Elm,Units: Type and constructor tags exclude parameters etc.

[niksilver]

This has been tested with

make units UNITS=simple-elm

which passes.

This is based on a problem encountered in Vim using ctags. It's possible this problem is entirely different (or maybe isn't even a problem!) with a different Vim configuration, a different editor, or a different use case.

To understand the problem being addressed, consider the following code:

type Param a
    = Cons a
    | Other a

baz : Param Int
baz =
    Cons 42

If the cursor is positioned under the Cons of the last line, then hitting Ctrl-] should take us to the definition of Cons on line 2. However, this fails because Vim identifies the tag under the cursor as the pure alphanumeric string Cons, but the tag in the tag file is Cons a.

Similarly for when the cursor is placed under Param in the third line from the bottom.

Also, if the definition of Param was

type Param a =
    Cons a
    | Other a

then its tag would be recorded as Param a =. Even though that's not the recommended way to format Elm it would still be helpful to generate a useful tag.

So this change is to have tags for types and constructors limited to the initial alphanumeric part, which solves all these problems.

[codecov]

Codecov Report

Merging #3297 (bad9fe8) into master (38c046d) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #3297   +/-   ##
=======================================
  Coverage   85.27%   85.27%           
=======================================
  Files         211      211           
  Lines       49826    49826           
=======================================
  Hits        42488    42488           
  Misses       7338     7338           

Impacted Files	Coverage Δ
optlib/elm.c	0.00% <ø> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 38c046d...bad9fe8. Read the comment docs.

[masatake]

@niksilver, thank you.
type variables (type parameters?) like a can be stored to signature: field.
Do you have an interest in filling the field for tags like:

Param	input.elm	/^type Param a$/;"	kind:type	signature:a
Cons	input.elm	/^    = Cons a$/;"	kind:constructor	signature:a	scope:type:Param	roles:def
Other	input.elm	/^    | Other a$/;"	kind:constructor	signature:a	scope:type:Param	roles:def

[niksilver]

Hello @masatake. Thank you for accepting my code change.

I'm currently trying to build a PackCC parser for Elm which will be more robust than the current regexp parser. As part of that I will see if I can add your suggestions. However, it's very slow going as I'm only a hobbyist developer. So I dare not give a timeframe, but it is something I'm working on.

[masatake]

I'm currently trying to build a PackCC parser for Elm which will be more robust than the current regexp parser. As part of that I will see if I can add your suggestions. However, it's very slow going as I'm only a hobbyist developer. So I dare not give a timeframe, but it is something I'm working on.

Interesting.

Have you ever considered using the --_mtable-regex-<LANG> option?
https://docs.ctags.io/en/latest/optlib.html?highlight=mtable-regex#adding-a-regex-to-a-regex-table

If you are interested in the option, I will write a small initial version of mtable-regex based Elm parser.

[niksilver]

Thanks, @masatake, that looks very promising. I'm going to pursue the PackCC parser for a while longer, but will revert to the mtable regexp approach if I feel like I'm losing the battle.

[masatake]

@niksilver does your ctags link to pcre2?
I would like to write an initial version of mtable based Elm parser using pcre2.

[niksilver]

Apparently no pcre2...

$ ./ctags --list-features
#NAME             DESCRIPTION
gnulib_regex      linked with the Gnulib regular expression library
iconv             can convert input/output encodings
option-directory  TO BE WRITTEN
optscript         can use the interpreter
packcc            has peg based parser(s)
regex             can use regular expression based pattern matching
wildcards         can use glob matching