texinode(Programmable Completion)(Completion Widgets)(Zsh Line Editor)(Top)
chapter(Programmable Completion)
cindex(completion, programmable)
cindex(completion, controlling)
findex(compctl)
sect(Description)
startlist()
list(tt(compctl) [ tt(-CDT) ] var(options) [ var(command) ... ])
list(tt(compctl) [ tt(-CDT) ] var(options) \
  [ tt(-x) var(pattern) var(options) tt(-) ... tt(--) ] \
  [ tt(PLUS()) var(options) [ tt(-x) ... tt(--) ] ... [tt(PLUS())] ] \
  [ var(command) ... ])
list(tt(compctl) tt(-M) var(match-specs) ...)
list(tt(compctl) tt(-L) [ tt(-CDTM) ] [ var(command) ... ])
list(tt(compctl) tt(PLUS()) var(command) ...)
endlist()

Control the editor's completion behavior according to the supplied set
of var(options).  Various editing commands, notably
tt(expand-or-complete-word), usually bound to tab, will
attempt to complete a word typed by the user, while others, notably
tt(delete-char-or-list), usually bound to ^D in EMACS editing
mode, list the possibilities; tt(compctl) controls what those
possibilities are.  They may for example be filenames (the most common
case, and hence the default), shell variables, or words from a
user-specified list.
startmenu()
menu(Command Flags)
menu(Option Flags)
menu(Alternative Completion)
menu(Extended Completion)
menu(Matching Control)
menu(Example)
endmenu()

texinode(Command Flags)(Option Flags)()(Programmable Completion)
sect(Command Flags)
Completion of the arguments of a command may be different for each
command or may use the default.  The behavior when completing the
command word itself may also be separately specified.  These
correspond to the following flags and arguments, all of which (except
for tt(-L)) may be combined with any combination of the
var(options) described subsequently in noderef(Option Flags):

startitem()
item(var(command) ...)(
controls completion for the named commands, which must be listed last
on the command line.  If completion is attempted for a command with a
pathname containing slashes and no completion definition is found, the
search is retried with the last pathname component. If the command starts
with a tt(=), completion is tried with the pathname of the command.

Any of the var(command) strings may be patterns of the form normally
used for filename generation.  These should be be quoted to protect them
from immediate expansion; for example the command string tt('foo*')
arranges for completion of the words of any command beginning with
tt(foo).  When completion is attempted, all pattern completions are
tried in the reverse order of their definition until one matches.  By
default, completion then procedes as normal, i.e. the shell will try to
generate more matches for the specific command on the command line; this
can be overridden by including tt(-tn) in the flags for the pattern
completion.

Note that aliases
are expanded before the command name is determined unless the
tt(COMPLETE_ALIASES) option is set.  Commands may not be combined
with the tt(-C), tt(-D) or tt(-T) flags.
)
item(tt(-C))(
controls completion when the command word itself is being completed.
If no tt(compctl -C) command has been issued,  the names of any
executable command (whether in the path or specific to the shell, such
as aliases or functions) are completed.
)
item(tt(-D))(
controls default completion behavior for the arguments of commands not
assigned any special behavior.  If no tt(compctl -D) command has
been issued, filenames are completed.
)
item(tt(-T))(
supplies completion flags to be used before any other processing is
done, even before processing for tt(compctl)s defined for specific
commands.  This is especially useful when combined with extended
completion (the tt(-x) flag, see noderef(Extended Completion) below).
Using this flag you can define default behavior
which will apply to all commands without exception, or you can alter
the standard behavior for all commands.  For example, if your access
to the user database is too slow and/or it contains too many users (so
that completion after `tt(~)' is too slow to be usable), you can use

indent(
tt(compctl -T -x 's[~] C[0,[^/]#]' -k friends -S/ -tn)
)

to complete the strings in the array tt(friends) after a `tt(~)'.
The tt(C[...]) argument is necessary so that this form of ~-completion is
not tried after the directory name is finished.
)
item(tt(-L))(
lists the existing completion behavior in a manner suitable for
putting into a start-up script; the existing behavior is not changed.
Any combination of the above forms, or the tt(-M) flag (which must
follow the tt(-L) flag), may be specified, otherwise all defined
completions are listed.  Any other flags supplied are ignored.
)
item(em(no argument))(
If no argument is given, tt(compctl) lists all defined completions
in an abbreviated form;  with a list of var(options), all completions
with those flags set (not counting extended completion) are listed.
)
endlist()

If the tt(PLUS()) flag is alone and followed immediately by the var(command)
list, the completion behavior for all the commands in the list is reset to
the default.  In other words, completion will subsequently use the
options specified by the tt(-D) flag.

The form with tt(-M) as the first and only option defines global
matching specifications, as described below in noderef(Matching Control).

texinode(Option Flags)(Alternative Completion)(Command Flags)(Programmable Completion)
sect(Option Flags)
startlist()
list([ tt(-fcFBdeaRGovNAIOPZEnbjrzu/) ])
list([ tt(-k) var(array) ] [ tt(-g) var(globstring) ] \
  [ tt(-s) var(subststring) ])
list([ tt(-K) var(function) ] [ tt(-i) var(function) ])
list([ tt(-Q) ] [ tt(-P) var(prefix) ] [ tt(-S) var(suffix) ])
list([ tt(-W) var(file-prefix) ] [ tt(-H) var(num pattern) ])
list([ tt(-q) ] [ tt(-X) var(explanation) ] [ tt(-Y) var(explanation) ])
list([ tt(-y) var(func-or-var) ] [ tt(-l) var(cmd) ] [ tt(-U) ])
list([ tt(-t) var(continue) ] [ tt(-J) var(name) ] [ tt(-V) var(name) ])
list([ tt(-M) var(match-spec) ])
endlist()

The remaining var(options) specify the type of command arguments
to look for during completion.  Any combination of these flags may be
specified; the result is a sorted list of all the possibilities.  The
options are as follows.
startmenu()
menu(Simple Flags)
menu(Flags with Arguments)
menu(Control Flags)
endmenu()

texinode(Simple Flags)(Flags with Arguments)()(Option Flags)
subsect(Simple Flags)
These produce completion lists made up by the shell itself:

startitem()
item(tt(-f))(
Filenames and filesystem paths.
)
item(tt(-/))(
Just filesystem paths.
)
item(tt(-c))(
Command names, including aliases, shell functions, builtins
and reserved words.
)
item(tt(-F))(
Function names.
)
item(tt(-B))(
Names of builtin commands.
)
item(tt(-m))(
Names of external commands.
)
item(tt(-w))(
Reserved words.
)
item(tt(-a))(
Alias names.
)
item(tt(-R))(
Names of regular (non-global) aliases.
)
item(tt(-G))(
Names of global aliases.
)
item(tt(-d))(
This can be combined with tt(-F), tt(-B), tt(-w),
tt(-a), tt(-R) and tt(-G) to get names of disabled
functions, builtins, reserved words or aliases.
)
item(tt(-e))(
This option (to show enabled commands) is in effect by default, but
may be combined with tt(-d); tt(-de) in combination with
tt(-F), tt(-B), tt(-w), tt(-a), tt(-R) and tt(-G)
will complete names of functions, builtins, reserved words or aliases
whether or not they are disabled.
)
item(tt(-o))(
Names of shell options (see
ifzman(zmanref(zshoptions))\
ifnzman(noderef(Options))\
).
)
item(tt(-v))(
Names of any variable defined in the shell.
)
item(tt(-N))(
Names of scalar (non-array) parameters.
)
item(tt(-A))(
Array names.
)
item(tt(-I))(
Names of integer variables.
)
item(tt(-O))(
Names of read-only variables.
)
item(tt(-p))(
Names of parameters used by the shell (including special parameters).
)
item(tt(-Z))(
Names of shell special parameters.
)
item(tt(-E))(
Names of environment variables.
)
item(tt(-n))(
Named directories.
)
item(tt(-b))(
Key binding names.
)
item(tt(-j))(
Job names:  the first word of the job leader's command line.  This is useful
with the tt(kill) builtin.
)
item(tt(-r))(
Names of running jobs.
)
item(tt(-z))(
Names of suspended jobs.
)
item(tt(-u))(
User names.
)
enditem()

texinode(Flags with Arguments)(Control Flags)(Simple Flags)(Option Flags)
subsect(Flags with Arguments)
These have user supplied arguments to determine how the list of
completions is to be made up:

startitem()
item(tt(-k) var(array))(
Names taken from the elements of tt($)var(array) (note that the `tt($)'
does not appear on the command line).
Alternatively, the argument var(array) itself may be a set
of space- or comma-separated values in parentheses, in which any
delimiter may be escaped with a backslash; in this case the argument
should be quoted.  For example,

indent(
nofill(tt(compctl -k "(cputime filesize datasize stacksize
	       coredumpsize resident descriptors)" limit))
)
)
item(tt(-g) var(globstring))(
The var(globstring) is expanded using filename globbing; it should be
quoted to protect it from immediate expansion. The resulting
filenames are taken as the possible completions.  Use `tt(*(/))' instead of
`tt(*/)' for directories.  The tt(fignore) special parameter is not
applied to the resulting files.  More than one pattern may be given
separated by blanks. (Note that brace expansion is em(not) part of
globbing.  Use the syntax `tt((either|or))' to match alternatives.)
)
item(tt(-s) var(subststring))(
The var(subststring) is split into words and these words are than
expanded using all shell expansion mechanisms (see
ifzman(zmanref(zshexpn))\
ifnzman(noderef(Expansion))\
).  The resulting words are taken as possible
completions.  The tt(fignore) special parameter is not applied to the
resulting files.  Note that tt(-g) is faster for filenames.
)
item(tt(-K) var(function))(
Call the given function to get the completions.  Unless the name
starts with an underscode, the function is
passed two arguments: the prefix and the suffix of the word on which
completion is to be attempted, in other words those characters before
the cursor position, and those from the cursor position onwards.  The
whole command line can be accessed with the tt(-c) and tt(-l) flags
of the tt(read) builtin. The
function should set the variable tt(reply) to an array containing
the completions (one completion per element); note that tt(reply)
should not be made local to the function.  From such a function the
command line can be accessed with the tt(-c) and tt(-l) flags to
the tt(read) builtin.  For example,

indent(
nofill(tt(function whoson { reply=(`users`); }
compctl -K whoson talk))
)

completes only logged-on users after `tt(talk)'.  Note that `tt(whoson)' must
return an array, so `tt(reply=`users`)' would be incorrect.
)
item(tt(-i) var(function))(
Like tt(-K), but the function is invoked in a context like that for
completion widgets, see 
ifzman(zmanref(zshzle))\
ifnzman(noderef(The zle Module))\
for more information.
)
item(tt(-H) var(num pattern))(
The possible completions are taken from the last var(num) history
lines.  Only words matching var(pattern) are taken.  If var(num) is
zero or negative the whole history is searched and if var(pattern) is
the empty string all words are taken (as with `tt(*)').  A typical
use is

indent(
tt(compctl -D -f PLUS() -H 0 '')
)

which forces completion to look back in the history list for a word if
no filename matches.
)
enditem()

texinode(Control Flags)()(Flags with Arguments)(Option Flags)
subsect(Control Flags)
These do not directly specify types of name to be completed, but
manipulate the options that do:

startitem()
item(tt(-Q))(
This instructs the shell not to quote any metacharacters in the possible
completions.  Normally the results of a completion are inserted into
the command line with any metacharacters quoted so that they are
interpreted as normal characters.  This is appropriate for filenames
and ordinary strings.  However, for special effects, such as inserting
a backquoted expression from a completion array (tt(-k)) so that
the expression will not be evaluated until the complete line is
executed, this option must be used.
)
item(tt(-P) var(prefix))(
The var(prefix) is inserted just before the completed string; any
initial part already typed will be completed and the whole var(prefix)
ignored for completion purposes.  For example,

indent(
tt(compctl -j -P "%" kill)
)

inserts a `%' after the kill command and then completes job names.
)
item(tt(-S) var(suffix))(
When a completion is found the var(suffix) is inserted after
the completed string.  In the case of menu completion the suffix is
inserted immediately, but it is still possible to cycle through the
list of completions by repeatedly hitting the same key.
)
item(tt(-W) var(file-prefix))(
With directory var(file-prefix):  for command, file, directory and
globbing completion (options tt(-c), tt(-f), tt(-/), tt(-g)), the file
prefix is implicitly added in front of the completion.  For example,

indent(
tt(compctl -/ -W ~/Mail maildirs)
)

completes any subdirectories to any depth beneath the directory
tt(~/Mail), although that prefix does not appear on the command line.
The var(file-prefix) may also be of the form accepted by the tt(-k)
flag, i.e. the name of an array or a literal list in parenthesis. In
this case all the directories in the list will be searched for
possible completions.
)
item(tt(-q))(
If used with a suffix as specified by the tt(-S) option, this
causes the suffix to be removed if the next character typed is a blank
or does not insert anything or if the suffix consists of only one character
and the next character typed is the same character; this the same rule used
for the tt(AUTO_REMOVE_SLASH) option.  The option is most useful for list
separators (comma, colon, etc.).
)
item(tt(-l) var(cmd))(
This option restricts the range
of command line words that are considered to be arguments.  If
combined with one of the extended completion patterns `tt(p[)...tt(])',
`tt(r[)...tt(])', or `tt(R[)...tt(])'  (see noderef(Extended Completion)
below) the range is restricted to the range of arguments
specified in the brackets.  Completion is then performed as if these
had been given as arguments to the var(cmd) supplied with the
option. If the var(cmd) string is empty the first word in the range
is instead taken as the command name, and command name completion
performed on the first word in the range.  For example,

indent(
tt(compctl -x 'r[-exec,;]' -l '' -- find)
)

completes arguments between `tt(-exec)' and the following `tt(;)' (or the end
of the command line if there is no such string) as if they were
a separate command line.
)
item(tt(-U))(
Use the whole list of possible completions, whether or not they
actually match the word on the command line.  The word typed so far
will be deleted.  This is most useful with a function (given by the
tt(-K) option) which can examine the word components passed to it
(or via the tt(read) builtin's tt(-c) and tt(-l) flags) and
use its own criteria to decide what matches.  If there is no
completion, the original word is retained.  Since the produced 
possible completions seldom seldom have interesting common prefixes
and suffixes, menucompletion is started immediatly if tt(AUTO_MENU) is
set and this flag is used.
)
item(tt(-y) var(func-or-var))(
The list provided by var(func-or-var) is displayed instead of the list
of completions whenever a listing is required; the actual completions
to be inserted are not affected.  It can be provided in two
ways. Firstly, if var(func-or-var) begins with a tt($) it defines a
variable, or if it begins with a left parenthesis a literal
array, which contains the list.  A variable may have been set by a
call to a function using the tt(-K) option.  Otherwise it contains the
name of a function which will be executed to create the list.  The
function will be passed as an argument list all matching completions,
including prefixes and suffixes expanded in full, and should set the
array var(reply) to the result.  In both cases, the display list will
only be retrieved after a complete list of matches has been created.

Note that the returned list does not have to correspond, even in
length, to the original set of matches, and may be passed as a scalar
instead of an array.  No special formatting of characters is
performed on the output in this case; in particular, newlines are
printed literally and if they appear output in columns is suppressed.
)
item(tt(-X) var(explanation))(
Print var(explanation) when trying completion on the current set of
options. A `tt(%n)' in this string is replaced by the number of
matches that were added for this explanation string.
The explanation only appears if completion was tried and there was
no unique match, or when listing completions. Explanation strings 
will be listed together with the matches of the group specified
together with the tt(-X) option (using the tt(-J) or tt(-V)
option). If the same explanation string is given to multiple tt(-X)
options, the string appears only once (for each group) and the number
of matches shown for the `tt(%n)' is the total number of all matches
for each of these uses. In any case, the explanation string will only
be shown if there was at least one match added for the explanation
string.

The sequences tt(%B), tt(%b), tt(%S), tt(%s), tt(%U), and tt(%u) specify
output attributes (bold, standout, and underline) and tt(%{...%}) can
be used to include literal escape sequences as in prompts.
)
item(tt(-Y) var(explanation))(
Identical to tt(-X), except that the var(explanation) first undergoes
expansion following the usual rules for strings in double quotes.
The expansion will be carried out after any functions are called for
the tt(-K) or tt(-y) options, allowing them to set variables.
)
item(tt(-t) var(continue))(
The var(continue)-string contains a character that specifies which set
of completion flags should be used next.  It is useful:

(i) With tt(-T), or when trying a list of pattern completions, when
tt(compctl) would usually continue with ordinary processing after
finding matches; this can be suppressed with `tt(-tn)'.

(ii) With a list of alternatives separated by tt(+), when tt(compctl)
would normally stop when one of the alternatives generates matches.  It
can be forced to consider the next set of completions by adding `tt(-t+)'
to the flags of the alternative before the `tt(+)'.

(iii) In an extended completion list (see below), when tt(compctl) would
normally continue until a set of conditions succeeded, then use only
the immediately following flags.  With `tt(-t-)', tt(compctl) will
continue trying extended completions after the next `tt(-)'; with
`tt(-tx)' it will attempt completion with the default flags, in other
words those before the `tt(-x)'.
)
item(tt(-J) var(name))(
This gives the name of the group the matches should be placed in. Groups
are listed and sorted separately; likewise, menucompletion will offer
the matches in the groups in the order in which the groups were
defined. If no group name is explicitly given, the matches are stored in
a group named var(default). The first time a group name is encountered,
a group with that name is created. After that all matches with the same
group name are stored in that group.

This can be useful with non-exclusive alternative completions.  For
example, in

indent(
tt(compctl -f -J files -t+ + -v -J variables foo)
)

both files and variables are possible completions, as the tt(-t+) forces
both sets of alternatives before and after the tt(+) to be considered at
once.  Because of the tt(-J) options, however, all files are listed
before all variables.
)
item(tt(-V) var(name))(
Like tt(-J), but matches within the group will not be sorted in listings
nor in menucompletion. These unsorted groups are in a different name
space than the sorted ones, so groups defined as tt(-J files) and tt(-V
files) are distinct.
)
item(tt(-M) var(match-spec))(
This defines additional matching control specifications that should be used
only when testing words for the list of flags this flag appears in. The format
of the var(match-spec) string is described below in noderef(Matching Control).
)
enditem()

texinode(Alternative Completion)(Extended Completion)(Option Flags)(Programmable Completion)
sect(Alternative Completion)
startlist()
list(tt(compctl) [ tt(-CDT) ] var(options) tt(PLUS()) var(options) [ tt(PLUS()) ... ] \
[ tt(PLUS()) ] var(command) ...)
endlist()

The form with `tt(PLUS())' specifies alternative options. Completion is
tried with the options before the first `tt(PLUS())'. If this produces no
matches completion is tried with the flags after the `tt(PLUS())' and so on. If
there are no flags after the last `tt(PLUS())' and a match has not been found
up to that point, default completion is tried.
If the list of flags contains a tt(-t) with a tt(PLUS()) character, the next
list of flags is used even if the current list produced matches.

texinode(Extended Completion)(Matching Control)(Alternative Completion)(Programmable Completion)
sect(Extended Completion)
startlist()
list(nofill(tt(compctl) [ tt(-CDT) ] var(options) \
tt(-x) var(pattern) var(options) tt(-) ... tt(--)
  [ var(command) ... ]))
list(nofill(tt(compctl) [ tt(-CDT) ] var(options) \
[ tt(-x) var(pattern) var(options) tt(-) ... tt(--) ]
  [ tt(PLUS()) var(options) [ tt(-x) ... tt(--) ] ... [tt(PLUS())] ] \
[ var(command) ... ]))
endlist()

The form with `tt(-x)' specifies extended completion for the
commands given; as shown, it may be combined with alternative
completion using `tt(PLUS())'.  Each var(pattern) is examined in turn; when a
match is found, the corresponding var(options), as described in
noderef(Option Flags) above, are used to generate possible
completions.  If no var(pattern) matches, the var(options) given
before the tt(-x) are used.

Note that each pattern should be supplied as a single argument and
should be quoted to prevent expansion of metacharacters by the
shell.

A var(pattern) is built of sub-patterns separated by commas; it
matches if at least one of these sub-patterns matches (they are
`or'ed). These sub-patterns are in turn composed of other
sub-patterns separated by white spaces which match if all of the
sub-patterns match (they are `and'ed).  An element of the
sub-patterns is of the form `var(c)tt([)...tt(][)...tt(])', where the pairs of
brackets may be repeated as often as necessary, and matches if any of
the sets of brackets match (an `or').  The example below makes this
clearer.

The elements may be any of the following:

startitem()
item(tt(s[)var(string)tt(])...)(
Matches if the current word on the command line starts with
one of the strings given in brackets.  The var(string) is not removed
and is not part of the completion.
)
item(tt(S[)var(string)tt(])...)(
Like tt(s[)var(string)tt(]) except that the var(string) is part of the
completion.
)
item(tt(p[)var(from)tt(,)var(to)tt(])...)(
Matches if the number of the current word is between one of
the var(from) and var(to) pairs inclusive. The comma and var(to)
are optional; var(to) defaults to the same value as var(from).  The
numbers may be negative: tt(-)var(n) refers to the var(n)'th last word
on the line.
)
item(tt(c[)var(offset)tt(,)var(string)tt(])...)(
Matches if the var(string) matches the word offset by
var(offset) from the current word position.  Usually var(offset)
will be negative.
)
item(tt(C[)var(offset)tt(,)var(pattern)tt(])...)(
Like tt(c) but using pattern matching instead.
)
item(tt(w[)var(index)tt(,)var(string)tt(])...)(
Matches if the word in position var(index) is equal
to the corresponding var(string).  Note that the word count is made
after any alias expansion.
)
item(tt(W[)var(index)tt(,)var(pattern)tt(])...)(
Like tt(w) but using pattern matching instead.
)
item(tt(n[)var(index)tt(,)var(string)tt(])...)(
Matches if the current word contains var(string).  Anything up to and
including the var(index)th occurrence of this string will not be
considered part of the completion, but the rest will.  var(index) may
be negative to count from the end: in most cases, var(index) will be
1 or -1.  For example,

indent(
tt(compctl -s '`users`' -x 'n[1,@]' -k hosts -- talk)
)

will usually complete usernames, but if you insert an tt(@) after the
name, names from the array var(hosts) (assumed to contain hostnames,
though you must make the array yourself) will be completed.  Other
commands such as tt(rcp) can be handled similarly.
)
item(tt(N[)var(index)tt(,)var(string)tt(])...)(
Like tt(n) except that the string will be
taken as a character class.  Anything up to and including the
var(index)th occurrence of any of the characters in var(string)
will not be considered part of the completion.
)
item(tt(m[)var(min)tt(,)var(max)tt(])...)(
Matches if the total number of words lies between var(min) and
var(max) inclusive.
)
item(tt(r[)var(str1)tt(,)var(str2)tt(])...)(
Matches if the cursor is after a word with prefix var(str1).  If there
is also a word with prefix var(str2) on the command line it matches
only if the cursor is before this word. If the comma and var(str2) are
omitted, it matches if the cursor is after a word with prefix var(str1).
)
item(tt(R[)var(str1)tt(,)var(str2)tt(])...)(
Like tt(r) but using pattern matching instead.
)
enditem()

texinode(Matching Control)(Example)(Extended Completion)(Programmable Completion)
sect(Matching Control)

It is possible by use of the tt(-M) var(spec) flag to specify how the
characters in the string to be completed (referred to here as the
command line) map onto the characters in the list of matches produced by
the completion code (referred to here as the trial completions).

The var(spec) consists of one or more matching descriptions separated by
whitespace. Each description consists of a letter followed by a colon,
then the patterns describing which character sequences on the line match
which character sequences in the trial completion.  Any sequence of characters not
handled in this fashion must match exactly, as usual.

The forms of var(spec) understood are as follows. In each case, the
form with an uppercase initial character retains the string already
typed on the command line as the final result of completion, while with
a lowercase initial character the string on the command line is changed
into the corresponding part of the trial completion.

startitem()
xitem(tt(m:)var(lpat)tt(=)var(tpat))
item(tt(M:)var(lpat)tt(=)var(tpat))(
Here, var(lpat) is a pattern that matches on the command line,
corresponding to var(tpat) which matches in the trial completion.
)
xitem(tt(l:)var(anchor)tt(|)var(lpat)tt(=)var(tpat))
item(tt(L:)var(anchor)tt(|)var(lpat)tt(=)var(tpat))(
These letters are for patterns that are anchored by another pattern on
the left side. Matching for var(lpat) and var(tpat) is as for tt(m) and
tt(M), but the pattern var(lpat) matched on the command line must be
preceeded by the pattern var(anchor).  The var(anchor) can be blank to
anchor the match to the start of the command line string; otherwise the
anchor can occur anywhere, but must match in both the command line and
trial completion strings.
)
xitem(tt(r:)var(lpat)tt(|)var(anchor)tt(=)var(tpat))
item(tt(R:)var(lpat)tt(|)var(anchor)tt(=)var(tpat))(
As tt(l) and tt(L) with the difference that the command line and trial
completion patterns are anchored on the right side.  Here an empty
var(anchor) forces the match to the end of the command line string.
)
enditem()

Each var(lpat), var(tpat) or var(anchor) is either an empty string or
consists of a sequence of literal characters (which may be quoted with a
backslash), question marks, character classes, and correspondence
classes; ordinary shell patterns are not used.  Literal characters match
only themselves, question marks match any character, and character
classes are formed as for globbing and match any character in the given
set.

Correspondence classes are defined like character classes, but with two
differences: they are delimited by a pair of braces, and negated classes
are not allowed, so the characters tt(!) and tt(^) have no special
meaning directly after the opening brace.  They indicate that a range of
characters on the line match a range of characters in the trial
completion, but (unlike ordinary character classes) paired according to
the corresponding position in the sequence. For example, to make any
lowercase letter on the line match the corresponding uppercase letter in
the trial completion, you can use `tt(m:{a-z}={A-Z})'.  More than one
pair of classes can occur, in which case the first class before the
tt(=) corresponds to the first after it, and so on.  If one side has
more such classes than the other side, the superfluous classes behave
like normal character classes. In anchor patterns correspondence classes
also behave like normal character classes.

The pattern var(tpat) may also be a single star, `tt(*)'. This means
that the pattern on the command line can match any number of characters
in the trial completion. In this case the pattern must be anchored (on
either side); the var(anchor) then determines how much of the trial
completion is to be included --- only the characters up to the next
appearance of the anchor will be matched.

Examples:

The option tt(-o) produces option names in all-lowercase form, without
underscores, and without the optional tt(no) at the beginning even
though the builtins tt(setopt) and tt(unsetopt) understand option names
with uppercase letters, underscores, and the optional tt(no).  The
following alters the matching rules so that the prefix tt(no) and any
underscore are ignored when trying to match the trial completions
generated and uppercase letters on the line match the corresponding
lowercase letters in the words:
indent(
nofill(tt(compctl -M 'L:|[nN][oO]= M:_= M:{A-Z}={a-z}' \ 
  -o setopt unsetopt))
)
The first part says that the pattern `tt([nN][oO])' at the beginning
(the empty anchor before the pipe symbol) of the string on the
line matches the empty string in the list of words generated by
completion, so it will be ignored if present. The second part does the
same for an underscore anywhere in the command line string, and the
third part uses correspondence classes so that any
uppercase letter on the line matches the corresponding lowercase
letter in the word. The use of the uppercase forms of the
specification characters (tt(L) and tt(M)) guarantees that what has
already been typed on the command line (in particular the prefix
tt(no)) will not be deleted.

The second example makes completion case insensitive.  By using
tt(compctl) with the tt(-M) option alone this applies to every
completion.  This is just the same as in the tt(setopt) example, except
here we wish to retain the characters in the list of completions:
indent(
tt(compctl -M 'm:{a-z}={A-Z}')
)
This makes lowercase letters match their uppercase counterparts.
To make uppercase letters match the lowercase forms as well:
indent(
tt(compctl -M 'm:{a-zA-Z}={A-Za-z}')
)

A nice example for the use of tt(*) patterns is partial word
completion. Sometimes you would like to make strings like tt(c.s.u)
complete to strings like tt(comp.source.unix), i.e. the word on the
command line consists of multiple parts, separated by a dot in this
example, where each part should be completed separately --- note,
however, that the case where each part of the word, i.e. tt(comp),
tt(source) and tt(unix) in this example, is to be completed separately
is a different problem to be solved by extended completion.  The
example can be handled by:
indent(
nofill(tt(compctl -M 'r:|.=* r:|=*' \ 
  -k '(comp.sources.unix comp.sources.misc ...)' ngroups))
)
The first specification says that tt(lpat) is the empty string, while
tt(anchor) is a dot; tt(tpat) is tt(*), so this can match anything
except for the `tt(.)' from the anchor in
the trial completion word.  So in tt(c.s.u), the matcher sees tt(c),
followed by the empty string, followed by the anchor `tt(.)', and
likewise for the second dot, and replaces the empty strings before the
anchors, giving tt(c)[tt(omp)]tt(.s)[tt(ources)]tt(.u)[tt(nix)], where
the last part of the completion is just as normal.

The second specification is needed to make this work when the cursor is
in the middle of the string on the command line and the option
tt(COMPLETE_IN_WORD) is set. In this case the completion code would
normally try to match trial completions that end with the string as
typed so far, i.e. it will only insert new characters at the cursor
position rather then at the end.  However in our example we would like
the code to recognise matches which contain extra characters after the
string on the line (the tt(nix) in the example).  Hence we say that the
empty string at the end of the string on the line matches any characters
at the end of the trial completion.

More generally, the specification
indent(
tt(compctl -M 'r:|[.,_-]=* r:|=*')
)
allows one to complete words with abbreviations before any of the
characters in the square brackets in any completion.  For example, to
complete tt(veryverylongfile.c) rather than tt(veryverylongheader.h)
with the above in effect, you can just type tt(very.c) before attempting
completion.

The form tt(compctl -M) that defines global matching actually accepts
any number of specification strings, unlike the case where the tt(-M)
option applies only to a particular command.  In this case, when
completion is attempted for any command, the code will try the
specifications in order until one matches.  This allows one to define
simple and fast matches to be used first, more powerful matchers as a
second choice, and so on.

For example, one can make the code match trial completions that contain
the string on the command line as a substring, not just at the
beginning.  Since this might produce more matches than we want,
we arrange for it to be tried only if the matchers described above don't
produce any matches:
indent(
tt(compctl -M 'r:|[.,_-]=* r:|=*' 'l:|=* r:|=*')
)
Here, if the string on the command line is tt(foo.bar), tt(compctl)
first tries matching tt(foo)var(anything)tt(.bar)var(anything), as
with the previous example.  If that fails, the two descriptions in the
second string after the tt(-M) say that the blanks at the beginning
and end of the string on the command line can match any set of
characters at the beginning or end of the trial completion, so it will
look for var(anything)tt(foo.bar)var(anything).

texinode(Example)()(Matching Control)(Programmable Completion)
sect(Example)
indent(
nofill(
tt(compctl -u -x 's[tt(PLUS())] c[-1,-f],s[-f+PLUS()]' \ 
  -g '~/Mail/*(:t)' - 's[-f],c[-1,-f]' -f -- mail))
)
This is to be interpreted as follows:

If the current command is tt(mail), then

indent(
if ((the current word begins with tt(PLUS()) and the previous word is tt(-f))
or (the current word begins with tt(-f+PLUS()))), then complete the
non-directory part (the `tt(:t)' glob modifier) of files in the directory
tt(~/Mail); else

if the current word begins with tt(-f) or the previous word was tt(-f), then
complete any file; else

complete user names.
)