Index all assigned variables #620
Conversation
Stopgap until we are in a better position to track the scopes of objects in a document. This causes a first definition to be the symbol of reference. Jumping from a reference to a definition then always moves up.
| // Super hacky but let's wait until the typed API to do better | ||
| if !lhs_text.starts_with("method(") && !lhs.is_identifier_or_string() { | ||
| return Ok(None); | ||
| } | ||
|
|
||
| let start = convert_point_to_position(contents, lhs.start_position()); | ||
| let end = convert_point_to_position(contents, lhs.end_position()); | ||
|
|
||
| Ok(Some(IndexEntry { | ||
| key: lhs_text.clone(), | ||
| range: Range { start, end }, | ||
| data: IndexEntryData::Variable { name: lhs_text }, | ||
| })) |
There was a problem hiding this comment.
Seems pretty reasonable even with the hackiness for method(
| indexer::IndexEntryData::Variable { name } => { | ||
| let completion = match completion_item_from_variable(name) { | ||
| Ok(item) => item, | ||
| Err(err) => { | ||
| log::error!("{err:?}"); | ||
| return; | ||
| }, | ||
| }; | ||
| completions.push(completion); | ||
| }, |
There was a problem hiding this comment.
So this ensures that for the workspace we now add variables to the completion list even if they don't exist yet in the R session. For example in a fresh restart of clock we now get these global constants:
I think that's probably fine, even though sending that to the console will result in an error. I guess its still useful when writing in the editor without loading the package?
However, note that after you load the package, the completion shows up with a different symbol:
There was a problem hiding this comment.
I think this will all get cleaned up once we build a stronger model of object scopes.
| IndexEntryData::Variable { name } => { | ||
| info.push(SymbolInformation { | ||
| name: name.clone(), | ||
| kind: SymbolKind::VARIABLE, | ||
| location: Location { | ||
| uri: Url::from_file_path(path).unwrap(), | ||
| range: entry.range, | ||
| }, | ||
| tags: None, | ||
| deprecated: None, | ||
| container_name: None, | ||
| }); | ||
| }, |
There was a problem hiding this comment.
Looks like workspace symbols are working here:
There was a problem hiding this comment.
So I guess workspace symbols are enough to power jump to definition, neat
| deprecated: None, | ||
| container_name: None, | ||
| }); | ||
| }, | ||
| }; | ||
| }); | ||
|
|
There was a problem hiding this comment.
It's somewhat interesting to note that for document symbols, i.e. with @, we were already tracking all variables due to how document_symbols() and eventually index_assignment() works.
It even already worked with method( calls. We don't check that the LHS is an identifier, we just accept anything. That does lead to some weirdness but its probably ok.
So it was just the following two places where we weren't indexing variables but we do now:
- Workspace level symbols (so
#nameworks across files, and so you can jump to defn) - Workspace level static analysis completions (even though we previously did get a completion item already due to runtime completions once the user loaded the package)
There was a problem hiding this comment.
oh I hadn't caught up on that, good catch
Quick fix to improve S7 workflow.
Addresses posit-dev/positron#5274
Also a nice improvement all around as you can jump to the definition of e.g. a data frame in a script.
Positron Release Notes
New Features
Bug Fixes