Support completion for field names inside recursive records. #1088
Conversation
This only returns field identifiers which are in the current scope. If a field declaration is in scope, it means that we have access to it and we should include it as a possible completion identiifer.
😎 |
There was a problem hiding this comment.
Code looks good, but I tried this inside the mantis benchmark and it didn't work. That being said, I don't know if it didn't because it was too slow, or something related to coc.nvim not liking the LSP. Does it work for you @ebresafegaga ? If you go inside types.ncl, pick a random record, and add a new filed that refers to one of its sibling (which is a record), do you get completion?
|
I'm not sure what is going on, but I know in some cases, if the program incomplete, the parser might not be able to correctly parse the record with good error recovery. I get completion if I do this (note the comma at the end): If the comma is not present, no completion is given |
|
Got it. In any case incomplete programs are a more general problem, and orthogonal to this PR. It's good to go for me 👍 |
Implements item 4 in #877
As a consequence of this change, we get completion on stdlib names (such as
string,array,contracts, etc), because they are record field names which are in scope.