Fix issue in keyword indentation with compound keywords

[ajhalme]

While playing around with Julia macros, I noticed that abstract type definitions inside a begin-end block were breaking the indentation in my Emacs.

I've never written anything in Emacs Lisp before, but the broken indentation was annoying me enough that I figured out how to fix it.

On closer inspection, I determined that the issue is with julia-last-open-block-pos and julia-at-keyword. The block position function determines the nesting level by scanning code preceding point one s-expressions at a time, keeping count of matched block starting and ending keywords. The julia-at-keyword function tries to match on keywords listed in julia-block-start-keywords, but unfortunately the matching doesn't quite work with compound keywords made up of more than one word.

Because julia-at-keyword uses (current-word), and because abstract and type are parsed as separate words, the matching misses the compound keyword abstract type (and primitive type and mutable struct). This results in the begin/end counting algorithm missing the compound keyword block beginnings, breaking the indentation.

This PR fixes this issue by introducing a new matching function specifically for compound keywords, julia-at-compound-keyword. This function looks at the current word and the one preceding it, and determines whether point is on the second word of a compound keyword. To make use of this function, I propose splitting the keyword list in two parts, with the compound words in their own list. The new function is applied as an additional clause in the block begin/end counting function.

I've included two tests that demonstrate the desired behaviour.

The performance cost from running the extra function should be negligible, though it is called for every s-expression step inside the matching algorithm.

[tpapp]

LGTM.

[tpapp]

@giordano @non-Jedi @FelipeLema If you have time a review would be appreciated.

Currently CI is not running, I ran the tests locally and the two font-lock failures seem unrelated.

[tpapp]

CI failure was local to a quirk my setup, sorry for the noise. Tests actually run clean.

[tpapp]

Closing and reopening to trigger CI.

[tpapp]

Incidentally, even if you have

never written anything in Emacs Lisp before

you write nice and clean code, so if you would be willing to review PRs, just comment in #97.

[tpapp]

@ajhalme: why did you close this?

[FelipeLema]

ohmydays I was looking forward to this... do you need any help to complete this?

      ∧_∧ 
　／＼(`･ω･)／＼ 
(　●　⊂　　 ⊃●　) 
　＼／/ 　 < ＼／ 
　∧_∧/⌒＼)∧_∧ 
（ ･ω･∩　∩･ω･ ) 
⊂　　 ノ　ヽ　　 つ 
`/　　/　　 ｜　　< 
(_ノヽ)　　 (_ノ＼_)