feat: request autocompletion without typing a letter #310
Conversation
juleswritescode
changed the title
There was a problem hiding this comment.
looks good to me overall. just would like to double check if we really need to instantiate the tree sitter parser with every request 😢. can't we just mutate the tree?
also: we will have a bunch of conflicts, and a few things will need to be handled differently. but no issues, just that we should merge yours first so I can fix the merge conflicts with mine.
| sql.push(c); | ||
| } | ||
|
|
||
| let mut parser = tree_sitter::Parser::new(); |
There was a problem hiding this comment.
isnt there a way to modify the tree sitter tree instead of reparsing the string?
There was a problem hiding this comment.
maybe we can pass the passer down to here?
There was a problem hiding this comment.
I think editing the node isn to possible, so maybe we should try to pass the existing tree sitter parser down.
I think I would shy away from that – we would mutate the tree that is also used for other features… seems like a source of very annoying bugs :/ Also, keep in mind that:
I added some benchmarks – on my machine, reusing the old treesitter takes about 500 nanoseconds, and creating a new one about 5-150 microseconds depending on the size of the sql. 
I mean, I can play a Counterspell when I see a hakbal in about that time, but I wouldn't mind waiting that long for completions |
|
@psteinroe Curious about your opinion: I tried moving the We could make the |
|
ah sorry, i didn't mean to move the mapper there. I was thinking to clone or copy the tree sitter parser to skip the initialisation costs. Will look at the pr later, thanks for trying! 🙏🏼 |
| @@ -24,3 +28,167 @@ impl IntoIterator for CompletionsResult { | |||
| self.items.into_iter() | |||
| } | |||
| } | |||
|
|
|||
| pub(crate) fn get_statement_for_completions<'a>( | |||
There was a problem hiding this comment.
why can't this be implemented as a Filter in parsed_document?
There was a problem hiding this comment.
because I need to peek into the next statement for the overlap check :/
If you have other ideas, let me know!
There was a problem hiding this comment.
makes sense. maybe we can find a good abstraction for this but for now lets roll with this.
So far, we were relying on complete tree-sitter trees for autocompletion.
This works fine if you type the first letter of a node, say:
select * from u|Now, the
uis considered a table node, and we can look for tables.But this PR makes it such that we can also request completions if we haven't typed the
uyet.It does so by checking whether the tree is valid and, if not, injecting a token to fix the tree-sitter tree.
Additionally, in the case of
select * from |, the workspace would consider the cursor to be outside the statement. I added a little offset so the cursor still matches a statement if it's 2 characters after the last token.