Kate

This article was originally published as a blog at http://www.alweb.dk/blog/anders/autoindentation at 2004/11/02.

NOTE Mon Nov 1 16:49:03 CET 2004: revised versions of the variables for both perl and bash. Also update kdelibs/kate/part.

Some time ago, when I convinced Laurence Withers to contribute a XML auto-indenter to Kate, I got the idea that we should also have a more flexible indenter that could be configured using the document variable system. The other day while commenting on this bug I was reminded again. So in one of my usual streaks away from my overfilled TODO list, I sat down and implemented one last night..

The commit log in a friendly formatting looks like this:

Add a variable based autoindenter.

This indenter matches a line and the first line with content above against
regular expressions to decide if a line should be [un]indented. When a newline
is entered, both the old and the new line is considered, and it can be triggered
to indent the current by typing one of a list of characters.

The regular expression patterns and trigger character list are read from document variables, which can be embedded directly in the document or read from a filetype configuration.

The following variables are used:

var-indent-indent-after

a regular expression pattern, if the first line with content above the current, indentation is added

var-indent-indent

a regular expression pattern, if the current line matches indent is added. this is not tested if the above matched.

var-indent-unindent

a regular expression pattern, if it matches the current line, indent is removed.

var-indent-triggerchars

a string of characters, when any is typed, the current line is reindented.

This should be usable for adding basic autoindentation to bash, lua, basic, perl...
Note that Kate's document variables currently can't contain a semicolon, (would be a common trigger char), I'm going to fix that.

How to use it

I'd really appreciate it if some of you would like help me testing this! If you run cvs HEAD and write bash or perl code, please consider using the configurations below, and let me know how it works out. If you want to define variables for another language, don't hesitate to come into #kate at irc.freenode.net and ask me for help.

The general idea is to store the variables needed to define indentation for language X in a filetype configuration, but the variables can of cause be stored in any location from where kate will read them. So far, I have defined variables that seems to work with bash script and with perl. It looks like this:

Variables required to autoindent bash script:

# kate: indent-mode varindent;
# kate: var-indent-indent-after (\{(?![^\}]*\})|\b(then|elif|else)\b(?!.+fi)|\\bdo\b(?!.+done)|\bcase\s+.+\s+in\b(?!.*esac)|\[\[);
# kate: var-indent-indent \$\{.*\};
# kate: var-indent-unindent ([}]\s*$|\b(fi|elif|else)\b|\bdone\b|\besac\b|\]\]);
# kate: var-indent-triggerchars {};

The entry from ~/.kde/share/config/katefiletyperc looks like this (NOTE that the Variables property has been broken into several lines, the continued lines are indented. The double backslashes seems to be a KConfig thing.):

[Bash]
Mimetypes=application/x-shellscript
Section=Scripts
Variables=kate: indent-mode varindent;
  var-indent-indent-after (\\{(?![^\\}]*\\})|\b(then|elif|else)\b(?!.+fi)\\bdo\b(?!.+done)|\\bcase\\s+\\s+in\\b(?!.*esac)|\\[\\[);
  var-indent-indent \\$\\{.*\\};
  var-indent-unindent (\\}|\\b(fi|elif|else)\\b|\\bdone\b|\\besac\\b|\\]\\]);
  var-indent-triggerchars {};
Wildcards=*.sh

Explanation:

1. The var-indent-indent-after variable matches any line that looks like an unfinished bash control loop, or like it has an opening curly brace or a double square bracket opening.
2. The var-indent-unindent variable matches any of the corresponding closing elements.
3. The var-indent-indent variable adds indentation if the line has variables enclosed in curly braces, in anticipation of the unindenter removing it again. This might seem a bit silly, but so far I haven't a better solution.
4. Finally, the var-indent-triggerchars variable causes immediate indentation when a curly brace is entered. (the textual closings will be unindented when/if a newline is entered.)

Variables required to autoindent perl code:

(hey, who'd have thought that those would be simpler?)

# kate: indent-mode varindent;
# kate: var-indent-indent-after (?:\{(?![^}]*\})|\((?![^\)]*\)));
# kate: var-indent-indent (?:\{.*\}|\(.*\));
# kate: var-indent-unindent [})];
# kate: var-indent-triggerchars });

... and the corresponding filetyperc entry:

[Perl]
Mimetypes=text/x-perl
Section=Scripts
Variables=kate: indent-mode varindent;
  var-indent-indent-after (?:\\{(?![^}]*\\})|\\((?![^\)]*\\)));
  var-indent-indent (?:\\{.*\\}|\\(.*\\));
  var-indent-unindent [})];
  var-indent-triggerchars });
Wildcards=*.pl;*.pm

This works much like the bash configuration, but simpler, since all i care for is indenting [curly brace] blocks and multiline parens. For perl, a semicolon shoudl really cause reindenting, but kate can't read that from a variable yet :-\

...uhm.. I'm going to have to create a working configuration for CSS ASAP!