Highlight visible kind application, fixes to visible type application #135
Conversation
sheaf
force-pushed
the
vta
branch
2 times, most recently
from
af0f4eb
to
be029bb
Compare
syntaxes/haskell.YAML-tmLanguage
Outdated
| @@ -573,6 +547,35 @@ repository: | |||
| | } # A block closed? Maybe this should also include `;`, because non-indentation based `do` | |||
| ) | |||
| | ^(?!\1\s+\S|\s*$) # at least one, no-further indented, non-whitespace character. I.e. a same-level declaration/implementation | |||
| type_application: | |||
| patterns: | |||
| - begin: '(?<=\s)(@)(\()' | |||
There was a problem hiding this comment.
I am not sure about the look-behind here. As your original issue (#73) quoted, the @ may not be preceeded by any end-of-identifier character, which is why I just copied the legal identifier character class regex. It may be that the majority there overlaps with the inversion of \s, but at least in cases where you have (my expr)@type it should probably still parse as a type application, what do you think?
There was a problem hiding this comment.
I realize this is a very niche case though 😄
There was a problem hiding this comment.
Good point. It seems a bit subtle indeed, as actually
f = (id)@([Int])is not parsed by GHC, whereas
f = [1,2,3]@[Int]
g = A{f=1}@Boolare both parsed as type applications. I'm not sure exactly then which are the relevant characters, do you have a suggestion for a regex? Like [\s\}\]] or something.
syntaxes/haskell.YAML-tmLanguage
Outdated
| name: meta.type-application.haskell | ||
| patterns: | ||
| - include: '#type_signature' | ||
| - begin: '(?<=\s)(@)\s*(\[)' |
There was a problem hiding this comment.
now that I think about it ... do you think we could somehow only trigger on the @ (plus look behind) and then dispatch into another set of begin-end rules for [ ], ( ) and ident expressions? The termination condition for the outer multiline match would be the concatenation of the inner end conditions, perhaps as a look-behind. This way we don't have to duplicate the complicated start condition so many times? I think it could actually work.
There was a problem hiding this comment.
I haven't been able to figure this one out. Feel free to give it a go, otherwise I think I'll just use the current solution.
|
I used the lookbehind (:+:)@(A B C)which is valid record at syntax. @JustusAdam I hope this is OK now. I don't know how to avoid the triple lookbehind for the visible type application, let me know if you have a solution. |
|
I just realised we are also allowed the following: f @'(a,b)
f @'[a,b,c]I'll add those to this PR soon. Edit: Done (+rebased). |
Type application is now detected by: space, then
@, then an expression of the following three forms:(,)[,][\p{Ll}_\p{Lu}\p{Lt}\p{Nd}'']I added some tricky situations to the test case to check that it's behaving as expected, but please add more cases if you can spot other issues/potential issues.
Note that we're doing better than GitHub on this one! 😄
Fixes test case
T0073.