Analyze completion site re: trailing parens #680
Conversation
| }, | ||
| "sourceLanguages": ["rust"] |
There was a problem hiding this comment.
I think this is a neutral-to-positive move for us.
| use crate::treesitter::NodeTypeExt; | ||
| use crate::treesitter::UnaryOperatorType; | ||
|
|
||
| #[allow(dead_code)] |
There was a problem hiding this comment.
Several of these fields are not actually used, but they were used along the way, i.e. when I had more logging in. I imagine this entire struct will go away when we do #681, so seems ok to have around for now.
| } | ||
| } | ||
|
|
||
| pub fn check_for_function_value(context: &DocumentContext, node_context: &NodeContext) -> bool { |
There was a problem hiding this comment.
I revised these to be less nested -- I liked that aspect of #369 once I discovered it.
| .clone() | ||
| } | ||
|
|
||
| #[test] |
There was a problem hiding this comment.
I decided to test this here after all, as opposed to testing individual completion sources. It feels like we just want to know that the completion is correct, regardless of which helper ultimately provided it. It didn't seem to make sense to test the helpers themselves, since there's nothing particularly tricky about them correctly using the no_trailing_parens boolean. Again #681 is going to present a big opportunity to revisit testing that relates to "is the CompletionItem correct and is it provided by the expected source(s)?"
There was a problem hiding this comment.
Used the basic structure of a test I found in #369.
Addresses posit-dev/positron#1818
Addresses posit-dev/positron#2338
Closes #369 (I just came across this earlier effort that stalled out in draft from)
I think of this as a "big and dumb" solution to this problem. But it works! The discussion I want to have is basically hinted at in #1818. I suspect it's time to create some formal
CompletionOptionstruct that gets formed (probably) inprovide_completions()that contains some analysis of the completion site. This will then be passed down to all the lower level functions that marshal completions.Summary on what we decided together:
bigsmall and dumb solution along these lines, to enjoy immediate improvements in our completions for?and insidedebug()and friends.