runtime/doc/usr_41.txt

*usr_41.txt*	For Vim version 7.0aa.  Last change: 2004 Oct 06

		     VIM USER MANUAL - by Bram Moolenaar

			      Write a Vim script


The Vim script language is used for the startup vimrc file, syntax files, and
many other things.  This chapter explains the items that can be used in a Vim
script.  There are a lot of them, thus this is a long chapter.

|41.1|	Introduction
|41.2|	Variables
|41.3|	Expressions
|41.4|	Conditionals
|41.5|	Executing an expression
|41.6|	Using functions
|41.7|	Defining a function
|41.8|	Exceptions
|41.9|	Various remarks
|41.10|	Writing a plugin
|41.11|	Writing a filetype plugin
|41.12|	Writing a compiler plugin

     Next chapter: |usr_42.txt|  Add new menus
 Previous chapter: |usr_40.txt|  Make new commands
Table of contents: |usr_toc.txt|

==============================================================================
*41.1*	Introduction					*vim-script-intro*

Your first experience with Vim scripts is the vimrc file.  Vim reads it when
it starts up and executes the commands.  You can set options to values you
prefer.  And you can use any colon command in it (commands that start with a
":"; these are sometimes referred to as Ex commands or command-line commands).
   Syntax files are also Vim scripts.  As are files that set options for a
specific file type.  A complicated macro can be defined by a separate Vim
script file.  You can think of other uses yourself.

Let's start with a simple example: >

	:let i = 1
	:while i < 5
	:  echo "count is" i
	:  let i = i + 1
	:endwhile
<
	Note:
	The ":" characters are not really needed here.  You only need to use
	them when you type a command.  In a Vim script file they can be left
	out.  We will use them here anyway to make clear these are colon
	commands and make them stand out from Normal mode commands.

The ":let" command assigns a value to a variable.  The generic form is: >

	:let {variable} = {expression}

In this case the variable name is "i" and the expression is a simple value,
the number one.
   The ":while" command starts a loop.  The generic form is: >

	:while {condition}
	:  {statements}
	:endwhile

The statements until the matching ":endwhile" are executed for as long as the
condition is true.  The condition used here is the expression "i < 5".  This
is true when the variable i is smaller than five.
   The ":echo" command prints its arguments.  In this case the string "count
is" and the value of the variable i.  Since i is one, this will print:

	count is 1 ~

Then there is another ":let i =" command.  The value used is the expression "i
+ 1".  This adds one to the variable i and assigns the new value to the same
variable.
   The output of the example code is:

	count is 1 ~
	count is 2 ~
	count is 3 ~
	count is 4 ~

	Note:
	If you happen to write a while loop that keeps on running, you can
	interrupt it by pressing CTRL-C (CTRL-Break on MS-Windows).


THREE KINDS OF NUMBERS

Numbers can be decimal, hexadecimal or octal.  A hexadecimal number starts
with "0x" or "0X".  For example "0x1f" is 31.  An octal number starts with a
zero.  "017" is 15.  Careful: don't put a zero before a decimal number, it
will be interpreted as an octal number!
   The ":echo" command always prints decimal numbers.  Example: >

	:echo 0x7f 036
<	127 30 ~

A number is made negative with a minus sign.  This also works for hexadecimal
and octal numbers.   A minus sign is also for subtraction.  Compare this with
the previous example: >

	:echo 0x7f -036
<	97 ~

White space in an expression is ignored.  However, it's recommended to use it
for separating items, to make the expression easier to read.  For example, to
avoid the confusion with a negative number, put a space between the minus sign
and the following number: >

	:echo 0x7f - 036

==============================================================================
*41.2*	Variables

A variable name consists of ASCII letters, digits and the underscore.  It
cannot start with a digit.  Valid variable names are:

	counter
	_aap3
	very_long_variable_name_with_underscores
	FuncLength
	LENGTH

Invalid names are "foo+bar" and "6var".
   These variables are global.  To see a list of currently defined variables
use this command: >

	:let

You can use global variables everywhere.  This also means that when the
variable "count" is used in one script file, it might also be used in another
file.  This leads to confusion at least, and real problems at worst.  To avoid
this, you can use a variable local to a script file by prepending "s:".  For
example, one script contains this code: >

	:let s:count = 1
	:while s:count < 5
	:  source other.vim
	:  let s:count = s:count + 1
	:endwhile

Since "s:count" is local to this script, you can be sure that sourcing the
"other.vim" script will not change this variable.  If "other.vim" also uses an
"s:count" variable, it will be a different copy, local to that script.  More
about script-local variables here: |script-variable|.

There are more kinds of variables, see |internal-variables|.  The most often
used ones are:

	b:name		variable local to a buffer
	w:name		variable local to a window
	g:name		global variable (also in a function)
	v:name		variable predefined by Vim


DELETING VARIABLES

Variables take up memory and show up in the output of the ":let" command.  To
delete a variable use the ":unlet" command.  Example: >

	:unlet s:count

This deletes the script-local variable "s:count" to free up the memory it
uses.  If you are not sure if the variable exists, and don't want an error
message when it doesn't, append !: >

	:unlet! s:count

When a script finishes, the local variables used there will not be
automatically freed.  The next time the script executes, it can still use the
old value.  Example: >

	:if !exists("s:call_count")
	:  let s:call_count = 0
	:endif
	:let s:call_count = s:call_count + 1
	:echo "called" s:call_count "times"

The "exists()" function checks if a variable has already been defined.  Its
argument is the name of the variable you want to check.  Not the variable
itself!  If you would do this: >

	:if !exists(s:call_count)

Then the value of s:call_count will be used as the name of the variable that
exists() checks.  That's not what you want.
   The exclamation mark ! negates a value.  When the value was true, it
becomes false.  When it was false, it becomes true.  You can read it as "not".
Thus "if !exists()" can be read as "if not exists()".
   What Vim calls true is anything that is not zero.  Only zero is false.


STRING VARIABLES AND CONSTANTS

So far only numbers were used for the variable value.  Strings can be used as
well.  Numbers and strings are the only two types of variables that Vim
supports.  The type is dynamic, it is set each time when assigning a value to
the variable with ":let".
   To assign a string value to a variable, you need to use a string constant.
There are two types of these.  First the string in double quotes: >

	:let name = "peter"
	:echo name
<	peter ~

If you want to include a double quote inside the string, put a backslash in
front of it: >

	:let name = "\"peter\""
	:echo name
<	"peter" ~

To avoid the need for a backslash, you can use a string in single quotes: >

	:let name = '"peter"'
	:echo name
<	"peter" ~

Inside a single-quote string all the characters are as they are.  The drawback
is that it's impossible to include a single quote.  A backslash is taken
literally as well, thus you can't use it to change the meaning of the
character after it.
   In double-quote strings it is possible to use special characters.  Here are
a few useful ones:

	\t		<Tab>
	\n		<NL>, line break
	\r		<CR>, <Enter>
	\e		<Esc>
	\b		<BS>, backspace
	\"		"
	\\		\, backslash
	\<Esc>		<Esc>
	\<C-W>		CTRL-W

The last two are just examples.  The  "\<name>" form can be used to include
the special key "name".
   See |expr-quote| for the full list of special items in a string.

==============================================================================
*41.3*	Expressions

Vim has a rich, yet simple way to handle expressions.  You can read the
definition here: |expression-syntax|.  Here we will show the most common
items.
   The numbers, strings and variables mentioned above are expressions by
themselves.  Thus everywhere an expression is expected, you can use a number,
string or variable.  Other basic items in an expression are:

	$NAME		environment variable
	&name		option
	@r		register

Examples: >

	:echo "The value of 'tabstop' is" &ts
	:echo "Your home directory is" $HOME
	:if @a > 5

The &name form can be used to save an option value, set it to a new value,
do something and restore the old value.  Example: >

	:let save_ic = &ic
	:set noic
	:/The Start/,$delete
	:let &ic = save_ic

This makes sure the "The Start" pattern is used with the 'ignorecase' option
off.  Still, it keeps the value that the user had set.


MATHEMATICS

It becomes more interesting if we combine these basic items.  Let's start with
mathematics on numbers:

	a + b		add
	a - b		subtract
	a * b		multiply
	a / b		divide
	a % b		modulo

The usual precedence is used.  Example: >

	:echo 10 + 5 * 2
<	20 ~

Grouping is done with braces.  No surprises here.  Example: >

	:echo (10 + 5) * 2
<	30 ~

Strings can be concatenated with ".".  Example: >

	:echo "foo" . "bar"
<	foobar ~

When the ":echo" command gets multiple arguments, it separates them with a
space.  In the example the argument is a single expression, thus no space is
inserted.

Borrowed from the C language is the conditional expression:

	a ? b : c

If "a" evaluates to true "b" is used, otherwise "c" is used.  Example: >

	:let i = 4
	:echo i > 5 ? "i is big" : "i is small"
<	i is small ~

The three parts of the constructs are always evaluated first, thus you could
see it work as:

	(a) ? (b) : (c)

==============================================================================
*41.4*	Conditionals

The ":if" commands executes the following statements, until the matching
":endif", only when a condition is met.  The generic form is:

	:if {condition}
	   {statements}
	:endif

Only when the expression {condition} evaluates to true (non-zero) will the
{statements} be executed.  These must still be valid commands.  If they
contain garbage, Vim won't be able to find the ":endif".
   You can also use ":else".  The generic form for this is:

	:if {condition}
	   {statements}
	:else
	   {statements}
	:endif

The second {statements} is only executed if the first one isn't.
   Finally, there is ":elseif":

	:if {condition}
	   {statements}
	:elseif {condition}
	   {statements}
	:endif

This works just like using ":else" and then "if", but without the need for an
extra ":endif".
   A useful example for your vimrc file is checking the 'term' option and
doing something depending upon its value: >

	:if &term == "xterm"
	:  " Do stuff for xterm
	:elseif &term == "vt100"
	:  " Do stuff for a vt100 terminal
	:else
	:  " Do something for other terminals
	:endif


LOGIC OPERATIONS

We already used some of them in the examples.  These are the most often used
ones:

	a == b		equal to
	a != b		not equal to
	a >  b		greater than
	a >= b		greater than or equal to
	a <  b		less than
	a <= b		less than or equal to

The result is one if the condition is met and zero otherwise.  An example: >

	:if v:version >= 600
	:  echo "congratulations"
	:else
	:  echo "you are using an old version, upgrade!"
	:endif

Here "v:version" is a variable defined by Vim, which has the value of the Vim
version.  600 is for version 6.0.  Version 6.1 has the value 601.  This is
very useful to write a script that works with multiple versions of Vim.
|v:version|

The logic operators work both for numbers and strings.  When comparing two
strings, the mathematical difference is used.  This compares byte values,
which may not be right for some languages.
   When comparing a string with a number, the string is first converted to a
number.  This is a bit tricky, because when a string doesn't look like a
number, the number zero is used.  Example: >

	:if 0 == "one"
	:  echo "yes"
	:endif

This will echo "yes", because "one" doesn't look like a number, thus it is
converted to the number zero.

For strings there are two more items:

	a =~ b		matches with
	a !~ b		does not match with

The left item "a" is used as a string.  The right item "b" is used as a
pattern, like what's used for searching.  Example: >

	:if str =~ " "
	:  echo "str contains a space"
	:endif
	:if str !~ '\.$'
	:  echo "str does not end in a full stop"
	:endif

Notice the use of a single-quote string for the pattern.  This is useful,
because backslashes need to be doubled in a double-quote string and patterns
tend to contain many backslashes.

The 'ignorecase' option is used when comparing strings.  When you don't want
that, append "#" to match case and "?" to ignore case.  Thus "==?" compares
two strings to be equal while ignoring case.  And "!~#" checks if a pattern
doesn't match, also checking the case of letters.  For the full table see
|expr-==|.


MORE LOOPING

The ":while" command was already mentioned.  Two more statements can be used
in between the ":while" and the ":endwhile":

	:continue		Jump back to the start of the while loop; the
				loop continues.
	:break			Jump forward to the ":endwhile"; the loop is
				discontinued.

Example: >

	:while counter < 40
	:  call do_something()
	:  if skip_flag
	:    continue
	:  endif
	:  if finished_flag
	:    break
	:  endif
	:  sleep 50m
	:endwhile

The ":sleep" command makes Vim take a nap.  The "50m" specifies fifty
milliseconds.  Another example is ":sleep 4", which sleeps for four seconds.

==============================================================================
*41.5*	Executing an expression

So far the commands in the script were executed by Vim directly.  The
":execute" command allows executing the result of an expression.  This is a
very powerful way to build commands and execute them.
   An example is to jump to a tag, which is contained in a variable: >

	:execute "tag " . tag_name

The "." is used to concatenate the string "tag " with the value of variable
"tag_name".  Suppose "tag_name" has the value "get_cmd", then the command that
will be executed is: >

	:tag get_cmd

The ":execute" command can only execute colon commands.  The ":normal" command
executes Normal mode commands.  However, its argument is not an expression but
the literal command characters.  Example: >

	:normal gg=G

This jumps to the first line and formats all lines with the "=" operator.
   To make ":normal" work with an expression, combine ":execute" with it.
Example: >

	:execute "normal " . normal_commands

The variable "normal_commands" must contain the Normal mode commands.
   Make sure that the argument for ":normal" is a complete command.  Otherwise
Vim will run into the end of the argument and abort the command.  For example,
if you start Insert mode, you must leave Insert mode as well.  This works: >

	:execute "normal Inew text \<Esc>"

This inserts "new text " in the current line.  Notice the use of the special
key "\<Esc>".  This avoids having to enter a real <Esc> character in your
script.

==============================================================================
*41.6*	Using functions

Vim defines many functions and provides a large amount of functionality that
way.  A few examples will be given in this section.  You can find the whole
list here: |functions|.

A function is called with the ":call" command.  The parameters are passed in
between braces, separated by commas.  Example: >

	:call search("Date: ", "W")

This calls the search() function, with arguments "Date: " and "W".  The
search() function uses its first argument as a search pattern and the second
one as flags.  The "W" flag means the search doesn't wrap around the end of
the file.

A function can be called in an expression.  Example: >

	:let line = getline(".")
	:let repl = substitute(line, '\a', "*", "g")
	:call setline(".", repl)

The getline() function obtains a line from the current file.  Its argument is
a specification of the line number.  In this case "." is used, which means the
line where the cursor is.
   The substitute() function does something similar to the ":substitute"
command.  The first argument is the string on which to perform the
substitution.  The second argument is the pattern, the third the replacement
string.  Finally, the last arguments are the flags.
   The setline() function sets the line, specified by the first argument, to a
new string, the second argument.  In this example the line under the cursor is
replaced with the result of the substitute().  Thus the effect of the three
statements is equal to: >

	:substitute/\a/*/g

Using the functions becomes more interesting when you do more work before and
after the substitute() call.


FUNCTIONS						*function-list*

There are many functions.  We will mention them here, grouped by what they are
used for.  You can find an alphabetical list here: |functions|.  Use CTRL-] on
the function name to jump to detailed help on it.

String manipulation:
	char2nr()		get ASCII value of a character
	nr2char()		get a character by its ASCII value
	escape()		escape characters in a string with a '\'
	strtrans()		translate a string to make it printable
	tolower()		turn a string to lowercase
	toupper()		turn a string to uppercase
	match()			position where a pattern matches in a string
	matchend()		position where a pattern match ends in a string
	matchstr()		match of a pattern in a string
	stridx()		first index of a short string in a long string
	strridx()		last index of a short string in a long string
	strlen()		length of a string
	substitute()		substitute a pattern match with a string
	submatch()		get a specific match in a ":substitute"
	strpart()		get part of a string
	expand()		expand special keywords
	type()			type of a variable
	iconv()			convert text from one encoding to another

Working with text in the current buffer:
	byte2line()		get line number at a specific byte count
	line2byte()		byte count at a specific line
	col()			column number of the cursor or a mark
	virtcol()		screen column of the cursor or a mark
	line()			line number of the cursor or mark
	wincol()		window column number of the cursor
	winline()		window line number of the cursor
	cursor()		position the cursor at a line/column
	getline()		get a line from the buffer
	setline()		replace a line in the buffer
	append()		append {string} below line {lnum}
	indent()		indent of a specific line
	cindent()		indent according to C indenting
	lispindent()		indent according to Lisp indenting
	nextnonblank()		find next non-blank line
	prevnonblank()		find previous non-blank line
	search()		find a match for a pattern
	searchpair()		find the other end of a start/skip/end

System functions and manipulation of files:
	browse()		put up a file requester
	glob()			expand wildcards
	globpath()		expand wildcards in a number of directories
	resolve()		find out where a shortcut points to
	fnamemodify()		modify a file name
	executable()		check if an executable program exists
	filereadable()		check if a file can be read
	filewritable()		check if a file can be written to
	isdirectory()		check if a directory exists
	getcwd()		get the current working directory
	getfsize()		get the size of a file
	getftime()		get last modification time of a file
	localtime()		get current time
	strftime()		convert time to a string
	tempname()		get the name of a temporary file
	delete()		delete a file
	rename()		rename a file
	system()		get the result of a shell command
	hostname()		name of the system

Buffers, windows and the argument list:
	argc()			number of entries in the argument list
	argidx()		current position in the argument list
	argv()			get one entry from the argument list
	bufexists()		check if a buffer exists
	buflisted()		check if a buffer exists and is listed
	bufloaded()		check if a buffer exists and is loaded
	bufname()		get the name of a specific buffer
	bufnr()			get the buffer number of a specific buffer
	winnr()			get the window number for the current window
	bufwinnr()		get the window number of a specific buffer
	winbufnr()		get the buffer number of a specific window
	getbufvar()		get a variable value from a specific buffer
	setbufvar()		set a variable in a specific buffer
	getwinvar()		get a variable value from a specific window
	setwinvar()		set a variable in a specific window

Folding:
	foldclosed()		check for a closed fold at a specific line
	foldclosedend()		like foldclosed() but return the last line
	foldlevel()		check for the fold level at a specific line
	foldtext()		generate the line displayed for a closed fold

Syntax highlighting:
	hlexists()		check if a highlight group exists
	hlID()			get ID of a highlight group
	synID()			get syntax ID at a specific position
	synIDattr()		get a specific attribute of a syntax ID
	synIDtrans()		get translated syntax ID

History:
	histadd()		add an item to a history
	histdel()		delete an item from a history
	histget()		get an item from a history
	histnr()		get highest index of a history list

Interactive:
	confirm()		let the user make a choice
	getchar()		get a character from the user
	getcharmod()		get modifiers for the last typed character
	input()			get a line from the user
	inputsecret()		get a line from the user without showing it
	inputdialog()		get a line from the user in a dialog
	inputresave		save and clear typeahead
	inputrestore()		restore typeahead

Vim server:
	serverlist()		return the list of server names
	remote_send()		send command characters to a Vim server
	remote_expr()		evaluate an expression in a Vim server
	server2client()		send a reply to a client of a Vim server
	remote_peek()		check if there is a reply from a Vim server
	remote_read()		read a reply from a Vim server
	foreground()		move the Vim window to the foreground
	remote_foreground()	move the Vim server window to the foreground

Various:
	mode()			get current editing mode
	visualmode()		last visual mode used
	hasmapto()		check if a mapping exists
	mapcheck()		check if a matching mapping exists
	maparg()		get rhs of a mapping
	exists()		check if a variable, function, etc. exists
	has()			check if a feature is supported in Vim
	cscope_connection()	check if a cscope connection exists
	did_filetype()		check if a FileType autocommand was used
	eventhandler()		check if invoked by an event handler
	getwinposx()		X position of the GUI Vim window
	getwinposy()		Y position of the GUI Vim window
	winheight()		get height of a specific window
	winwidth()		get width of a specific window
	libcall()		call a function in an external library
	libcallnr()		idem, returning a number
	getreg()		get contents of a register
	getregtype()		get type of a register
	setreg()		set contents and type of a register

==============================================================================
*41.7*	Defining a function

Vim enables you to define your own functions.  The basic function declaration
begins as follows: >

	:function {name}({var1}, {var2}, ...)
	:  {body}
	:endfunction
<
	Note:
	Function names must begin with a capital letter.

Let's define a short function to return the smaller of two numbers.  It starts
with this line: >

	:function Min(num1, num2)

This tells Vim that the function is named "Min" and it takes two arguments:
"num1" and "num2".
   The first thing you need to do is to check to see which number is smaller:
   >
	:  if a:num1 < a:num2

The special prefix "a:" tells Vim that the variable is a function argument.
Let's assign the variable "smaller" the value of the smallest number: >

	:  if a:num1 < a:num2
	:    let smaller = a:num1
	:  else
	:    let smaller = a:num2
	:  endif

The variable "smaller" is a local variable.  Variables used inside a function
are local unless prefixed by something like "g:", "a:", or "s:".

	Note:
	To access a global variable from inside a function you must prepend
	"g:" to it.  Thus "g:count" inside a function is used for the global
	variable "count", and "count" is another variable, local to the
	function.

You now use the ":return" statement to return the smallest number to the user.
Finally, you end the function: >

	:  return smaller
	:endfunction

The complete function definition is as follows: >

	:function Min(num1, num2)
	:  if a:num1 < a:num2
	:    let smaller = a:num1
	:  else
	:    let smaller = a:num2
	:  endif
	:  return smaller
	:endfunction

A user defined function is called in exactly the same way as a builtin
function.  Only the name is different.  The Min function can be used like
this: >

	:echo Min(5, 8)

Only now will the function be executed and the lines be interpreted by Vim.
If there are mistakes, like using an undefined variable or function, you will
now get an error message.  When defining the function these errors are not
detected.

When a function reaches ":endfunction" or ":return" is used without an
argument, the function returns zero.

To redefine a function that already exists, use the ! for the ":function"
command: >

	:function!  Min(num1, num2, num3)


USING A RANGE

The ":call" command can be given a line range.  This can have one of two
meanings.  When a function has been defined with the "range" keyword, it will
take care of the line range itself.
  The function will be passed the variables "a:firstline" and "a:lastline".
These will have the line numbers from the range the function was called with.
Example: >

	:function Count_words() range
	:  let n = a:firstline
	:  let count = 0
	:  while n <= a:lastline
	:    let count = count + Wordcount(getline(n))
	:    let n = n + 1
	:  endwhile
	:  echo "found " . count . " words"
	:endfunction

You can call this function with: >

	:10,30call Count_words()

It will be executed once and echo the number of words.
   The other way to use a line range is by defining a function without the
"range" keyword.  The function will be called once for every line in the
range, with the cursor in that line.  Example: >

	:function  Number()
	:  echo "line " . line(".") . " contains: " . getline(".")
	:endfunction

If you call this function with: >

	:10,15call Number()

The function will be called six times.


VARIABLE NUMBER OF ARGUMENTS

Vim enables you to define functions that have a variable number of arguments.
The following command, for instance, defines a function that must have 1
argument (start) and can have up to 20 additional arguments: >

	:function Show(start, ...)

The variable "a:1" contains the first optional argument, "a:2" the second, and
so on.  The variable "a:0" contains the number of extra arguments.
   For example: >

	:function Show(start, ...)
	:  echohl Title
	:  echo "Show is " . a:start
	:  echohl None
	:  let index = 1
	:  while index <= a:0
	:    echo "  Arg " . index . " is " . a:{index}
	:    let index = index + 1
	:  endwhile
	:  echo ""
	:endfunction

This uses the ":echohl" command to specify the highlighting used for the
following ":echo" command.  ":echohl None" stops it again.  The ":echon"
command works like ":echo", but doesn't output a line break.


LISTING FUNCTIONS

The ":function" command lists the names and arguments of all user-defined
functions: >

	:function
<	function Show(start, ...) ~
	function GetVimIndent() ~
	function SetSyn(name) ~

To see what a function does, use its name as an argument for ":function": >

	:function SetSyn
<	1     if &syntax == '' ~
	2       let &syntax = a:name ~
	3     endif ~
	   endfunction ~


DEBUGGING

The line number is useful for when you get an error message or when debugging.
See |debug-scripts| about debugging mode.
   You can also set the 'verbose' option to 12 or higher to see all function
calls.  Set it to 15 or higher to see every executed line.


DELETING A FUNCTION

To delete the Show() function: >

	:delfunction Show

You get an error when the function doesn't exist.

==============================================================================
*41.8*	Exceptions

Let's start with an example: >

	:try
	:   read ~/templates/pascal.tmpl
	:catch /E484:/
	:   echo "Sorry, the Pascal template file cannot be found."
	:endtry

The ":read" command will fail if the file does not exist.  Instead of
generating an error message, this code catches the error and gives the user a
nice message instead.

For the commands in between ":try" and ":endtry" errors are turned into
exceptions.  An exception is a string.  In the case of an error the string
contains the error message.  And every error message has a number.  In this
case, the error we catch contains "E484:".  This number is guaranteed to stay
the same (the text may change, e.g., it may be translated).

When the ":read" command causes another error, the pattern "E484:" will not
match in it.  Thus this exception will not be caught and result in the usual
error message.

You might be tempted to do this: >

	:try
	:   read ~/templates/pascal.tmpl
	:catch
	:   echo "Sorry, the Pascal template file cannot be found."
	:endtry

This means all errors are caught.  But then you will not see errors that are
useful, such as "E21: Cannot make changes, 'modifiable' is off".

Another useful mechanism is the ":finally" command: >

	:let tmp = tempname()
	:try
	:   exe ".,$write " . tmp
	:   exe "!filter " . tmp
	:   .,$delete
	:   exe "$read " . tmp
	:finally
	:   call delete(tmp)
	:endtry

This filters the lines from the cursor until the end of the file through the
"filter" command, which takes a file name argument.  No matter if the
filtering works, something goes wrong in between ":try" and ":finally" or the
user cancels the filtering by pressing CTRL-C, the "call delete(tmp)" is
always executed.  This makes sure you don't leave the temporary file behind.

More information about exception handling can be found in the reference
manual: |exception-handling|.

==============================================================================
*41.9*	Various remarks

Here is a summary of items that apply to Vim scripts.  They are also mentioned
elsewhere, but form a nice checklist.

The end-of-line character depends on the system.  For Unix a single <NL>
character is used.  For MS-DOS, Windows, OS/2 and the like, <CR><LF> is used.
This is important when using mappings that end in a <CR>.  See |:source_crnl|.


WHITE SPACE

Blank lines are allowed and ignored.

Leading whitespace characters (blanks and TABs) are always ignored.  The
whitespaces between parameters (e.g. between the 'set' and the 'cpoptions' in
the example below) are reduced to one blank character and plays the role of a
separator, the whitespaces after the last (visible) character may or may not
be ignored depending on the situation, see below.

For a ":set" command involving the "=" (equal) sign, such as in: >

	:set cpoptions    =aABceFst

the whitespace immediately before the "=" sign is ignored.  But there can be
no whitespace after the "=" sign!

To include a whitespace character in the value of an option, it must be
escaped by a "\" (backslash)  as in the following example: >

	:set tags=my\ nice\ file

The same example written as >

	:set tags=my nice file

will issue an error, because it is interpreted as: >

	:set tags=my
	:set nice
	:set file


COMMENTS

The character " (the double quote mark) starts a comment.  Everything after
and including this character until the end-of-line is considered a comment and
is ignored, except for commands that don't consider comments, as shown in
examples below.  A comment can start on any character position on the line.

There is a little "catch" with comments for some commands.  Examples: >

	:abbrev dev development		" shorthand
	:map <F3> o#include		" insert include
	:execute cmd			" do it
	:!ls *.c			" list C files

The abbreviation 'dev' will be expanded to 'development     " shorthand'.  The
mapping of <F3> will actually be the whole line after the 'o# ....' including
the '" insert include'.  The "execute" command will give an error.  The "!"
command will send everything after it to the shell, causing an error for an
unmatched '"' character.
   There can be no comment after ":map", ":abbreviate", ":execute" and "!"
commands (there are a few more commands with this restriction).  For the
":map", ":abbreviate" and ":execute" commands there is a trick: >

	:abbrev dev development|" shorthand
	:map <F3> o#include|" insert include
	:execute cmd			|" do it

With the '|' character the command is separated from the next one.  And that
next command is only a comment.

Notice that there is no white space before the '|' in the abbreviation and
mapping.  For these commands, any character until the end-of-line or '|' is
included.  As a consequence of this behavior, you don't always see that
trailing whitespace is included: >

	:map <F4> o#include

To avoid these problems, you can set the 'list' option when editing vimrc
files.


PITFALLS

Even bigger problem arises in the following example: >

	:map ,ab o#include
	:unmap ,ab

Here the unmap command will not work, because it tries to unmap ",ab ".  This
does not exist as a mapped sequence.  An error will be issued, which is very
hard to identify, because the ending whitespace character in ":unmap ,ab " is
not visible.

And this is the same as what happens when one uses a comment after an 'unmap'
command: >

	:unmap ,ab     " comment

Here the comment part will be ignored.  However, Vim will try to unmap
',ab     ', which does not exist.  Rewrite it as: >

	:unmap ,ab|    " comment


RESTORING THE VIEW

Sometimes you want to make a change and go back to where cursor was.
Restoring the relative position would also be nice, so that the same line
appears at the top of the window.
   This example yanks the current line, puts it above the first line in the
file and then restores the view: >

	map ,p ma"aYHmbgg"aP`bzt`a

What this does: >
	ma"aYHmbgg"aP`bzt`a
<	ma			set mark a at cursor position
	  "aY			yank current line into register a
	     Hmb		go to top line in window and set mark b there
		gg		go to first line in file
		  "aP		put the yanked line above it
		     `b		go back to top line in display
		       zt	position the text in the window as before
			 `a	go back to saved cursor position


PACKAGING

To avoid your function names to interfere with functions that you get from
others, use this scheme:
- Prepend a unique string before each function name.  I often use an
  abbreviation.  For example, "OW_" is used for the option window functions.
- Put the definition of your functions together in a file.  Set a global
  variable to indicate that the functions have been loaded.  When sourcing the
  file again, first unload the functions.
Example: >

	" This is the XXX package

	if exists("XXX_loaded")
	  delfun XXX_one
	  delfun XXX_two
	endif

	function XXX_one(a)
		... body of function ...
	endfun

	function XXX_two(b)
		... body of function ...
	endfun

	let XXX_loaded = 1

==============================================================================
*41.10*	Writing a plugin				*write-plugin*

You can write a Vim script in such a way that many people can use it.  This is
called a plugin.  Vim users can drop your script in their plugin directory and
use its features right away |add-plugin|.

There are actually two types of plugins:

  global plugins: For all types of files.
filetype plugins: Only for files of a specific type.

In this section the first type is explained.  Most items are also relevant for
writing filetype plugins.  The specifics for filetype plugins are in the next
section |write-filetype-plugin|.


NAME

First of all you must choose a name for your plugin.  The features provided
by the plugin should be clear from its name.  And it should be unlikely that
someone else writes a plugin with the same name but which does something
different.  And please limit the name to 8 characters, to avoid problems on
old Windows systems.

A script that corrects typing mistakes could be called "typecorr.vim".  We
will use it here as an example.

For the plugin to work for everybody, it should follow a few guidelines.  This
will be explained step-by-step.  The complete example plugin is at the end.


BODY

Let's start with the body of the plugin, the lines that do the actual work: >

 14	iabbrev teh the
 15	iabbrev otehr other
 16	iabbrev wnat want
 17	iabbrev synchronisation
 18		\ synchronization
 19	let s:count = 4

The actual list should be much longer, of course.

The line numbers have only been added to explain a few things, don't put them
in your plugin file!


HEADER

You will probably add new corrections to the plugin and soon have several
versions laying around.  And when distributing this file, people will want to
know who wrote this wonderful plugin and where they can send remarks.
Therefore, put a header at the top of your plugin: >

  1	" Vim global plugin for correcting typing mistakes
  2	" Last Change:	2000 Oct 15
  3	" Maintainer:	Bram Moolenaar <[email protected]>

About copyright and licensing: Since plugins are very useful and it's hardly
worth restricting their distribution, please consider making your plugin
either public domain or use the Vim |license|.  A short note about this near
the top of the plugin should be sufficient.  Example: >

  4	" License:	This file is placed in the public domain.


LINE CONTINUATION, AVOIDING SIDE EFFECTS		*use-cpo-save*

In line 18 above, the line-continuation mechanism is used |line-continuation|.
Users with 'compatible' set will run into trouble here, they will get an error
message.  We can't just reset 'compatible', because that has a lot of side
effects.  To avoid this, we will set the 'cpoptions' option to its Vim default
value and restore it later.  That will allow the use of line-continuation and
make the script work for most people.  It is done like this: >

 11	let s:save_cpo = &cpo
 12	set cpo&vim
 ..
 42	let &cpo = s:save_cpo

We first store the old value of 'cpoptions' in the s:save_cpo variable.  At
the end of the plugin this value is restored.

Notice that a script-local variable is used |s:var|.  A global variable could
already be in use for something else.  Always use script-local variables for
things that are only used in the script.


NOT LOADING

It's possible that a user doesn't always want to load this plugin.  Or the
system administrator has dropped it in the system-wide plugin directory, but a
user has his own plugin he wants to use.  Then the user must have a chance to
disable loading this specific plugin.  This will make it possible: >

  6	if exists("loaded_typecorr")
  7	  finish
  8	endif
  9	let loaded_typecorr = 1

This also avoids that when the script is loaded twice it would cause error
messages for redefining functions and cause trouble for autocommands that are
added twice.


MAPPING

Now let's make the plugin more interesting: We will add a mapping that adds a
correction for the word under the cursor.  We could just pick a key sequence
for this mapping, but the user might already use it for something else.  To
allow the user to define which keys a mapping in a plugin uses, the <Leader>
item can be used: >

 22	  map <unique> <Leader>a  <Plug>TypecorrAdd

The "<Plug>TypecorrAdd" thing will do the work, more about that further on.

The user can set the "mapleader" variable to the key sequence that he wants
this mapping to start with.  Thus if the user has done: >

	let mapleader = "_"

the mapping will define "_a".  If the user didn't do this, the default value
will be used, which is a backslash.  Then a map for "\a" will be defined.

Note that <unique> is used, this will cause an error message if the mapping
already happened to exist. |:map-<unique>|

But what if the user wants to define his own key sequence?  We can allow that
with this mechanism: >

 21	if !hasmapto('<Plug>TypecorrAdd')
 22	  map <unique> <Leader>a  <Plug>TypecorrAdd
 23	endif

This checks if a mapping to "<Plug>TypecorrAdd" already exists, and only
defines the mapping from "<Leader>a" if it doesn't.  The user then has a
chance of putting this in his vimrc file: >

	map ,c  <Plug>TypecorrAdd

Then the mapped key sequence will be ",c" instead of "_a" or "\a".


PIECES

If a script gets longer, you often want to break up the work in pieces.  You
can use functions or mappings for this.  But you don't want these functions
and mappings to interfere with the ones from other scripts.  For example, you
could define a function Add(), but another script could try to define the same
function.  To avoid this, we define the function local to the script by
prepending it with "s:".

We will define a function that adds a new typing correction: >

 30	function s:Add(from, correct)
 31	  let to = input("type the correction for " . a:from . ": ")
 32	  exe ":iabbrev " . a:from . " " . to
 ..
 36	endfunction

Now we can call the function s:Add() from within this script.  If another
script also defines s:Add(), it will be local to that script and can only
be called from the script it was defined in.  There can also be a global Add()
function (without the "s:"), which is again another function.

<SID> can be used with mappings.  It generates a script ID, which identifies
the current script.  In our typing correction plugin we use it like this: >

 24	noremap <unique> <script> <Plug>TypecorrAdd  <SID>Add
 ..
 28	noremap <SID>Add  :call <SID>Add(expand("<cword>"), 1)<CR>

Thus when a user types "\a", this sequence is invoked: >

	\a  ->  <Plug>TypecorrAdd  ->  <SID>Add  ->  :call <SID>Add()

If another script would also map <SID>Add, it would get another script ID and
thus define another mapping.

Note that instead of s:Add() we use <SID>Add() here.  That is because the
mapping is typed by the user, thus outside of the script.  The <SID> is
translated to the script ID, so that Vim knows in which script to look for
the Add() function.

This is a bit complicated, but it's required for the plugin to work together
with other plugins.  The basic rule is that you use <SID>Add() in mappings and
s:Add() in other places (the script itself, autocommands, user commands).

We can also add a menu entry to do the same as the mapping: >

 26	noremenu <script> Plugin.Add\ Correction      <SID>Add

The "Plugin" menu is recommended for adding menu items for plugins.  In this
case only one item is used.  When adding more items, creating a submenu is
recommended.  For example, "Plugin.CVS" could be used for a plugin that offers
CVS operations "Plugin.CVS.checkin", "Plugin.CVS.checkout", etc.

Note that in line 28 ":noremap" is used to avoid that any other mappings cause
trouble.  Someone may have remapped ":call", for example.  In line 24 we also
use ":noremap", but we do want "<SID>Add" to be remapped.  This is why
"<script>" is used here.  This only allows mappings which are local to the
script. |:map-<script>|  The same is done in line 26 for ":noremenu".
|:menu-<script>|


<SID> AND <Plug>					*using-<Plug>*

Both <SID> and <Plug> are used to avoid that mappings of typed keys interfere
with mappings that are only to be used from other mappings.  Note the
difference between using <SID> and <Plug>:

<Plug>	is visible outside of the script.  It is used for mappings which the
	user might want to map a key sequence to.  <Plug> is a special code
	that a typed key will never produce.
	To make it very unlikely that other plugins use the same sequence of
	characters, use this structure: <Plug> scriptname mapname
	In our example the scriptname is "Typecorr" and the mapname is "Add".
	This results in "<Plug>TypecorrAdd".  Only the first character of
	scriptname and mapname is uppercase, so that we can see where mapname
	starts.

<SID>	is the script ID, a unique identifier for a script.
	Internally Vim translates <SID> to "<SNR>123_", where "123" can be any
	number.  Thus a function "<SID>Add()" will have a name "<SNR>11_Add()"
	in one script, and "<SNR>22_Add()" in another.  You can see this if
	you use the ":function" command to get a list of functions.  The
	translation of <SID> in mappings is exactly the same, that's how you
	can call a script-local function from a mapping.


USER COMMAND

Now let's add a user command to add a correction: >

 38	if !exists(":Correct")
 39	  command -nargs=1  Correct  :call s:Add(<q-args>, 0)
 40	endif

The user command is defined only if no command with the same name already
exists.  Otherwise we would get an error here.  Overriding the existing user
command with ":command!" is not a good idea, this would probably make the user
wonder why the command he defined himself doesn't work.  |:command|


SCRIPT VARIABLES

When a variable starts with "s:" it is a script variable.  It can only be used
inside a script.  Outside the script it's not visible.  This avoids trouble
with using the same variable name in different scripts.  The variables will be
kept as long as Vim is running.  And the same variables are used when sourcing
the same script again. |s:var|

The fun is that these variables can also be used in functions, autocommands
and user commands that are defined in the script.  In our example we can add
a few lines to count the number of corrections: >

 19	let s:count = 4
 ..
 30	function s:Add(from, correct)
 ..
 34	  let s:count = s:count + 1
 35	  echo s:count . " corrections now"
 36	endfunction

First s:count is initialized to 4 in the script itself.  When later the
s:Add() function is called, it increments s:count.  It doesn't matter from
where the function was called, since it has been defined in the script, it
will use the local variables from this script.


THE RESULT

Here is the resulting complete example: >

  1	" Vim global plugin for correcting typing mistakes
  2	" Last Change:	2000 Oct 15
  3	" Maintainer:	Bram Moolenaar <[email protected]>
  4	" License:	This file is placed in the public domain.
  5
  6	if exists("loaded_typecorr")
  7	  finish
  8	endif
  9	let loaded_typecorr = 1
 10
 11	let s:save_cpo = &cpo
 12	set cpo&vim
 13
 14	iabbrev teh the
 15	iabbrev otehr other
 16	iabbrev wnat want
 17	iabbrev synchronisation
 18		\ synchronization
 19	let s:count = 4
 20
 21	if !hasmapto('<Plug>TypecorrAdd')
 22	  map <unique> <Leader>a  <Plug>TypecorrAdd
 23	endif
 24	noremap <unique> <script> <Plug>TypecorrAdd  <SID>Add
 25
 26	noremenu <script> Plugin.Add\ Correction      <SID>Add
 27
 28	noremap <SID>Add  :call <SID>Add(expand("<cword>"), 1)<CR>
 29
 30	function s:Add(from, correct)
 31	  let to = input("type the correction for " . a:from . ": ")
 32	  exe ":iabbrev " . a:from . " " . to
 33	  if a:correct | exe "normal viws\<C-R>\" \b\e" | endif
 34	  let s:count = s:count + 1
 35	  echo s:count . " corrections now"
 36	endfunction
 37
 38	if !exists(":Correct")
 39	  command -nargs=1  Correct  :call s:Add(<q-args>, 0)
 40	endif
 41
 42	let &cpo = s:save_cpo

Line 33 wasn't explained yet.  It applies the new correction to the word under
the cursor.  The |:normal| command is used to use the new abbreviation.  Note
that mappings and abbreviations are expanded here, even though the function
was called from a mapping defined with ":noremap".

Using "unix" for the 'fileformat' option is recommended.  The Vim scripts will
then work everywhere.  Scripts with 'fileformat' set to "dos" do not work on
Unix.  Also see |:source_crnl|.  To be sure it is set right, do this before
writing the file: >

	:set fileformat=unix


DOCUMENTATION						*write-local-help*

It's a good idea to also write some documentation for your plugin.  Especially
when its behavior can be changed by the user.  See |add-local-help| for how
they are installed.

Here is a simple example for a plugin help file, called "typecorr.txt": >

  1	*typecorr.txt*	Plugin for correcting typing mistakes
  2
  3	If you make typing mistakes, this plugin will have them corrected
  4	automatically.
  5
  6	There are currently only a few corrections.  Add your own if you like.
  7
  8	Mappings:
  9	<Leader>a   or   <Plug>TypecorrAdd
 10		Add a correction for the word under the cursor.
 11
 12	Commands:
 13	:Correct {word}
 14		Add a correction for {word}.
 15
 16							*typecorr-settings*
 17	This plugin doesn't have any settings.

The first line is actually the only one for which the format matters.  It will
be extracted from the help file to be put in the "LOCAL ADDITIONS:" section of
help.txt |local-additions|.  The first "*" must be in the first column of the
first line.  After adding your help file do ":help" and check that the entries
line up nicely.

You can add more tags inside ** in your help file.  But be careful not to use
existing help tags.  You would probably use the name of your plugin in most of
them, like "typecorr-settings" in the example.

Using references to other parts of the help in || is recommended.  This makes
it easy for the user to find associated help.


FILETYPE DETECTION					*plugin-filetype*

If your filetype is not already detected by Vim, you should create a filetype
detection snippet in a separate file.  It is usually in the form of an
autocommand that sets the filetype when the file name matches a pattern.
Example: >

	au BufNewFile,BufRead *.foo			set filetype=foofoo

Write this single-line file as "ftdetect/foofoo.vim" in the first directory
that appears in 'runtimepath'.  For Unix that would be
"~/.vim/ftdetect/foofoo.vim".  The convention is to use the name of the
filetype for the script name.

You can make more complicated checks if you like, for example to inspect the
contents of the file to recognize the language.  Also see |new-filetype|.


SUMMARY							*plugin-special*

Summary of special things to use in a plugin:

s:name			Variables local to the script.

<SID>			Script-ID, used for mappings and functions local to
			the script.

hasmapto()		Function to test if the user already defined a mapping
			for functionality the script offers.

<Leader>		Value of "mapleader", which the user defines as the
			keys that plugin mappings start with.

:map <unique>		Give a warning if a mapping already exists.

:noremap <script>	Use only mappings local to the script, not global
			mappings.

exists(":Cmd")		Check if a user command already exists.

==============================================================================
*41.11*	Writing a filetype plugin	*write-filetype-plugin* *ftplugin*

A filetype plugin is like a global plugin, except that it sets options and
defines mappings for the current buffer only.  See |add-filetype-plugin| for
how this type of plugin is used.

First read the section on global plugins above |41.10|.  All that is said there
also applies to filetype plugins.  There are a few extras, which are explained
here.  The essential thing is that a filetype plugin should only have an
effect on the current buffer.


DISABLING

If you are writing a filetype plugin to be used by many people, they need a
chance to disable loading it.  Put this at the top of the plugin: >

	" Only do this when not done yet for this buffer
	if exists("b:did_ftplugin")
	  finish
	endif
	let b:did_ftplugin = 1

This also needs to be used to avoid that the same plugin is executed twice for
the same buffer (happens when using an ":edit" command without arguments).

Now users can disable loading the default plugin completely by making a
filetype plugin with only this line: >

	let b:did_ftplugin = 1

This does require that the filetype plugin directory comes before $VIMRUNTIME
in 'runtimepath'!

If you do want to use the default plugin, but overrule one of the settings,
you can write the different setting in a script: >

	setlocal textwidth=70

Now write this in the "after" directory, so that it gets sourced after the
distributed "vim.vim" ftplugin |after-directory|.  For Unix this would be
"~/.vim/after/ftplugin/vim.vim".  Note that the default plugin will have set
"b:did_ftplugin", but it is ignored here.


OPTIONS

To make sure the filetype plugin only affects the current buffer use the >

	:setlocal

command to set options.  And only set options which are local to a buffer (see
the help for the option to check that).  When using |:setlocal| for global
options or options local to a window, the value will change for many buffers,
and that is not what a filetype plugin should do.

When an option has a value that is a list of flags or items, consider using
"+=" and "-=" to keep the existing value.  Be aware that the user may have
changed an option value already.  First resetting to the default value and
then changing it often a good idea.  Example: >

	:setlocal formatoptions& formatoptions+=ro


MAPPINGS

To make sure mappings will only work in the current buffer use the >

	:map <buffer>

command.  This needs to be combined with the two-step mapping explained above.
An example of how to define functionality in a filetype plugin: >

	if !hasmapto('<Plug>JavaImport')
	  map <buffer> <unique> <LocalLeader>i <Plug>JavaImport
	endif
	noremap <buffer> <unique> <Plug>JavaImport oimport ""<Left><Esc>

|hasmapto()| is used to check if the user has already defined a map to
<Plug>JavaImport.  If not, then the filetype plugin defines the default
mapping.  This starts with |<LocalLeader>|, which allows the user to select
the key(s) he wants filetype plugin mappings to start with.  The default is a
backslash.
"<unique>" is used to give an error message if the mapping already exists or
overlaps with an existing mapping.
|:noremap| is used to avoid that any other mappings that the user has defined
interferes.  You might want to use ":noremap <script>" to allow remapping
mappings defined in this script that start with <SID>.

The user must have a chance to disable the mappings in a filetype plugin,
without disabling everything.  Here is an example of how this is done for a
plugin for the mail filetype: >

	" Add mappings, unless the user didn't want this.
	if !exists("no_plugin_maps") && !exists("no_mail_maps")
	  " Quote text by inserting "> "
	  if !hasmapto('<Plug>MailQuote')
	    vmap <buffer> <LocalLeader>q <Plug>MailQuote
	    nmap <buffer> <LocalLeader>q <Plug>MailQuote
	  endif
	  vnoremap <buffer> <Plug>MailQuote :s/^/> /<CR>
	  nnoremap <buffer> <Plug>MailQuote :.,$s/^/> /<CR>
	endif

Two global variables are used:
no_plugin_maps		disables mappings for all filetype plugins
no_mail_maps		disables mappings for a specific filetype


USER COMMANDS

To add a user command for a specific file type, so that it can only be used in
one buffer, use the "-buffer" argument to |:command|.  Example: >

	:command -buffer  Make  make %:r.s


VARIABLES

A filetype plugin will be sourced for each buffer of the type it's for.  Local
script variables |s:var| will be shared between all invocations.  Use local
buffer variables |b:var| if you want a variable specifically for one buffer.


FUNCTIONS

When defining a function, this only needs to be done once.  But the filetype
plugin will be sourced every time a file with this filetype will be opened.
This construct make sure the function is only defined once: >

	:if !exists("*s:Func")
	:  function s:Func(arg)
	:    ...
	:  endfunction
	:endif
<

UNDO							*undo_ftplugin*

When the user does ":setfiletype xyz" the effect of the previous filetype
should be undone.  Set the b:undo_ftplugin variable to the commands that will
undo the settings in your filetype plugin.  Example: >

	let b:undo_ftplugin = "setlocal fo< com< tw< commentstring<"
		\ . "| unlet b:match_ignorecase b:match_words b:match_skip"

Using ":setlocal" with "<" after the option name resets the option to its
global value.  That is mostly the best way to reset the option value.

This does require removing the "C" flag from 'cpoptions' to allow line
continuation, as mentioned above |use-cpo-save|.


FILE NAME

The filetype must be included in the file name |ftplugin-name|.  Use one of
these three forms:

	.../ftplugin/stuff.vim
	.../ftplugin/stuff_foo.vim
	.../ftplugin/stuff/bar.vim

"stuff" is the filetype, "foo" and "bar" are arbitrary names.


SUMMARY							*ftplugin-special*

Summary of special things to use in a filetype plugin:

<LocalLeader>		Value of "maplocalleader", which the user defines as
			the keys that filetype plugin mappings start with.

:map <buffer>		Define a mapping local to the buffer.

:noremap <script>	Only remap mappings defined in this script that start
			with <SID>.

:setlocal		Set an option for the current buffer only.

:command -buffer	Define a user command local to the buffer.

exists("*s:Func")	Check if a function was already defined.

Also see |plugin-special|, the special things used for all plugins.

==============================================================================
*41.12*	Writing a compiler plugin		*write-compiler-plugin*

A compiler plugin sets options for use with a specific compiler.  The user can
load it with the |:compiler| command.  The main use is to set the
'errorformat' and 'makeprg' options.

Easiest is to have a look at examples.  This command will edit all the default
compiler plugins: >

	:next $VIMRUNTIME/compiler/*.vim

Use |:next| to go to the next plugin file.

There are two special items about these files.  First is a mechanism to allow
a user to overrule or add to the default file.  The default files start with: >

	:if exists("current_compiler")
	:  finish
	:endif
	:let current_compiler = "mine"

When you write a compiler file and put it in your personal runtime directory
(e.g., ~/.vim/compiler for Unix), you set the "current_compiler" variable to
make the default file skip the settings.

The second mechanism is to use ":set" for ":compiler!" and ":setlocal" for
":compiler".  Vim defines the ":CompilerSet" user command for this.  However,
older Vim versions don't, thus your plugin should define it then.  This is an
example: >

  if exists(":CompilerSet") != 2
    command -nargs=* CompilerSet setlocal <args>
  endif
  CompilerSet errorformat&		" use the default 'errorformat'
  CompilerSet makeprg=nmake

When you write a compiler plugin for the Vim distribution or for a system-wide
runtime directory, use the mechanism mentioned above.  When
"current_compiler" was already set by a user plugin nothing will be done.

When you write a compiler plugin to overrule settings from a default plugin,
don't check "current_compiler".  This plugin is supposed to be loaded
last, thus it should be in a directory at the end of 'runtimepath'.  For Unix
that could be ~/.vim/after/compiler.

==============================================================================

Next chapter: |usr_42.txt|  Add new menus

Copyright: see |manual-copyright|  vim:tw=78:ts=8:ft=help:norl: