Rewite indentation engine

[taku0]

New indentation engine, based on that of swift-mode.

See doc/indentation_logic/pages.pdf for its design.

Pros:

• Better indentation for most cases
• Electric indentations
• Comprehensive tests

Cons:

• Complex
• Hard to read (sorry for a monolithic big commit)
• Hard to maintain
• Slower; indent-region on entire sample.kt takes a couple of seconds

Options:

• Merge this branch as is (I wish so) or with minor tweaks.
• Remove codes for corner cases and make it works only for practical cases.
• Reject this PR and fix problems with other method.

[gregghz]

do you have any sense of what can be done about performance? I'd say a couple seconds for one file might be too slow.

[taku0]

Two measures may improve the performance:

• Caching syntactic information
• Moving some code from lexer to indentation logic

Caching syntactic information

When indenting a line, we analyze the text around the point. In most cases, analyzing a few lines is sufficient, but it may analyzes more lines in some cases. Caching analyzed information may improve the performance when indenting a bulk of lines. Actually, we analyze the position of brackets eagerly and cache them. Caching more information would complicate the code and might be slower for simple cases.

Moving some code from lexer to indentation logic

The code is separated into two layers: the lexer and the indentation logic.

The lexer is a pair of functions, kotlin-mode--forward-token and kotlin-mode--backward-token. For most cases, it just looks a char or a word, but for some tokens, it scans around for a few lines to distinguish ambiguous token types. The distinction of the types is not always required in the indentation logic, so moving some code into the indentation logic may improve the performance. It will complicate the indentation logic and may violate the abstraction.

[knobo]

Has development of this patch stopped?

[taku0]

@knobo
I have rebased on the master branch and resolved conflicts.
This PR doesn't address Kotlin language updates, if any, since Jan 2020.

[taku0]

Add support for value class.
https://kotlinlang.org/docs/inline-classes.html

[taku0]

Rebased and fixed conflicts.

[pyloolex]

@taku0, did you get a chance to verify #74?

[taku0]

Is there any possibility that this PR will be merged?
I have improved performance so that indent-region on entire sample.kt takes only 0.164911s now.
Maintainability could still be a problem. If you add me to Emacs-Kotlin-Mode-Maintainers, I'll be glad to maintain it.

[gregghz]

@taku0 Thanks for this work. I'm happy to add you as a maintainer. Myself and @russel haven't been very active here. I'll give @russel a few days to chime in, but if i don't hear anything I'll go ahead and add you as an owner/maintainer.

[taku0]

@gregghz Thank you. I'm working on GitHub Actions for tests and linter for each Emacs versions. After that, I will investigate #55. Maybe porting code of swift-mode will fix it.