Increase performance by 50%

[kisaragi-hiu]

After this package was released, parinfer-mode was made, which vendored this package as this package isn't in a package archive. That package was eventually deprecated as the author started recommending parinfer-rust-mode instead.

As a long time user of Parinfer, I found the pure elisp version to be perfectly usable, with a few low-hanging fruit optimizations to be taken (it was needlessly running parinferlib-indent-mode twice when opening files, for example), so I've been focused on optimizing and improving my own fork of parinfer-mode instead of moving onto parinfer-rust-mode.

As part of this effort I also experimented with some changes to this library, eventually landing on the two main changes here: lexical binding and using global variables instead of passing a hash table around. These add up to a 50% improvement in performance (perf.el reports indent-mode going from 0.540383s to 0.278933s on my machine with the same number of GCs).

As this repository is not archived I figured I should probably upstream these changes.

• performance
  • Use global variables instead of passing around a result hash table.
  • This is safe as state is always reset by parinferlib--create-initial-result (now renamed to parinferlib--initialize)
  • This brought perf.el's indent mode time from 0.51s down to 0.38s on my machine
  • Use lexical binding
  • This further brought perf.el's indent mode time from 0.38s down to 0.28s on my machine
• tests
  • use a Makefile (this made it quicker for me to run tests with a task runner plugin)
  • test byte-compiled version
  • performance: average over 30 runs
• style
  • Make the package header conform to convension
  • use (when lst) not (when (parinferlib--not-empty? lst)) as empty list = nil
  • remove indent_size option in editorconfig as that produces incorrect indentation for emacs lisp

[oakmac]

This is really wonderful and impressive ❤️ 🤓

Thank you for the high quality contribution!