path: root/doc/info/auto-auto-complete.texinfo

\input texinfo   @c -*-texinfo-*-

@c %**start of header
@setfilename auto-auto-complete.info
@settitle auto-auto-complete
@afourpaper
@documentencoding UTF-8
@documentlanguage en
@finalout
@c %**end of header


@dircategory Development
@direntry
* auto-auto-complete: (auto-auto-complete).        Autogenerate shell auto-completion scripts
@end direntry


@copying
Copyright @copyright{} 2014, 2015 Mattias Andrée

@quotation
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3 or
any later version published by the Free Software Foundation; with no
Invariant Sections, with no Front-Cover Texts, and with no Back-Cover
Texts. A copy of the license is included in the section entitled
``GNU Free Documentation License''.
@end quotation
@end copying

@ifnottex
@node Top
@top auto-auto-complete -- Autogenerate shell auto-completion scripts
@insertcopying
@end ifnottex

@titlepage
@title auto-auto-complete
@subtitle Autogenerate shell auto-completion scripts
@author by Mattias Andrée (maandree)

@page
@c @center `'
@vskip 0pt plus 1filll
@insertcopying
@end titlepage

@contents



@menu
* Overview::                        Brief overview of @command{auto-auto-complete}.
* Invoking::                        Invocation of @command{auto-auto-complete}.
* Syntax::                          The auto-auto-complete syntax.
* GNU Free Documentation License::  Copying and sharing this manual.
@end menu



@node Overview
@chapter Overview

@command{auto-auto-complete} provides a LISP-like
@footnote{A sane alternative to using XML.} declarative
language for creating auto-completion scripts for commands
in a shell-agnostic way. The current version of
@command{auto-auto-complete} can use such files to generate
auto-completion scripts for the @command{bash}, @command{zsh}
and @command{fish} shells.

@command{auto-auto-complete}'s language limited in comparsion
to for example raw auto-completion scripts for the @command{bash}
shell, however it is well enough for most projects.



@node Invoking
@chapter Invoking

@command{auto-auto-complete} recognises two options:

@table @option
@item -o
@itemx --output OUTPUT_FILE
Specifies the pathname of the generated file.

@item -s
@itemx -f
@itemx --file
@itemx --source SOURCE_FILE
Specifies the pathname of the auto-auto-complete script.
@end table

Both of these options must be used. Additionally
the shell that the generate file should be generated
for must be specified; this is done by adding name
of the shell as a stand-along argument, for example
@command{auto-auto-complete bash --output mycmd.bash --source mycmd.autocomplete}

It is also possible to define variables that can be
used the auto-auto-complete script. If you, for example,
want to give the variable @var{command} the value
@code{mycmd}, add the argument @option{command=mycmd}.
It is also possible to define arrays, for example
if you want the variable @var{srcopt} to be an array
of the for values @code{-s}, @code{-f}, @code{--source}
and @code{--file}, add the arguments @option{srcopt=-s},
@option{srcopt=-f}, @option{srcopt=--source} and
@option{srcopt=--file}. It is not possible to have
variable whose name begin with a dash (`-').

Alternatively you can run
@command{auto-auto-complete SHELL --where COMMAND}
(alternatively with @option{-w} instead of
@option{--where}). This will print the pathname
you should use for the generated file when installing
it. However the path prefix will not be included,
so if your package is installed to @file{/usr}
@footnote{The command being installed to @file{/usr/bin}
or @file{/usr/sbin}.}, you sould prepend @file{/usr}
to the output.



@node Syntax
@chapter Syntax

@command{auto-auto-complete} uses a LISP-like free form
syntax. Valid whitespace is normal blank space, horizontal
tab space@footnote{Also know simply as tab.},
carriage return@footnote{The first character in a new line
in for example the HTTP protocol and in Window's encoding
for new lines, it was the new line character in the classical
Mac operating systems}, line feed (new line) and form feed
(new page). Comments can be started with either a semicolon
(;) or a hash (#). Comments end at the next following
new line, which may either be a carriage return, line feed
or form feed. Comments cannot be started inside quotes.

The backslash character (\) can be used to force the
following character to be parsed verbatim, this is called
escaping. It is highly discourage to use this to escape
new lines, especially if the new line encoding used in the
document is carrige return–line feed, as that would only
escape the carrige return. There is also a set of characters
that have a special meaning when they are escaped:

@table @asis
@item a
Audible bell character.
@item b
Backspace character.
@item e
Escape character.
@item f
Form feed character.
@item n
Line feed character.
@item r
Carriage character.
@item t
Horizontal tab space character.
@item v
Vertical tab space character.
@item 0
Null character.
@end table

Quotes, either simple quotes (') or double quotes (")
can be used to parse all character verbatim except
backslash (\). A quote ends at the next quote character
that is not escaped by a backslash (\) and is identical
to the opening quote character. This is especially useful
for escaping whitespace and round brackets.

The rest of this chapter will demonstrate how to write a
script by example of @command{ponysay} (because it uses
most of the syntax).

The first thing we do is to declare which command the
script is for. We do this by creating the root brackets
and put the name of the command as the first element.

@example
(ponysay)
@end example

However, @command{ponysay} has a very similar command
called @command{ponythink}. It is sensible to let the
same script be used for auto-complete for both commands,
to do this we utilise that we can define variables
when we invoke @command{auto-auto-complete}.

@example
((value command))
@end example

Now when we compile this script we need invoke
@command{auto-auto-complete} with either
@option{command=ponysay} or @option{command=ponythink}.
If we want @command{ponysay} to be used if we
do not specify a value for @var{command} we instead
write:

@example
((value command ponysay))
@end example

Remember that we could give a variable multiple values.
This can also be done here. However in this example
we only want one value. For example, @command{((value var a b))}
would generate @command{(a b)} if @var{var} has not been set.

@command{ponysay} recognises the options @option{-h} and
@option{--help} for printing a summary of recognised options.
These options does not take any arguments and hence are specified
with @code{(unargumented)}.

@example
((value command ponysay)
  (unargumented (options -h --help)
                (desc 'Show summary of options'))
)
@end example

This only specifies that these option exists, if we also want
the generated scripts to suggest @option{--help} we need to
add @code{(complete --help)}.

@example
((value command ponysay)
  (unargumented (options -h --help)
                (complete --help)
                (desc 'Show summary of options'))
)
@end example

Now that we have our first option, lets add a few others, to
keep the example short, we will skip the most of the options.

@example
((value command ponysay)
  (unargumented (options -h --help)
                (complete --help)
                (desc 'Show summary of options'))
  (unargumented (options -l --list)
                (complete --list)
                (desc 'List all MLP:FiM ponies'))
  (unargumented (options +l ++list)
                (complete ++list)
                (desc 'List all non-MLP:FiM ponies'))
  (unargumented (options -X --256-colours --256colours --x-colours)
                (desc 'Use xterm colours'))
)
@end example

Now (especially if we had added all options) we have many
@code{(unargumented)} blocks. We can use @code{(multiple)}
so we do not have to write @code{unargumented} so many times.

@example
((value command ponysay)
  (multiple unargumented
    ((options -h --help) (complete --help)
     (desc 'Show summary of options'))
    ((options -l --list) (complete --list)
     (desc 'List all MLP:FiM ponies'))
    ((options +l ++list) (complete ++list)
     (desc 'List all non-MLP:FiM ponies'))
    ((options -X --256-colours --256colours --x-colours)
     (desc 'Use xterm colours'))
  )
)
@end example

To keep this example short we will truncate this to:

@example
((value command ponysay)
  (multiple unargumented ...) ;We have cut out the options.
)
@end example

@command{ponysay} also have a number of options that does
take an argument. We will add a few of them.

@example
((value command ponysay)
  (multiple unargumented ...) ;We have cut out the options.
  (multiple argumented
    ((options -f --file --pony) (complete --file --pony)
     (desc 'Specify the pony that should printed'))
    ((options -b --bubble --balloon) (complete --balloon)
     (desc 'Specify message balloon style'))
    ((options -W --wrap) (complete --wrap)
     (desc 'Specify wrapping column'))
    ((options +c --colour) (complete --colour)
     (desc 'Specify colour of the balloon, balloon link and message'))
  )
)
@end example

Just like @command{ponysay --help} prints @code{--wrap COLUMN}
to indicate that the argument for @option{--wrap} is is an
index of the column where the message printed by @command{ponysay}
is wrapped, shells could display the text @code{COLUMN} as
a placeholder for the next argument when you have typed
@option{--wrap}. To enable this in shells that support it,
we use @code{(arg)}.

@example
((value command ponysay)
  (multiple unargumented ...) ;We have cut out the options.
  (multiple argumented
    ((options -f --file --pony) (complete --file --pony) (arg PONY)
     (desc 'Specify the pony that should printed'))
    ((options -b --bubble --balloon) (complete --balloon) (arg STYLE)
     (desc 'Specify message balloon style'))
    ((options -W --wrap) (complete --wrap) (arg COLUMN)
     (desc 'Specify wrapping column)')
    ((options +c --colour) (complete --colour) (arg ANSI-COLOUR)
     (desc 'Specify colour of the balloon, balloon link and message'))
  )
)
@end example

The next step now is to specify the type of argument the options
want. To do this we use @code{(files)}. The elements in @code{(files)}
specify what type of file the shell should suggest. Multiple
type can be used. Tehe recognsied ones are:

@table @code
@item -0
Do not suggest files, or do not suggest files of types specified
after @code{-0}.
@item -a
Suggest all files.
@item -f
Suggest regular files and pipes.
@item -r
Suggest regular files but not pipes.
@item -p
Suggest pipes but not regular files.
@item -d
Suggest directories.
@item -l
Suggest symlinks. This is suggest by default,
but @code{-0} can be used to stop this.
@item -s
Suggest sockets.
@item -D
Suggest doors.
@item -b
Suggest block devices.
@item -c
Suggest character devices.
@item -S
Suggest block devices and character devices.
@end table

@example
((value command ponysay)
  (multiple unargumented ...) ;We have cut out the options.
  (multiple argumented
    ((options -f --file --pony) (complete --file --pony) (arg PONY)
     (files -f)
     (desc 'Specify the pony that should printed'))
    ((options -b --bubble --balloon) (complete --balloon) (arg STYLE)
     (files -f)
     (desc 'Specify message balloon style'))
    ((options -W --wrap) (complete --wrap) (arg COLUMN)
     (files -0)
     (desc Specify wrapping column))
    ((options +c --colour) (complete --colour) (arg ANSI-COLOUR)
     (files -0)
     (desc 'Specify colour of the balloon, balloon link and message'))
  )
)
@end example

@code{(files)} can also be used to specify patterns (using
@code{sh}-globbing) for the filenames of the files to suggest.
For example @code{--pony} in @command{ponysay} should only,
in respect to files, suggest files that end with @code{.pony}.

@example
((value command ponysay)
  (multiple unargumented ...) ;We have cut out the options.
  (multiple argumented
    ((options -f --file --pony) (complete --file --pony) (arg PONY)
     (files -f *.pony)
     (desc 'Specify the pony that should printed'))
    ((options -b --bubble --balloon) (complete --balloon) (arg STYLE)
     (files -f *.say)
     (desc 'Specify message balloon style'))
    ((options -W --wrap) (complete --wrap) (arg COLUMN)
     (files -0)
     (desc Specify wrapping column))
    ((options +c --colour) (complete --colour) (arg ANSI-COLOUR)
     (files -0)
     (desc 'Specify colour of the balloon, balloon link and message'))
  )
)
@end example

For @option{--balloon} files ending with @code{.say} should be
suggested if the completion is for @command{ponysay}, but for
@command{ponythink} @code{.think}-files should be suggest.
We can use @code{(case)} to select this based on the value of
the first element in the root block, which is the name of the command.

@example
((value command ponysay)
  (multiple unargumented ...) ;We have cut out the options.
  (multiple argumented
    ((options -f --file --pony) (complete --file --pony) (arg PONY)
     (files -f *.pony)
     (desc 'Specify the pony that should printed'))
    ((options -b --bubble --balloon) (complete --balloon) (arg STYLE)
     (files -f (case (ponysay *.say) (ponythink *.think)))
     (desc 'Specify message balloon style'))
    ((options -W --wrap) (complete --wrap) (arg COLUMN)
     (files -0)
     (desc Specify wrapping column))
    ((options +c --colour) (complete --colour) (arg ANSI-COLOUR)
     (files -0)
     (desc 'Specify colour of the balloon, balloon link and message'))
  )
)
@end example

Another part of options with arguments is suggestions that are not
based on filenames.

@example
((value command ponysay)
  (multiple unargumented ...) ;We have cut out the options.
  (multiple argumented
    ((options -f --file --pony) (complete --file --pony) (arg PONY)
     (suggest pony-f) (files -f *.pony)
     (desc 'Specify the pony that should printed'))
    ((options -b --bubble --balloon) (complete --balloon) (arg STYLE)
     (suggest balloon)
     (files -f (case (ponysay *.say) (ponythink *.think)))
     (desc 'Specify message balloon style'))
    ((options -W --wrap) (complete --wrap) (arg COLUMN)
     (suggest wrap) (files -0)
     (desc Specify wrapping column))
    ((options +c --colour) (complete --colour) (arg ANSI-COLOUR)
     (files -0)
     (desc 'Specify colour of the balloon, balloon link and message'))
  )
  (suggestion pony-f)  ;We will fill this in later...
  (suggestion balloon) ;We will fill this in later...
  (suggestion wrap)    ;We will fill this in later...
)
@end example

Lets cut out the options again to make this shorter.

@example
((value command ponysay)
  (multiple unargumented ...) ;We have cut out the options.
  (multiple argumented ...)   ;We have cut out the options.
  (suggestion pony-f)  ;We will fill this in later...
  (suggestion balloon) ;We will fill this in later...
  (suggestion wrap)    ;We will fill this in later...
)
@end example

Another part of ponysay is that it will take also
argument that are not associated with an option make
make a message it prints out of thiat.

@example
((value command ponysay)
  (default (arg MESSAGE) (files -0) (suggest message)
   (desc 'Message spoken by the pony'))
  (multiple unargumented ...) ;We have cut out the options.
  (multiple argumented ...)   ;We have cut out the options.
  (suggestion message) ;We will fill this in later...
  (suggestion pony-f)  ;We will fill this in later...
  (suggestion balloon) ;We will fill this in later...
  (suggestion wrap)    ;We will fill this in later...
)
@end example

A rather unusual part of @command{ponysay} is that
it has variadic options. A variadic option is a option
that takes all following arguments, unconditionally.
For example, in @command{ponysay} you can write
@code{--ponies twilight trixie} instead of
@code{--pony twilight --pony trixie}. @code{(variadic)}
is used to declare a variadic option. @code{(bind)}
becomes interesting here; because @command{ponysay}'s
variadic options have non-variadic counterparts, it
is pleasant to reuse the non-variadic options'
configurations. @code{(bind)} will copy everything
that is missing except @code{(options)} and
@code{(complete)}. In this example we will not
use @code{(complete)} because we do not want variadic
options to be suggest but we will use @code{(desc)}
because we want to adjust the descriptions.

@example
((value command ponysay)
  (default (arg MESSAGE) (files -0) (suggest message)
   (desc 'Message spoken by the pony'))
  (multiple unargumented ...) ;We have cut out the options.
  (multiple argumented ...)   ;We have cut out the options.
  (variadic (options --f --files --ponies) (bind -f)
            (desc 'Specify the ponies that may be printed'))
  (suggestion message) ;We will fill this in later...
  (suggestion pony-f)  ;We will fill this in later...
  (suggestion balloon) ;We will fill this in later...
  (suggestion wrap)    ;We will fill this in later...
)
@end example

Notice that we used @code{-f} for the element in @code{(bind)},
this is because we want @option{--f}, @option{--files} and
@code{--ponies} to have the same configurations (with exception
for the description) as the @option{-f} option.

Once again, to make the example shorter we will cut out
some parts.

@example
( ;We have cut out everything but the (suggestion):s.
  (suggestion message) ;We will fill this in later...
  (suggestion pony-f)  ;We will fill this in later...
  (suggestion balloon) ;We will fill this in later...
  (suggestion wrap)    ;We will fill this in later...
)
@end example

For @code{(suggestion message)} we want the word
`MESSAGE' to be suggested to let the user know
that the non-option arguments make up the message
that is printed (if used).

@example
( ;We have cut out everything but the (suggestion):s.
  (suggestion message (verbatim MESSAGE))
  (suggestion pony-f)  ;We will fill this in later...
  (suggestion balloon) ;We will fill this in later...
  (suggestion wrap)    ;We will fill this in later...
)
@end example

For @code{(suggestion pony-f)} we want, in addition
to the .pony-files which as already been configured,
.pony-files from @file{/usr/share/ponysay/ponies}
to be suggested without the .pony-suffix.

@example
( ;We have cut out everything but the (suggestion):s.
  (suggestion message (verbatim MESSAGE))
  (suggestion pony-f (ls "'/usr/share/ponysay/ponies'" .pony))
  (suggestion balloon) ;We will fill this in later...
  (suggestion wrap)    ;We will fill this in later...
)
@end example

However, if the shell supports executing comments to
get suggetions we want to utilise this.

@example
( ;We have cut out everything but the (suggestion):s.
  (suggestion message (verbatim MESSAGE))
  (suggestion pony-f (exec "'/usr/bin/ponysay'" --onelist)
                     (noexec ls "'/usr/share/ponysay/ponies'" .pony))
  (suggestion balloon) ;We will fill this in later...
  (suggestion wrap)    ;We will fill this in later...
)
@end example

@code{(suggestion balloon)} will work very similarly.

@example
( ;We have cut out everything but the (suggestion):s.
  (suggestion message (verbatim MESSAGE))
  (suggestion pony-f (exec "'/usr/bin/ponysay'" --onelist)
                     (noexec ls "'/usr/share/ponysay/ponies'" .pony))
  (suggestion balloon (exec "'/usr/bin/ponysay'" --balloonlist)
                      (no-exec ls "'/usr/share/ponysay/balloons'"
                       (case (ponysay .say) (ponythink .think))))
  (suggestion wrap) ;We will fill this in later...
)
@end example

The first thing we want to do for @option{--wrap}
is to give it some default suggestion.

@example
( ;We have cut out everything but (suggestion wrap).
  (suggestion wrap (verbatim none inherit 100 60))
)
@end example

The next step is to suggest the terminal's
width minus 10 columns. In the Bash shell this
can be calculated with
@command{$(( $(stty size <&2 | cut -d ' ' -f 2) - 10 ))}.

@example
( ;We have cut out everything but (suggestion wrap).
  (suggestion wrap
     (verbatim none inherit 100 60)
     (calc (pipe (stdin-fd (stty size) (stderr)) (cut -d ' ' -f 2)) - 10)
  )
)
@end example

As seen here @code{(pipe (a) (b) (c))} translates into
@code{(a | b | c)}. There are a few similar blocks that
can be used.

@table @code
@item (fullpipe (a) (b) (c))
@code{(a |& b |& c)}, or equivalently:
@code{(a 2>&1 | b 2>&1 | c)}
@item (cat (a) (b) (c))
@code{(a ; b ; c)}
@item (and (a) (b) (c))
@code{(a && b && c)}
@item (or (a) (b) (c))
@code{(a || b || c)}
@end table

It was also shown that @code{(stdin-fd (a) (stderr))} translates
into @code{a <&2}. @code{(stdin)}, @code{(stdout)} and @code{(stderr)}
translates into @code{0}, @code{1} and @code{2}, respectively.
Additional @code{(stdin-fd a b)} translates into @code{a <&b},
@code{(stdout-fd a b)} into @code{a >&b}, @code{(stderr-fd a b)} into @code{a 2>&b}
and @code{(fd-fd a b c)} into @code{a b<>&c}.
You can also redirect to files:

@table @code
@item (stdin a b)
@code{a < b}
@item (stdout a b)
@code{a > b}
@item (stderr a b)
@code{a 2> b}
@item (fd a b c)
@code{a b> c}
@end table



@node GNU Free Documentation License
@appendix GNU Free Documentation License
@include fdl.texinfo

@bye