kakoune/doc/interfacing.asciidoc
2014-03-02 02:08:17 +00:00

124 lines
4.1 KiB
Plaintext

Interfacing Kakoune with external programs
==========================================
In order to interact with the external world, Kakoune uses the shell, mainly
through the +%sh{ ... }+ string type, and it's control socket.
Basic interaction
-----------------
For synchronous operations, +%sh{ ... }+ blocks are easy to use, they behave
similarly to +$( ... )+ shell construct.
For example, one can echo the current time in Kakoune status line using:
[source,bash]
----
:echo %sh{ date }
----
For asynchronous operations, the Kakoune Unix stream socket can be used. This
is the same socket that Kakoune clients connect to. It is available through
+kak_session+ environment variable: the socket is +/tmp/kak-${kak_session}+
For example, we can echo a message in Kakoune in 10 seconds with:
[source,bash]
----
:nop %sh{ (
sleep 10
echo "eval -client '$kak_client' 'echo sleep ended'" |
kak -p ${kak_session}
) >& /dev/null < /dev/null & }
----
* The +nop+ command is used so that any eventual output from the
+%sh{ ... }+ is not interpreted by Kakoune
* When writing to the socket, Kakoune has no way to guess in which
client's context the command should be evaluated. A temporary
context is used, which does not have any user interface, so if we want
to interact with the user, we need to use the +eval+ command, with
it's +-client+ option to send commands to a specific client.
* For the command to run asynchronously, we wrap it in a sub shell
with parenthesis, redirect it's +std{in,err,out}+ to +/dev/null+, and
run it in background with +&+. Using this pattern, the shell does
not wait for this sub shell to finish before quitting.
Interactive output
------------------
It is a frequent interaction mode to run a program and display it's output
in a Kakoune buffer.
The common pattern to do that is to use a fifo buffer:
[source,bash]
-----
%sh{
# Create a temporary fifo for communication
output=$(mktemp -d -t kak-temp-XXXXXXXX)/fifo
mkfifo ${output}
# run command detached from the shell
( run command here >& ${output} ) >& /dev/null < /dev/null &
# Open the file in Kakoune and add a hook to remove the fifo
echo "edit! -fifo %{output} *buffer-name*
hook buffer BufClose .* %{ nop %sh{ rm -r $(dirname ${output}} }"
}
-----
This is a very simple example, most of the time, the echo command will as
well contains
-----
set buffer filetype <...>
-----
and some hooks for this filetype will have been written
Completion candidates
---------------------
Most of the time, filetype specific completion should be provided by
external programs.
external completions are provided using the +completions+ option, which
have the following format.
----
line.column[+len]@timestamp:candidate1:candidate2:...
----
the first element of this string list specify where and when this completions
applies, the others are simply completion candidates.
As a completion program may take some time to compute the candidates, it should
run asynchronously. In order to do that, the following pattern may be used:
[source,bash]
-----
# Declare the option which will store the temporary filename
decl str plugin_filename
%sh{
# ask Kakoune to write current buffer to temporary file
filename=$(mktemp -t kak-temp.XXXXXXXX)
echo "set buffer plugin_filename '$filename'
write '$filename'"
}
# End the %sh{} so that it's output gets executed by Kakoune.
# Use a nop so that any eventual output of this %sh does not get interpreted.
nop %sh{ ( # launch a detached shell
buffer="${kak_opt_plugin_filename}"
line="${kak_cursor_line}"
column="${kak_cursor_column}"
# run completer program and put output in colon separated format
candidates=$(completer $buffer $line $column | completer_filter)
# remove temporary file
rm $buffer
# generate completion option value
completions="$line.$column@$kak_timestamp:$candidates"
# write to Kakoune socket for the buffer that triggered the completion
echo "set buffer=${kak_bufname} completions '$completions'" |
kak -p ${kak_session}
) >& /dev/null < /dev/null & }
-----