Fix some comments disappearing in array sugar. #5947
Conversation
The ids for `Array.get` and `Array.set` used to desugar array access have location that overlaps with the arguments, stealing the comments. This takes care of some examples similar to the one in #5946 but not that specific example.
|
Note: with this the error message misses a location for: let a = [1]
module Array = {}
let _ = a[0]the message being: |
res_syntax/src/res_core.ml
Outdated
| @@ -2009,20 +2008,18 @@ and parseBracketAccess p expr startPos = | |||
| let accessExpr = parseConstrainedOrCoercedExpr p in | |||
| Parser.expect Rbracket p; | |||
| Parser.eatBreadcrumb p; | |||
| let rbracket = p.prevEndPos in | |||
There was a problem hiding this comment.
I'm not able to access the computer now so have not run it to see if it works.
Can we use stringStart position got rbracket for loc?
There was a problem hiding this comment.
I've tried a few different locations, such as [ or ] only, but that already breaks some printing tests.
I could get it to work with mkLoc startPos startPos, which does not break tests. The error message for Array.get not found has at least the right line in that case.
res_syntax/src/res_core.ml
Outdated
| match p.token with | ||
| | Equal -> | ||
| Parser.leaveBreadcrumb p ExprArrayMutation; | ||
| Parser.next p; | ||
| let rhsExpr = parseExpr p in | ||
| let arraySet = | ||
| Location.mkloc (Longident.Ldot (Lident "Array", "set")) arrayLoc | ||
| Location.mknoloc (Longident.Ldot (Lident "Array", "set")) | ||
| in | ||
| let endPos = p.prevEndPos in | ||
| let arraySet = | ||
| Ast_helper.Exp.apply ~loc:(mkLoc startPos endPos) |
There was a problem hiding this comment.
For now, this could be alternatives:
Ast_helper.Exp.apply ~loc:(mkLoc lbracket rbracket)What do you think?
There was a problem hiding this comment.
The example case's ouput looks like this:
// This comment will vanish
(
incidents
->Belt.Array.keep(({status}) => status === #OPEN)
->Belt.SortArray.stableSortBy((a, b) => compare(a.createdTime, b.createdTime))
)[0]
The comment location looks weird, but at least we wouldn't lost the loc for error msg.
There was a problem hiding this comment.
I have a separate question. Why does it matter what location one gives to array functions given that they are not printed? I'd like to see where in the printing mechanism that matters, and whether one can make it be ignored.
Fixes #5946
The ids for
Array.getandArray.setused to desugar array access have location that overlaps with the arguments, stealing the comments.This takes care of some examples similar to the one in #5946 but not that specific example.