' operator works as a filter, as expected. In addition, the ^X-! command runs a shell command and captures its output in a specific buffer, called "[Output]". This is almost identical to ":e !cmd", except that in that case the buffer is named according to the command name. These output capture commands are most useful in conjunction with the "error finder", '^X-^X', described below. On systems supporting job control, ^Z will suspend vile. Text manipulation command: -------------------------- Remember, these are only the new or different commands. The standard vi set should still work. Undo ("u") and line-undo ("U") are available for all commands. They should be a little less capricious than their vi counterparts, since they do not share the default yank register for their operation. Also, line-undo ("U") is available until the next change anywhere in the file, rather than until you leave the line. Unfortunately, the cursor position after an undo may not always be the same as it would be in vi. The vi "global" command is present, in its non-interactive form only. So is the "substitute" command. These both look pretty different while they're being used than they do in vi, and since the searching is done right after the pattern is entered, there can be a long delay while you're trying to finish typing your complete command. You can type commands just as you would have in vi, i.e. ":g/oldpat/s//newstring/" will work. But you won't see any of the '/' characters. Try it-- you'll get the idea. Line ranges are not possible on ":g", but they are on ":s". The ":v" counterpart to ":g" is not implemented. The ":g" command can be followed by any of l (list), p (print), < (left shift), > (right shift), r (read file), d (delete), L (lower case), U (upper case), ~ (flip case), put (append yanked text), Put (prepend yanked text), s (substitute), trim (trim trailing whitespace). For example, ":g/pattern/Put" will insert the contents of the default yank register just above every line containing "pattern". The ":g" command can only be used over the entire file -- smaller regions are not permitted. The substitute command can be followed by any of 'g', a digit, or 'p', to do the substitution for all occurrences, the n'th occurrence, or to print the resulting line respectively. The ":&" and '&' commands work much as they do in vi, and repeat the last substitution. The '^A-&' command is a new operator (see below), so it can work on regions: for example use '&}' to "repeat the previous substitution over the rest of this paragraph". Operators --------- Vi has a class of commands known as "operators". Operator commands are always immediately followed by a motion command. The text affected by an operator is bounded by the initial position, and the cursor position after the motion is completed. Thus the delete operator ('d') can be followed by the word motion command ('w'), causing the next word to be deleted. The sequence "dG" will delete through the end of the file, and "d/junk" will delete to the next occurrence of the string "junk". Operators can all be "stuttered" to affect lines. Thus "dd" deletes one line, "4dd" affects 4 lines, etc. Some operators in vile can be forced to affect whole lines, though the motion wouldn't normally imply it, by using the ^X form of the command. For example, "d%" (assuming you are on a curly brace) will delete a C-style block of code. "^X-d%" will delete that same area, plus anything else on the lines containing the curly- brace endpoints. Note that some operators always affect whole lines, no matter how the motion is specified. For instance, "!w" will always filter an entire line, and not just a single word. There are several new operator commands: ^A-~ Is the operator form of the '~' command, so "^A-~~" changes the case of all characters on the current line, "^A-~w" does it to a word, "3^A-~}" does it for 3 paragraphs, etc. (In vile, the simple '~' command will take a repeat count, unlike some versions of vi. If you wish it to be an operator, rebind '~' to the "flip-til" command.) ^A-u Like ^A-~, but converts the region to upper case. ^A-l Like ^A-~, but converts the region to lower case. ^A-f Format the region based on the current fill column. The initial indentation of both the first and second lines of each "paragraph" in the region are preserved, and all subsequent lines get the second line's indentation. This makes indented/outdented paragraphs (like this one) work correctly. (This is intentionally _not_ the same behavior obtained by using "!fmt", since that behavior is obviously available elsewhere.) The usual usage of this command is "^A-f}", which reformats the current paragraph. The re-formatting begins again with each new paragraph, where a paragraph has the same boundaries used by the '{' and '}' commands -- i.e. blank lines, or lines beginning in .I .L .P .Q or .b. This make is possible to use "3^A-f}" or "^A-fG" to reformat multiple paragraphs. The reformatting operation know about both C and shell comments, and will do the "right" thing with lines that start with '#' or '*' characters. ^X-s For every occurrence of a search string, put in a replace- ment string. This is similar to "s/pattern/replacement/g" over the region. ^A-& Is an operator in vile, similar to the traditional & command. It repeats the last substitution over the specified region. (Unlike the '&' command, this one will remember trailing g, p, l, or numeric options.) ^X-d Delete the region, including the lines it starts and ends on. ^X-c Change the region, including the lines it starts and ends on. ^X-y Yank the region, including the lines it starts and ends on. ^A-t Trim trailing whitespace from all lines in the region. ^A- Convert tabs to spaces, using the current tabstop value. ^A- Convert as many spaces to tabs as appropriate. Text insertion -------------- ^X-p Causes the previously yanked or deleted text, no matter how it was obtained, to be inserted after the current line. Usually text that did not consist of whole lines where it came from is inserted immediately following the cursor. ^X-P As above, but the text is put before the current line. Thus "dw" followed by a "p" command does a normal insertion of the deleted word, whereas "^X-p" results in the word being inserted on a line by itself. R vi's overwrite mode is supported. Note that the combination of overwrite mode and the (ANSI) arrow keys can be used to give a "picture drawing" mode of operation: anything you type goes into the buffer without moving adjacent text, and you can move around using the arrow keys without leaving overwrite mode. Hint: start with a buffer full of lines that consist entirely of blanks (as opposed to blank lines). Searching --------- ^X-/ Does a forward search for the "word" located under the cursor. ^X-? Does a reverse search for the "word" located under the cursor. ^A-/ Does not do a search, but sets the search pattern to the "word" under the cursor. Useful for "picking up" a word from one buffer, and searching for it in another. The following two commands may not always be present in vile, depending on how it was built: ^X-S Incremental forward searching. As you enter the search string, the cursor is advanced to the next match with what you've typed so far. Use ^F and ^R to continue the search forward or in reverse, using the current pattern. ^X-R As above, but in reverse. Tags ---- Vile supports vi-style "tags" files. ":ta" or ":tag" allows you to enter a tagname to locate. Changes to that file and location. ^] Uses the identifier currently under the cursor as the tagname. ^T or ^X-^] or ":pop" - pops to the file and location just previous to the last tag command. When one of these commands is used, vile will look for a file named "tags" in the current directory, and load it into a hidden buffer for use during tag searches. This buffer is editable if you wish (":e tags"), but will not appear in the buffer lists. If a buffer named "tags" is already available when a tag is first requested, it will be used instead of a file called "tags", and of course will remain visible. If the ":ta" form of the command is used, tag matches must be exact, unless "taglength" is set non-zero, in which case the first tag matching that many characters will be used. If your tags file is not named "tags", simply edit it and name the buffer "tags" with the ":rename-buffer" command. ^A-o Remove all blank lines containing or immediately following the current line. Miscellaneous commands ---------------------- ^X-^X The "error finder". Goes to the next file/line error pair specified in the last buffer captured from a command's output. This buffer is usually created with the ^X-! command. For example, "-!cc -c junk.c" puts all of the compiler output into the buffer named "[Output]". Repeatedly hitting ^X-^X will position the editor at each error in turn, and will eventually start over again at the top. The "Entering directory XXX" and "Leavind directory XXX" messages that GNU make puts out with the -w switch are honored, so that files are found with the correct path. (Tip: I use the following macro to quickly grep a source directory for the string under the cursor: use-register g load-register "^X!egrep -n : *.[chs]^M" where the ^X and ^M are each single control characters, entered using ^V to escape them. Then I invoke with @g to execute.) ^X-t Set or report on the tab-stop width. To set, the spacing must precede the command, as in "4^X-t". The "set tabstop" command described below does the same thing. The status output indicates whether the buffer is currently using the local or global copy of the tabstop value. ^X-f Set the local fill-column to be used with ^A-f and auto-wrap mode on insert. The default value is 7/8's of the screen size, with a maximum of 70. Since arguments come before commands, you type: 65^X-f. The "set fillcol" command does the same thing. The status output indicates whether the buffer is currently using the local or global copy of the tabstop value. ^X-x Set encryption key. (not well tested, but hopefully not broken) CRYPT must have been on when vile was built for this to do anything. K Count prefix. The first time you type it, it is equivalent to an argument of 4 to the following command. If you repeat it, it becomes worth 16, the next time 64, etc... % In addition to finding matching braces, brackets, and parentheses, the '%' command will find matching #ifdef's and C style comments. If the cursor is on the # of "#ifdef", the '%' command will find the matching "#endif" or "#else". On an "#else" it will find "#endif", and on "#endif" it will go back up to the "#if". If the cursor is on any part of a "/*" or "*/" sequence, it will find the appropriate corresponding C comment endpoint. Editor modes ------------ Modes are associated with either windows or buffers, and are usually inherited from a set of global modes. The value of the mode will track changes to the global modes, until the "local" value is set independently of the global one. To set or change a global mode, use ":set", to remove the mode, use ":unset", ":setno", or ":set" with the modename prefixed with "no". To set and reset local modes, use ":setl", ":unsetl", ":setlno". To display modes, use ":setall", ":modes", ":set all". Local values are only shown if they differ from the global values, whether they have been independently set or not. The possible modes (with synonyms in parentheses, and a trailing B or W indicating whether the mode belongs to buffers or windows) are: autoindent (ai) During insert, newly created lines inherit their leading indent from the previous line in the buffer. (B) autosave (as) Automatic file saving. Writes the file after every 256 characters of inserted text. Other file changes are not counted. (B) cmode C-code mode.. Maintains current indentation level automatically during insert, like autoindent, above. If a line ends with a '{', then the next line indents further. If a line begins with a '}', it is lined up with the line containing its matching paired brace. If a line starts with '#' it is moved to column 1. If the global cmode is set, then the buffer's C mode is turned on automatically only for files ending in a character from c-suffixes. Additionally, if a line begins with a '#' character, it will not shift right with the '>>' command. (B) crypt Causes files to be encrypted. This is NOT compatible with the UNIX crypt(1) routines. (B) c-tabstop Will prompt for a new value for spacing of tabstops for use if a file is in cmode. (B) comments A regular expression defining commented paragraph delimiters. This is used in addition to the "paragraphs" expression (see below) when reformatting a region. The net effect is that paragraphs inside of comments are preserved when reformatting, but are not reachable with the '}' and '{' motions. (B) dos When writing the buffer, terminate lines with CR/LF pairs, rather than the usual single LF. On input, if the global DOS mode is set, then incoming CR/LF pairs are taken as line terminators, and the local DOS mode is set on the buffer if the majority of lines ended that way. If global DOS mode is _not_ set, then incoming CR characters will be visible on the screen. Setting this makes editing binary files unreliable. (B) fillcol (fc) Will prompt for a new value for the fill column, where auto-wrapping and region formatting will break lines. (B) ignorecase (ic) Text searches normally match the pattern exactly. With this set, searches are case-insensitive. (B) list (li) The buffer will be displayed with tabs and newlines made visible, instead of as whitespace. (W) magic Allow meta-characters in search strings. See the section "Regular Expressions" for more detail. (B) paragraphs A regular expression defining where the "next-paragraph" ('}') and "previous-paragraph" ('{') commands will go. (B) sections A regular expression defining where the "next-section" (']') and "previous-section" ('[') commands will go. (B) showmatch (sm) During insert, if a closing '}', ']', or ')' character is typed, the cursor will highlight the matching member of the pair for about a quarter second. (B) showmode (smd) Causes an indicator on the modeline to indicate what mode vile is currently in: insert (I), replace (R), or command (none). (B) sideways Will prompt for a new value for the sideways scroll offset, which allow display of a section of code normally off the screen to the right. Also affected by the ^X-^R and ^X-^L commands. (W) tabinsert (ti) Allow the physical insertion of tab characters into the buffer. If "notabinsert", and an attempt is made to insert a tab character by explicitly typing it or by using shiftwidth or the line shifting commands, then the appropriate number of space characters will be inserted instead. Use '^V^I' to insert a real tab, and remember that pre-existing tabs will not be affected. Use '^A-SP' to eliminate tabs from a region of text. (B) tabstop (ts) Will prompt for a new value for spacing of normal tabstops. (B) taglength (tl) Sets the significant length for tags. If non-zero, the first tag matching in that many characters will be used, otherwise, tags must match the lookup pattern exactly. This will not effect tags picked up from the cursor -- they are always matched exactly. (B) tagsrelative (tr) Causes files looked up via the tags mechanism to be found reletive to the location of the tags file, rather than the current directory. (B) terse vile produces more "status" messages than vi, which may become annoying at low baud rates. Setting terse mode will suppress many of these. (B) wrapscan (ws) Text searches will continue from past the bottom of the file to the top, and vice-versa. (B) wrapwords (ww) Similar to vi's auto-wrap mode. While inserting, words are moved to the next line if the current line gets too long. Unlike vi, wrapping is only attempted when a space is typed. (B) view View the file only. No changes are permitted. This is set automatically for the output of shell commands. (B) Special Character Expansion --------------------------- As in vi, the % and # characters typed while responding to a prompt will expand to the current or "alternate" filename. In addition, the colon character (":") expands to the identifier name under the cursor. Expansion of ! to the last command run is not implemented. The expansion can be suppressed by prefixing with a '\'. Key Rebinding ------------- The vi "map", "map!", and "abbr" commands are not currently supported. There is a key rebinding facility (if vile is built to include it), which is invoked as follows. One must know the "english" name for the command being rebound. Use ":describe-bindings" or ":apropos string" to find englishnames containing "string". Then use the command: ":bind-key englishname keyseq" where keyseq is the exact keyboard sequence (i.e. single character, or ^X or ^A followed by a single character) to which the command should be bound. In a .vilerc file, keyseq should be the printable representation of the sequence, e.g. ^A-a or ^X-S. Commands can be bound to function keys, which are either ANSI escape sequences of the form "ESC [ c" or "ESC O c". Thus the ANSI up-arrow key, which produces "ESC [ A" is bindable as 'FN-A' (and is bound to "previous-line" by default), and the 4 ANSI function keys, which produce "ESC O P" through "ESC O S" (and sometimes more) are bound to 'FN-P', 'FN-Q', 'FN-R', and 'FN-S'. Any "function" key is available on any terminal using the '#' prefix. Entering the sequence "# A" is equivalent to hitting the ANSI up-arrow in the above paragraph, and hitting "# 3" will execute whatever is bound to 'FN-3' (which is "execute-macro-3" by default. Examples: To cause the / and ? commands to perform incremental searches, use: bind-key incremental-search / bind-key reverse-incremental-search ? To make ^N and ^P switch windows instead of cause motion by lines, try: bind-key next-window ^N bind-key previous-window ^P To cause the space bar to move forward by pages, as in the "more" command, use: bind-key next-page (Space and tab can be represented with the strings: "" and "".) The englishname "rebind-key" is synonymous with "bind-key". Note that even the ^A and ^X prefix characters can be rebound, using the dummy functions "cntl_a-prefix" and "cntl_x-prefix". Even if they are rebound, however, the binding list and bind-key commands will refer to them as ^A and ^X. Note that when interactive, ^A is typed using the control key. In a file, however, it should be a caret (^) followed by a letter. Macros ------ The first type of macro in vile, is for temporary, quick macro usage, and lets you record a macro as you execute vile commands. You can then replay those keystrokes with a single key. ^X-( Begin recording a keyboard macro. The keystrokes you type are recorded, until you use ^X-). ^X-) Finish recording a keyboard macro. ^X-& Execute the keyboard macro. The vi '@' command is present as well, and can be used to execute the contents of a named register as if it were entered at the keyboard. To make this more useful, the "load-register" command will allow preloading a named register, from .vilerc file. For example: use-register a load-register ihello^[ will load register 'a' with a command to insert the word "hello". (The ^[ should be a real ESC character, entered by preceding it with ^V.) A better example, is this: use-register w load-register ":!chmod +w %^M:w^M" which makes the current file writable and writes it. (Again, use ^V to get the CR characters into the .vilerc file.) Programmed macros ----------------- Vile can also be extended (though I confess this has only been lightly used or tested) by defining macros and then binding execution of those macros to key sequences. For example, if the following lines appear in a .vilerc file: 1 store-macro 5 delete-til next-word ~endm bind-key execute-macro-1 ^A-1 then when ^A-1 is executed, 5 words will be deleted. The "-til" suffix on an englishname denotes that it is a vi operator style command, and expects to be followed by a motion command. Also, 1 store-macro 5 delete-til lines ~endm would be the equivalent of "5dd" since the word "lines" represents the stuttered 'dd' style of operation. There are some built-in variables that can be used in macros to gain access to parts of vile status, and parts of the current buffer. Built-in variables are accessed by name, prefixed with the '$' character. (User-defined variables can also be set and used; their names are prefixed with the '%' character.) $pagelen number of screen lines in use by editor $pagewid current screen width $curcol current column position of cursor $curline current line in file $cbufname current buffer name $cfilname current file name $sres current screen resolution $debug macro debugging -- set true for line by line tracing $status returns the status of the last command $palette current palette string $lastkey last keyboard char struck $version current version number (read-only) $progname returns current prog name - "vile" (read-only) $seed current random number seed $discmd display commands on command line $disinp display command line input characters $wline number of lines in current window $cwline line offset in current window $search search pattern $replace replacement pattern $match last matched magic pattern (read only) $kill some of the kill register (read only) $tpause length to pause for paren matching (pc versions) $pending typeahead pending flag (read only) $llength length of current line (read only) $char character under the cursor $line text of current line $word current "word" $identifier current punctuated word $pathname current path-like word $directory current directory $terse be terse -- suppress messages $font current font name (X version only) There are also functions available, which can act on those variables, or on hard-coded values. Operations are expressed in prefix notation, so to add to numbers you would say "&add 3 5". func no. of name args -------------- &add 2 add two numbers together &sub 2 subtraction &tim 2 multiplication &div 2 division &mod 2 modulus &neg 1 negate &cat 2 concatenate string &lef 2 left string(string, len) &rig 2 right string(string, pos) &mid 3 mid string(string, pos, len) ¬ 1 logical not &equ 2 logical equality &les 2 logical less than &gre 2 logical greater than &seq 2 string logical equality check &sle 2 string logical less than &sgr 2 string logical greater than &ind 1 evaluate indirect value &and 2 logical and &or 2 logical or &len 1 string length &upp 1 uppercase string &low 1 lower case string &tru 1 Truth of the universe logical test &asc 1 char to integer conversion &chr 1 integer to char conversion >k 0 get 1 character &rnd 1 get a random number from 1 to n &abs 1 absolute value of a number &sin 2 find the index of one string in another &env 1 retrieve a system environment var &bin 1 lookup what function name is bound to key &rd 1 is a file readable &wr 1 is a file writable To prevent vile from thinking that a failed command is an error in the macro, you can put "~force" in front of it. So, to write a macro which will run the "man" command on the identifier under the cursor, where you don't really consider it an error if the command fails, you might use: 9 store-macro ~force shell-command &cat "man " $identifier ~endm bind-key execute-macro-9 ^X-m The "screen-search-forward" command could be re-implemented as: 10 store-macro ~force search-forward $identifier ~endm bind-key execute-macro-10 ^X-/ And the "screen-search-pattern-grab" command, normally bound to ^A-/, could be implemented with: 11 store-macro set-variable $search $identifier write-message &cat "Search pattern is now " $search ~endm bind-key execute-macro-11 ^A-/ A variation, which will not find the identifier if it is a substring of another word, is: 12 store-macro ~force search-forward &cat "\\<" &cat $identifier "\\>" ~endm This works by surrounding the string with the \< and \> meta- characters. Primitive flow-of-control within a macro may be obtained with certain directives. Macro directives start with a "~" and include: ~if cond conditional execution ~else ~endif ~return Return (terminating current macro) ~goto *label Jump to a label in the current macro ~force Force macro to continue...even if command fails ~while cond Execute a loop if the condition is true ~break ~endwhile ~endm End a macro Line Labels begin with a "*" as the first nonblank char, like: *LBL01 Regular Expressions ------------------- Searches use regular expressions, which, as in vi, may be magic by default or not. There are new magic metacharacters. The code that implements the expressions is based directly on Henry Spencer's regexp code. Quoting from the original man page: "A regular expression is zero or more branches, separated by `\|'. It matches anything that matches one of the branches." "A branch is zero or more pieces, concatenated. It matches a match for the first, followed by a match for the second, etc." "A piece is an atom possibly followed by `*', `\+', or `\?'. An atom followed by `*' matches a sequence of 0 or more matches of the atom. An atom followed by `\+' matches a sequence of 1 or more matches of the atom. An atom followed by `\?' matches a match of the atom, or the null string." [i.e., `\?' matches 0 or 1 occurrences] "An atom is a regular expression in backslashed parentheses (matching a match for the regular expression), a range (see below), `.' (matching any single character), `^' (matching the null string at the beginning of the input string), `$' (matching the null string at the end of the input string), a `\' followed by a single character (matching that character), or a single character with no other significance (matching that character)." [ In addition, vile atoms may be: \< and \>, which match the beginning and end of a "word", \s and \S, which match sequences of whitespace and "darkspace", \w and \W, which match alphanumerics (including '_') and non-alphanumerics, \d and \D, which match digits and non-digits, and \p and \P, which match printable and non-printable characters (whitespace is printable). (Note: only \s will match the start or end of a line (since newlines count as whitespace), but since no vile regular expression will span more than one line, it will not _cross_ the newline.) ] "A range is a sequence of characters enclosed in `[]'. It normally matches any single character from the sequence. If the sequence begins with `^', it matches any single character not from the rest of the sequence. If two characters in the sequence are separated by `-', this is shorthand for the full list of ASCII characters between them (e.g. `[0-9]' matches any decimal digit). To include a literal `]' in the sequence, make it the first character (following a possible `^'). To include a literal `-', make it the first or last character." Additionally for vile, if "ignorecase" is set, then all literal matches, including those in character classes, are done without regard to upper and lower case. In magic mode, the following set of metacharacters must be preceded by '\' to hide their special meaning: * [ . ^ $ These characters are special if they _are_ preceded with a '\': ? + ( ) | < > If magic mode is _not_ on, only ^ and $ are special if not escaped. All of the following must be preceded with a '\' to be special, otherwise they are taken literally: ? + ( ) | * [ . < > Command files ------------- On startup (in the absence of '@' arguments -- see Invocation below), vile will attempt to read the file ".vilerc" in the current directory, and then in $HOME, and will execute the commands found therein. It is possible to nest such executions. For example, you might keep your general default settings in your home directory, and put the following lines in a .vilerc in a directory in which you prefer tabs be set to 4 spaces: source "$HOME/.vilerc" set tabstop 8 The quotes surrounding the filename part are necessary to protect the name from interpretation as a vile variable. Files can be executed at any time with the ":source" (or ":execute-file") command. There are (mostly untested) control structures available, such as ~while, ~if, etc. Refer to the source code for more information. Nesting of source'd files is limited to a depth of ten, to prevent infinite recursion. Invocation ---------- If set in the environment, the variable VILEINIT is used as a set of vile commands to initialize the editor. If unset, vile looks for files called "./.vilerc" and "$HOME/.vilerc, and executes the first one found. As a not-very-useful example of a VILEINIT sequence, the following setting will exactly recreate the default (i.e. no VILEINIT variable) behavior: export VILEINIT VILEINIT=" ~if &rd \"./.vilerc\" source \"./.vilerc\" ~else ~if &rd \"$HOME/.vilerc\" source \"$HOME/.vilerc\" ~endif ~endif" The command line options are fairly straightforward. Vile does not support running _any_ command after a '+', as does vi, but does support the commonly used '+400' to go to line 400 and '+/foo' to search for foo. Command files (see below) can be explicitly executed on startup by prefixing them with the '@' character, as in "vile @mycmds file.c". This will suppress the operation of VILEINIT and the .vilerc files. Use "vile -?" to get the usage message. Differences ------------ Of course, this really isn't vi. Some of the following differences deserve changing, others do not. The parser for the ':' commands is not very vi-like. For instance, ":e" will prompt you for a file name. Most commands remember their last argument, and will present it as the default reply to their prompt. The backspace, line kill, word kill, job control, etc. characters are not rebindable. They are, however, read from the tty settings on startup. There is no expansion of ! in filenames or shell escapes. The command ":!!" does rerun the previous shell command. Occurrences of '#' and '%' are recognized and expanded to the previous or current filename. Other punctuation (e.g. '~') may be expanded by your shell (sh, csh), since it is handed filenames for expansion if they contain any of these characters: * ? ~ [ ] $ { } Paragraph and section boundaries, for the {, }, [, and ] commands are configurable, but may not exactly match those in vi. They are expressed via regular expressions. The vi ":map", ":map!", and ":abbr" are not supported. There is no special lisp support. But then, when was the last time you heard of a lisp programmer that used vi? Of course, ex and open mode aren't there. Most, but not all, of the word-motion-with-operator and end-of-line anomalies have been recreated. One missing anomaly: In vile, "dw" on the last word of a line ending in whitespace deletes the trailing whitespace. Vi does not delete the whitespace. If you use "cd" to change directories, all buffers whose filenames do not already begin with '/' will be made absolute, by prepending the current directory name to the filename. Sample .vilerc -------------- Here's a sample .vilerc, to help get you started with the syntax and style of vile macros: set autoindent set fillcol 75 set cmode set c-tabstop 4 ; sitting on a brace, run the C code block through indent 1 store-macro write-file filter-til goto-matching-fence "indent -st" ~endm ; format the current paragraph 2 store-macro previous-paragraph filter-til next-paragraph "/usr/ucb/fmt" ~endm ; put 'fprintf(stderr,"\n");" into the file, and ; position the cursor to add to the string 3 store-macro ~force next-line beginning-of-line insert-string "fprintf(stderr, \"\\n\");\n" ; six back, because it counts the newline 6 backward-character insert-chars ~endm ; insert '> ' in front of every line from here til the end ; (this works because without argument, the default action ; for "goto-line" is to go to the end of file. 4 store-macro substitute-til goto-line "^" "> " ~endm ; the next three reproduce vile version three behavior bind-key next-window ^N bind-key previous-window ^P bind-key split-current-window ^T bind-key execute-macro-1 ^A-C bind-key execute-macro-2 ^A-F bind-key execute-macro-3 ^A-K bind-key execute-macro-4 ^A-M Note on Scrolling ----------------- If scrolling is really slow, or really ugly (the status line hops around a lot), and you're using a vt102 or compatible terminal that you think should be able to scroll okay, then the problem is almost certainly your termcap entry, which probably doesn't describe the scrolling regions adequately. Most standard termcap entries are insufficient. Here's one that should work: dy|vt102|vt-102|dec vt102:\ :do=^J:co#80:li#24:cl=50\E[;H\E[2J:\ :le=^H:bs:cm=5\E[%i%d;%dH:nd=2\E[C:up=2\E[A:\ :ce=3\E[K:cd=50\E[J:so=2\E[7m:se=2\E[m:us=2\E[4m:ue=2\E[m:\ :md=2\E[1m:mr=2\E[7m:mb=2\E[5m:me=2\E[m:is=\E[1;24r\E[24;1H:\ :rs=\E>\E[?3l\E[?4l\E[?5l\E[?7h\E[?8h:ks=\E[?1h\E=:ke=\E[?1l\E>:\ :ku=\EOA:kd=\EOB:kr=\EOC:kl=\EOD:kb=^H:\ :ho=\E[H:k1=\EOP:k2=\EOQ:k3=\EOR:k4=\EOS:pt:sr=5\EM:vt#3:\ :sc=\E7:rc=\E8:cs=\E[%i%d;%dr:vs=\E[?7l:ve=\E[?7h: The crucial entries are cs and sr -- without both of them, vile will attempt to use dl and al, and will scroll by deleting and inserting a block of lines, giving the ugly behavior you see. If you define PRETTIER_SCROLL when building vile, it will alternately delete and insert lines, instead of deleting them all and then inserting them all. This makes it look a _little_ better, but it's slower. ( The terminfo entry equivalent to the above termcap entry: vt102|vt-102|dec vt102, xon, cols#80, lines#24, vt#3, bel=^G, blink=\E[5m$<2>, bold=\E[1m$<2>, clear=\E[;H\E[2J$<50>, cnorm=\E[?7h, cr=\r, csr=\E[%i%p1%d;%p2%dr, cub1=\b, cud1=\n, cuf1=\E[C$<2>, cup=\E[%i%p1%d;%p2%dH$<5>, cuu1=\E[A$<2>, cvvis=\E[?7l, ed=\E[J$<50>, el=\E[K$<3>, home=\E[H, ht=\t, ind=\n, is2=\E[1;24r\E[24;1H, kbs=\b, kcub1=\EOD, kcud1=\EOB, kcuf1=\EOC, kcuu1=\EOA, kf1=\EOP, kf2=\EOQ, kf3=\EOR, kf4=\EOS, rc=\E8, rev=\E[7m$<2>, ri=\EM$<5>, rmkx=\E[?1l\E>, rmso=\E[m$<2>, rmul=\E[m$<2>, rs2=\E>\E[?3l\E[?4l\E[?5l\E[?7h\E[?8h, sc=\E7, sgr0=\E[m$<2>, smkx=\E[?1h\E=, smso=\E[7m$<2>, smul=\E[4m$<2>, ) Credits ------- This code has been written by a _lot_ of people. Names appearing within comments in the micro-Emacs source code are: Dave Conroy, Daniel Lawrence, John Gamble, Roger Ove, Dana Hoggatt, Jon Reid, Steve Wilhite, George Jones, Adam Fritz, D.R.Banks, Bob McNamara. In addition, some of the "ex" code is by Steve Kirkendall, author of the vi clone called "elvis". The regular expression code and documentation is by Henry Spencer. The sample .vilerc was created by Dave Lemke; he also contributed code for new features and bug fixes. Eric Krohn has done some excellent testing and enhancement. The changes to create vile from micro-Emacs were all done by Paul Fox, who can be reached at pgf@cayman.com. (By the way, this is not the same Paul Fox that did the Crisp editor.)