dotfiles/zsh/modules/completion/external/zsh-completions-howto.org

441 lines
26 KiB
Org Mode
Raw Normal View History

2018-04-05 13:06:54 +02:00
* Intro
The official documentation for writing zsh completion functions is difficult to understand, and doesn't give many examples.
At the time of writing this document I was able to find a few other tutorials on the web, however those tutorials only
explain a small portion of the capabilities of the completion system. This document aims to cover areas not explained elsewhere,
with examples, so that you can learn how to write more advanced completion functions. I do not go into all the details, but will
give enough information and examples to get you up and running. If you need more details you can look it up for yourself in the
[[http://zsh.sourceforge.net/Doc/Release/Completion-System.html#Completion-System][official documentation]].
Please make any scripts that you create publically available for others (e.g. by forking this repo and making a [[id:64bcd501-b0f0-48c7-b8e2-07af708b95ec][pull request]]).
Also if you have any more information to add or improvements to make to this tutorial, please do.
* Getting started
** Telling zsh which function to use for completing a command
Completion functions for commands are stored in files with names beginning with an underscore _, and these files should
be placed in a directory listed in the $fpath variable.
You can add a directory to $fpath by adding a line like this to your ~/.zshrc file:
#+BEGIN_SRC sh
fpath=(~/newdir $fpath)
#+END_SRC
The first line of a completion function file can look something like this:
#+BEGIN_SRC sh
#compdef foobar
#+END_SRC
This tells zsh that the file contains code for completing the foobar command.
This is the format that you will use most often for the first line, but you can also use the same file for completing
several different functions if you want. See [[http://zsh.sourceforge.net/Doc/Release/Completion-System.html#Autoloaded-files][here]] for more details.
You can also use the compdef command directly (e.g. in your ~/.zshrc file) to tell zsh which function to use for completing
a command like this:
#+BEGIN_SRC sh
> compdef _function foobar
#+END_SRC
or to use the same completions for several commands:
#+BEGIN_SRC sh
> compdef _function foobar goocar hoodar
#+END_SRC
or if you want to supply arguments:
#+BEGIN_SRC sh
> compdef '_function arg1 arg2' foobar
#+END_SRC
See [[http://zsh.sourceforge.net/Doc/Release/Completion-System.html#Functions-4][here]] for more details.
** Completing generic gnu commands
Many [[http://www.gnu.org/][gnu]] commands have a standardized way of listing option descriptions (when the --help option is used).
For these commands you can use the _gnu_generic function for automatically creating completions, like this:
#+BEGIN_SRC sh
> compdef _gnu_generic foobar
#+END_SRC
or to use _gnu_generic with several different commands:
#+BEGIN_SRC sh
> compdef _gnu_generic foobar goocar hoodar
#+END_SRC
This line can be placed in your ~/.zshrc file.
** Copying completions from another command
If you want a command, say cmd1, to have the same completions as another, say cmd2, which has already had
completions defined for it, you can do this:
#+BEGIN_SRC sh
> compdef cmd1=cmd2
#+END_SRC
This can be useful for example if you have created an alias for a command to help you remember it.
* Writing your own completion functions
A good way to get started is to look at some already defined completion functions.
On my linux installation these are found in /usr/share/zsh/functions/Completion/Unix
and /usr/share/zsh/functions/Completion/Linux and a few other subdirs.
You will notice that the _arguments function is used a lot in these files.
This is a utility function that makes it easy to write simple completion functions.
The _arguments function is a wrapper around the compadd builtin function.
The compadd builtin is the core function used to add completion words to the command line, and control its behaviour.
However, most of the time you will not need to use compadd, since there are many utility functions such as _arguments
and _describe which are easier to use.
For very basic completions the _describe function should be adequate
** Utility functions
Here is a list of some of the utility functions that may be of use.
The full list of utility functions, with full explanations, is available [[http://zsh.sourceforge.net/Doc/Release/Completion-System.html#Completion-Functions][here]].
Examples of how to use these functions are given in the next section.
*** main utility functions for overall completion
| _alternative | Can be used to generate completion candidates from other utility functions or shell code. |
| _arguments | Used to specify how to complete individual options & arguments for a command with unix style options. |
| _describe | Used for creating simple completions consisting of single words with descriptions (but no actions). Easier to use than _arguments |
| _gnu_generic | Can be used to complete options for commands that understand the `--help' option. |
| _regex_arguments | Creates a function for matching commandline arguments with regular expressions, and then performing actions/completions. |
*** functions for performing complex completions of single words
| _values | Used for completing arbitrary keywords (values) and their arguments, or comma separated lists of such combinations. |
| _combination | Used to complete combinations of values, for example pairs of hostnames and usernames. |
| _multi_parts | Used for completing multiple parts of words separately where each part is separated by some char, e.g. for completing partial filepaths: /u/i/sy -> /usr/include/sys |
| _sep_parts | Like _multi_parts but allows different separators at different parts of the completion. |
*** functions for completing specific types of objects
| _path_files | Used to complete filepaths. Take several options to control behaviour. |
| _files | Calls _path_files with all options except -g and -/. These options depend on file-patterns style setting. |
| _net_interfaces | Used for completing network interface names |
| _users | Used for completing user names |
| _groups | Used for completing group names |
| _options | Used for completing the names of shell options. |
| _parameters | Used for completing the names of shell parameters/variables (can restrict to those matching a pattern). |
*** functions for handling cached completions
If you have a very large number of completions you can save them in a cache file so that the completions load quickly.
| _cache_invalid | indicates whether the completions cache corresponding to a given cache identifier needs rebuilding |
| _retrieve_cache | retrieves completion information from a cache file |
| _store_cache | store completions corresponding to a given cache identifier in a cache file |
*** other functions
| _message | Used for displaying help messages in places where no completions can be generated. |
| _regex_words | Can be used to generate arguments for the _regex_arguments command. This is easier than writing the arguments manually. |
| _guard | Can be used in the ACTION of specifications for _arguments and similar functions to check the word being completed. |
*** Actions
Many of the utility functions such as _arguments, _regex_arguments, _alternative and _values may include an action
at the end of an option/argument specification. This action indicates how to complete the corresponding argument.
The actions can take one of the following forms:
| ( ) | Argument is required but no matches are generated for it. |
| (ITEM1 ITEM2) | List of possible matches |
| ((ITEM1\:'DESC1' ITEM2\:'DESC2')) | List of possible matches, with descriptions. Make sure to use different quotes than those around the whole specification. |
| ->STRING | Set $state to STRING and continue ($state can be checked in a case statement after the utility function call) |
| FUNCTION | Name of a function to call for generating matches or performing some other action, e.g. _files or _message |
| {EVAL-STRING} | Evaluate string as shell code to generate matches. This can be used to call a utility function with arguments, e.g. _values or _describe |
| =ACTION | Inserts a dummy word into completion command line without changing the point at which completion takes place. |
Not all action types are available for all utility functions that use them. For example the ->STRING type is not available in the
_regex_arguments or _alternative functions.
** Writing simple completion functions using _describe
The _describe function can be used for simple completions where the order and position of the options/arguments is
not important. You just need to create an array parameter to hold the options & their descriptions, and then pass
the parameter name as an argument to _describe. The following example creates completion candidates -c and -d, with
the descriptions (note this should be put in a file called _cmd in some directory listed in $fpath).
#+BEGIN_SRC sh
#compdef cmd
local -a options
options=('-c:description for -c opt' '-d:description for -d opt')
_describe 'values' options
#+END_SRC
You can use several different lists separated by a double hyphen e.g. like this:
#+BEGIN_SRC sh
local -a options arguments
options=('-c:description for -c opt' '-d:description for -d opt')
arguments=('e:description for e arg' 'f:description for f arg')
_describe 'values' options -- arguments
#+END_SRC
The _describe function can be used in an ACTION as part of a specification for _alternative, _arguments or _regex_arguments.
In this case you will have to put it in braces with its arguments, e.g. 'TAG:DESCRIPTION:{_describe 'values' options}'
** Writing completion functions using _alternative
Like _describe, this function performs simple completions where the order and position of options/arguments is not important.
However, unlike _describe, you can call execute shell code or call functions to obtain the completion candidates.
As arguments it takes a list of specifications each in the form 'TAG:DESCRIPTION:ACTION' where TAG is a tag name,
DESCRIPTION is a description, and ACTION is one of the action types listed previously (apart from the ->STRING and =ACTION forms).
For example:
#+BEGIN_SRC sh
_alternative 'args:custom args:(a b c)' 'files:filenames:_files'
#+END_SRC
The first specification adds completion candidates a, b & c, and the second specification calls the _files function
for completing filepaths.
We could split the specifications over several lines with \ and add descriptions to each of the custom args like this:
#+BEGIN_SRC sh
_alternative 'args:custom args:((a\:"description a" b\:"description b" c\:"description c"))'\
'files:filenames:_files'
#+END_SRC
If we want to call _files with arguments we can put it in braces, like this:
#+BEGIN_SRC sh
_alternative 'args:custom args:((a\:"description a" b\:"description b" c\:"description c"))'\
'files:filenames:{_files -/}'
#+END_SRC
To use parameter expansion to create our list of completions we must use double quotes to quote the specifications,
e.g:
#+BEGIN_SRC sh
_alternative "dirs:user directories:($userdirs)"\
"pids:process IDs:($(ps -A o pid=))"
#+END_SRC
In this case the first specification adds the words stored in the $userdirs variable, and the second specification
evaluates 'ps -A o pid=' to get a list of pids to use as completion candidates.
We can use other utility functions such as _values in the ACTION to perform more complex completions, e.g:
#+BEGIN_SRC sh
_alternative "dirs:user directories:($userdirs)"\
'opts:comma separated opts:{_values -s , a b c}'
#+END_SRC
this will complete the items in $userdirs, aswell as a comma separated list containing a, b &/or c.
As with _describe, the _alternative function can itself be used in an ACTION as part of a specification for _arguments
or _regex_arguments.
** Writing completion functions using _arguments
With the _arguments function you can create more sophisticated completion functions.
Like the _alternative function, _arguments takes a list of specification strings as arguments.
These specification strings can be for specifying options and any corresponding option arguments (e.g. -f filename),
or command arguments.
Basic option specifications take the form '-OPT[DESCRIPTION]', e.g. like this:
#+BEGIN_SRC sh
_arguments '-s[sort output]' '--l[long output]' '-l[long output]'
#+END_SRC
Arguments for the option can be specified after the option description in this form '-OPT[DESCRIPTION]:MESSAGE:ACTION',
where MESSAGE is a message to display and ACTION can be any of the forms mentioned in the ACTIONS section above.
For example:
#+BEGIN_SRC sh
_arguments '-f[input file]:filename:_files'
#+END_SRC
Command argument specifications take the form 'N:MESSAGE:ACTION' where N indicates that it is the Nth command argument,
and MESSAGE & ACTION are as before. If the N is omitted then it just means the next command argument (after any that have
already been specified). If a double colon is used at the start (after N) then the argument is optional.
For example:
#+BEGIN_SRC sh
_arguments '-s[sort output]' '1:first arg:_net_interfaces' '::optional arg:_files' ':next arg:(a b c)'
#+END_SRC
here the first arg is a network interface, the next optional arg is a file name, the last arg can be either a, b or c,
and the -s option may be completed at any position.
The _arguments function allows the full set of ACTION forms listed in the ACTION section above.
This means that you can use actions for selecting case statement branches like this:
#+BEGIN_SRC sh
_arguments '-m[music file]:filename:->files' '-f[flags]:flag:->flags'
case "$state" in
files)
local -a music_files
music_files=( Music/**/*.{mp3,wav,flac,ogg} )
_multi_parts / music_files
;;
flags)
_values -s , 'flags' a b c d e
;;
esac
#+END_SRC
In this case paths to music files are completed stepwise descending down directories using the _multi_parts function,
and the flags are completed as a comma separated list using the _values function.
I have just given you the basics of _arguments specifications here, you can also specify mutually exclusive options,
repeated options & arguments, options beginning with + insead of -, etc. For more details see the [[http://zsh.sourceforge.net/Doc/Release/Completion-System.html#Completion-System][official documentation]].
Also have a look at the tutorials mentioned at the end of this document, and the completion functions in the [[https://github.com/vapniks/zsh-completions/tree/master/src][src directory]].
** Writing completion functions using _regex_arguments and _regex_words
If you have a complex command line specification with several different possible argument sequences then
the _regex_arguments function may be what you need.
_regex_arguments creates a completion function whose name is given by the first argument.
Hence you need to first call _regex_arguments to create the completion function, and then call that function,
e.g. like this:
#+BEGIN_SRC sh
_regex_arguments _cmd OTHER_ARGS..
_cmd "$@"
#+END_SRC
The OTHER_ARGS should be sequences of specifications for matching & completing words on the command line.
These sequences can be separated by '|' to represent alternative sequences of words.
You can use bracketing to arbitrary depth to specify alternate subsequences, but the brackets must be backslashed like this \( \)
or quoted like this '(' ')'.
For example:
#+BEGIN_SRC sh
_regex_arguments _cmd SEQ1 '|' SEQ2 \( SEQ2a '|' SEQ2b \)
_cmd "$@"
#+END_SRC
this specifies a command line matching either SEQ1, or SEQ2 followed by SEQ2a or SEQ2b.
Each specification in a sequence must contain a / PATTERN/ part at the start followed by an optional ':TAG:DESCRIPTION:ACTION'
part.
Each PATTERN is a regular expression to match a word on the command line. These patterns are processed sequentially
until we reach a pattern that doesn't match at which point any corresponding ACTION is performed to obtain completions
for that word. Note that there needs to be a pattern to match the initial command itself.
See below for further explanation about PATTERNs.
The ':TAG:DESCRIPTION:ACTION' part is interpreted in the same way as for the _alternative function specifications,
except that it has an extra : at the start, and now all of the possible ACTION formats listed previously are allowed.
Here is an example:
#+BEGIN_SRC sh
_regex_arguments _hello /$'[^\0]##\0'/ \( /$'word1(a|b|c)\0'/ ':word:first word:(word1a word1b word1c)' '|'\
/$'word11(a|b|c)\0'/ ':word:first word:(word11a word11b word11c)' \( /$'word2(a|b|c)\0'/ ':word:second word:(word2a word2b word2c)'\
'|' /$'word22(a|b|c)\0'/ ':word:second word:(word22a word22b word22c)' \) \)
_cmd "$@"
#+END_SRC
in this case the first word can be word1 or word11 followed by an a, b or c, and if the first word contains 11 then a second
word is allowed which can be word2 followed by and a, b, or c, or a filename.
If this sounds too complicated a much simpler alternative is to use the _regex_words function for creating
specifications for _regex_arguments.
*** Patterns
You may notice that the / PATTERN/ specs in the previous example don't look like normal regular expressions.
Often a string parameter in the form $'foo\0' is used. This is so that the \0 in the string is interpreted correctly
as a null char which is used to separate words in the internal representation. If you don't include the \0 at the end
of the pattern you may get problems matching the next word. If you need to use the contents of a variable in a pattern,
you can double quote it so that it gets expanded and then put a string parameter containing a null char afterwards,
like this: "$somevar"$'\0'
The regular expression syntax for patterns seems to be a bit different from normal regular expressions,
and I can't find documentation anywhere.
However I have managed to work out what the following special chars are for:
| * | wildcard - any number of chars |
| ? | wildcard - single char |
| # | zero or more of the previous char (like * in a normal regular expression) |
| ## | one or more of the previous char (like + in a normal regular expression) |
*** _regex_words
The _regex_words function makes it much easier to create specifications for _regex_arguments.
The results of calling _regex_words can be stored in a variable which can then be used instead
of a specification for _regex_arguments.
To create a specification using _regex_words you supply it with a tag followed by a description followed by a list
of specifications for individual words. These specifications take the form 'WORD:DESCRIPTION:SPEC' where WORD is the
word to be completed, DESCRIPTION is a description for it, and SPEC can be another variable created by _regex_words
specifying words that come after the current word or blank if there are no further words.
For example:
#+BEGIN_SRC sh
_regex_words firstword 'The first word' 'word1a:a word:' 'word1b:b word:' 'word1c:c word'
#+END_SRC
the results of this function call will be stored in the $reply array, and so we should store it in another array
before $reply gets changed again, like this:
#+BEGIN_SRC sh
local -a firstword
_regex_words word 'The first word' 'word1a:a word:' 'word1b:b word:' 'word1c:c word'
firstword="$reply[@]"
#+END_SRC
we could then use it with _regex_arguments like this:
#+BEGIN_SRC sh
_regex_arguments _cmd /$'[^\0]##\0'/ "$firstword[@]"
_cmd "$@"
#+END_SRC
Note that I have added an extra pattern for the initial command word itself.
Here is a more complex example where we call _regex_words for different words on the command line
#+BEGIN_SRC sh
local -a firstword firstword2 secondword secondword2
_regex_words word1 'The second word' 'woo:tang clan' 'hoo:not me'
secondword=("$reply[@]")
_regex_words word2 'Another second word' 'yee:thou' 'haa:very funny!'
secondword2=("$reply[@]")
_regex_words commands 'The first word' 'foo:do foo' 'man:yeah man' 'chu:at chu'
firstword=("$reply[@]")
_regex_words word4 'Another first word' 'boo:scare somebody:$secondword' 'ga:baby noise:$secondword'\
'loo:go to the toilet:$secondword2'
firstword2=("$reply[@]")
_regex_arguments _hello /$'[^\0]##\0'/ "${firstword[@]}" "${firstword2[@]}"
_hello "$@"
#+END_SRC
In this case the first word can be one of "foo", "man", "chu", "boo", "ga" or "loo".
If the first word is "boo" or "ga" then the second word can be "woo" or "hoo",
and if the first word is "loo" then the second word can be "yee" or "haa", in the other
cases there is no second word.
For a good example of the usage of _regex_words have a look at the _ip function.
** complex completions with _values, _sep_parts, & _multi_parts
The _values, _sep_parts & _multi_parts functions can be used either on their own, or as ACTIONs in specifications for
_alternative, _arguments or _regex_arguments. The following examples may be instructive.
See the [[http://zsh.sourceforge.net/Doc/Release/Completion-System.html#Completion-System][official documentation]] for more info.
Space separated list of mp3 files:
#+BEGIN_SRC sh
_values 'mp3 files' ~/*.mp3
#+END_SRC
Comma separated list of session id numbers:
#+BEGIN_SRC sh
_values -s , 'session id' "${(uonzf)$(ps -A o sid=)}"
#+END_SRC
Completes foo@news:woo, or foo@news:laa, or bar@news:woo, etc:
#+BEGIN_SRC sh
_sep_parts '(foo bar)' @ '(news ftp)' : '(woo laa)'
#+END_SRC
Complete some MAC addresses one octet at a time:
#+BEGIN_SRC sh
_multi_parts : '(00:11:22:33:44:55 00:23:34:45:56:67 00:23:45:56:67:78)'
#+END_SRC
** Adding completion words directly using compadd
For more fine grained control you can use the builtin compadd function to add completion words directly.
This function has many different options for controlling how completions are displayed and how text on the command line
can be altered when words are completed. Read the [[http://zsh.sourceforge.net/Doc/Release/Completion-System.html#Completion-System][official documentation]] for full details.
Here I just give a few simple examples.
Add some words to the list of possible completions:
#+BEGIN_SRC sh
compadd foo bar blah
#+END_SRC
As above but also display an explanation:
#+BEGIN_SRC sh
compadd -X 'Some completions' foo bar blah
#+END_SRC
As above but automatically insert a prefix of "what_" before the completed word:
#+BEGIN_SRC sh
compadd -P what_ foo bar blah
#+END_SRC
As above but automatically insert a suffix of "_todo" after the completed word:
#+BEGIN_SRC sh
compadd -S _todo foo bar blah
#+END_SRC
As above but automatically remove the "_todo" suffix if a blank char is typed after the suffix:
#+BEGIN_SRC sh
compadd -P _todo -q foo bar blah
#+END_SRC
Add words in array $wordsarray to the list of possible completions
#+BEGIN_SRC sh
compadd -a wordsarray
#+END_SRC
* Testing & debugging
To reload a completion function:
#+BEGIN_SRC sh
> unfunction _func
> autoload -U _func
#+END_SRC
The following functions can be called to obtain useful information.
If the default keybindings don't work you can try pressing Alt+x and then enter the command name.
| Function | Default keybinding | Description |
|-----------------+--------------------+--------------------------------------------------------------------------------------------------------------------------------|
| _complete_help | Ctrl+x h | displays information about context names, tags, and completion functions used when completing at the current cursor position |
| _complete_help | Alt+2 Ctrl+x h | as above but displays even more information |
| _complete_debug | Ctrl+x ? | performs ordinary completion, but captures in a temporary file a trace of the shell commands executed by the completion system |
* Gotchas (things to watch out for)
Remember to include a #compdef line at the beginning of the file containing the completion function.
Take care to use the correct type of quoting for specifications to _arguments or _regex_arguments:
use double quotes if there is a parameter that needs to be expanded in the specification, single quotes otherwise,
and make sure to use different quotes around item descriptions.
Check that you have the correct number of :'s in the correct places for specifications for _arguments,
_alternative, _regex_arguments, etc.
Remember to include an initial pattern to match the command word when using _regex_arguments (it does not need a matching action).
Remember to put a null char $'\0' at the end of any PATTERN argument for _regex_arguments
* Tips
Sometimes you have a situation where there is just one option that can come after a subcommand, and zsh will complete this
automatically when tab is pressed after the subcommand. If instead you want it listed with its description before completing
you can add another empty option (i.e. \:) to the ACTION like this ':TAG:DESCRIPTION:((opt1\:"description for opt1" \:))'
Note this only applies to utility functions that use ACTIONs in their specification arguments (_arguments, _regex_arguments, etc.)
* Other resources
[[http://wikimatze.de/writing-zsh-completion-for-padrino.html][Here]] is a nicely formatted short tutorial showing basic usage of the _arguments function,
and [[http://www.linux-mag.com/id/1106/][here]] is a slightly more advanced tutorial using the _arguments function.
[[http://zsh.sourceforge.net/Doc/Release/Completion-System.html#Completion-System][Here]] is the zshcompsys man page.