perf(lsp): scan project based on ScanKind
#6099
Conversation
| let execution = Execution::new_format(VcsTargeted { | ||
| staged: false, | ||
| changed: false, | ||
| }); |
There was a problem hiding this comment.
I really like this change, because it felt weird that project_key needed to be part of Execution.
github-actions
bot
added
A-Project
| pub fn parse_html_with_cache(source: &str, cache: &mut NodeCache) -> HtmlParse { | ||
| tracing::debug_span!("Parsing phase").in_scope(move || { | ||
| let mut parser = HtmlParser::new(source); | ||
| let mut parser = HtmlParser::new(source); |
There was a problem hiding this comment.
All these logs have been removed because they cause a lot of noise
| "biome_did_change_workspace_settings", | ||
| "workspace/didChangeWatchedFiles", | ||
| if let Some(base_path) = self.session.base_path() { | ||
| if let Some(folders) = self.session.get_workspace_folders() { |
There was a problem hiding this comment.
I noticed that we weren't considering workspace folders when the configuration changed. This should fix some bugs.
| watch: true, | ||
| force: false, | ||
| scan_kind: ScanKind::Project, | ||
| scan_kind, |
There was a problem hiding this comment.
Ah, if I understand correctly this is the main reason for this PR, right? The LSP no longer always needs to use ScanKind::Project, but it can use an actual scan_kind determined by the workspace.
There was a problem hiding this comment.
Yes! If you open a project that doesn't use project rules, the memory used by the biome process (daemon) will still be low because we don't crawl and save all files.
ematipico
force-pushed
the
chore/smarter-lsp-scanner
branch
from
9d7468a to
2dcd162
Compare
CodSpeed Performance ReportMerging #6099 will not alter performanceComparing Summary
|
ematipico
force-pushed
the
chore/smarter-lsp-scanner
branch
from
51f69bd to
fc449d5
Compare
Summary
This PR optimises the LSP scanner by computing the
ScanKindvariant when we create a project. The business logic we used in the CLI has been moved inside the Workspace. When we create a project, the Workspace also tells us how to scan it. This is very useful and functional. This is very convenient because it will allow us to tweak the LSP scanner by taking LSP settings into account.To implement this functionality, the method
open_projectnow accepts theskipandonlyparameters because they need to be factored into account when deciding how to scan a project. Of course, the LSP doesn't have any of those yet.Additionally, the CLI adds some further checks, which are the same as before. That's why the method
comput_scan_kindnow returns anOption, so we canunwrap_orand use one computed from the workspace.The rest of the changes are mostly code moved around, and unit tests updated.
Test Plan
CI should stay green, snapshots shouldn't change.