crates/biome_html_syntax/src/element_ext.rs (1)

295-297: New public method lacks rustdoc and has a somewhat generic signature.

This public method would benefit from a brief rustdoc comment explaining its purpose. The method name is_style_scoped suggests checking for the scoped attribute specifically, but the parameter allows any attribute name—consider whether a more descriptive name or dedicated methods (e.g., is_style_scoped() without params, has_style_attribute(name)) would be clearer.

Suggested documentation

+    /// Returns `true` if this is a `<style>` tag that has the specified attribute.
+    ///
+    /// This is commonly used to detect scoped styles in frameworks like Vue (`<style scoped>`).
     pub fn is_style_scoped(&self, attribute_name: &str) -> bool {
         self.is_style_tag() && self.find_attribute_by_name(attribute_name).is_some()
     }

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_html_syntax/src/element_ext.rs` around lines 295 - 297, The new
public function is_style_scoped currently lacks rustdoc and has a generic
signature that accepts any attribute name; add a brief Rustdoc above pub fn
is_style_scoped(&self, attribute_name: &str) -> bool describing that it checks
whether this element is a <style> tag and contains the given attribute, and then
either (A) rename the method to has_style_attribute(&self, name: &str) to
reflect the general check and add rustdoc, or (B) keep is_style_scoped() as a
no-arg method that checks specifically for the "scoped" attribute and implement
a new pub fn has_style_attribute(&self, name: &str) -> bool that calls
self.is_style_tag() && self.find_attribute_by_name(name).is_some(); update
callers accordingly and document both methods; reference the existing helper
find_attribute_by_name when writing the docs.

crates/biome_html_analyze/tests/spec_tests.rs (1)

216-219: Scope snapshot normalisation more narrowly.

Line 218 replaces backslashes across the whole snapshot, which can also alter fixture source text. Prefer normalising only rendered path fragments.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_html_analyze/tests/spec_tests.rs` around lines 216 - 219, The
current unconditional snapshot.replace('\\', "/") (inside the needs_module_graph
branch) can alter fixture source text; change it to only normalize backslashes
inside rendered path fragments by applying the replacement to matched path-like
substrings rather than the whole snapshot. Locate the needs_module_graph block
and replace the global call on snapshot with a targeted transformation that
scans snapshot for path patterns (e.g., Windows drive or path-like regex
matches) and replaces backslashes with forward slashes only within those
matches, preserving other text in snapshot.

crates/biome_module_graph/tests/snap/mod.rs (1)

48-54: Stabilise snapshot file ordering once, then iterate without cloning.

Sorting by path in constructors avoids order-dependent snapshots and lets assert_snapshot iterate self.files directly.

Proposed patch

-        let files = fs
+        let mut files: Vec<_> = fs
             .files
             .read()
             .iter()
             .map(|(file, entry)| {
                 let content = entry.lock();
                 let content = String::from_utf8_lossy(content.as_slice()).into_owned();
                 (file.as_str().to_string(), content)
             })
             .collect();
+        files.sort_by(|(a, _), (b, _)| a.cmp(b));
         Self {
             module_graph,
             files,
             resolver: None,
         }

-    pub fn from_files(module_graph: &'a ModuleGraph, files: Vec<(String, String)>) -> Self {
+    pub fn from_files(
+        module_graph: &'a ModuleGraph,
+        mut files: Vec<(String, String)>,
+    ) -> Self {
+        files.sort_by(|(a, _), (b, _)| a.cmp(b));
         Self {
             module_graph,
             files,
             resolver: None,
         }
     }

-        let files: Vec<_> = self.files.clone();
-
         let dependency_data = self.module_graph.data();
-        for (file_name, source_code) in &files {
+        for (file_name, source_code) in &self.files {

Also applies to: 65-69

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_module_graph/tests/snap/mod.rs` around lines 48 - 54, The
constructor from_files should stabilise file ordering by sorting the incoming
files Vec by path before storing it (e.g. call files.sort_by(|a,b|
a.0.cmp(&b.0)) or equivalent) so snapshots are order-independent, and then store
the sorted Vec in the files field (module_graph, files, resolver) rather than
leaving it unsorted; apply the same deterministic sort change to the other
constructor variant handling files (the one around lines 65-69) and ensure
callers/consumers (e.g. tests using assert_snapshot) iterate self.files directly
without cloning.

crates/biome_module_graph/src/html_module_info/visitor.rs (1)

72-79: Drop the now-unused stylesheet parameter from visit_html_element.

Since stylesheet extraction moved to visit_self_closing_element, this parameter is dead weight.

Suggested patch

             if let Some(element) = HtmlElement::cast(node.clone()) {
                 self.visit_html_element(
                     element,
-                    &mut referenced_classes,
-                    &mut imported_stylesheets,
+                    &mut referenced_classes,
                 );
             } else if let Some(element) = HtmlSelfClosingElement::cast(node) {
                 self.visit_self_closing_element(element, &mut imported_stylesheets);
             }
@@
     fn visit_html_element(
         &self,
         element: HtmlElement,
         referenced_classes: &mut Vec<CssClassReference>,
-        _imported_stylesheets: &mut Vec<ResolvedPath>,
     ) {

Also applies to: 131-136

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_module_graph/src/html_module_info/visitor.rs` around lines 72 -
79, The call sites pass imported_stylesheets into visit_html_element but
stylesheet extraction was moved to visit_self_closing_element, so remove the
now-unused stylesheet parameter from the visit_html_element signature and all
its call sites (e.g., change the call inside the HtmlElement branch that
currently calls visit_html_element(element, &mut referenced_classes, &mut
imported_stylesheets) to only pass element and &mut referenced_classes), and
update the other occurrence referenced in the comment (around the second call at
the 131-136 region) to match; ensure visit_self_closing_element still accepts
&mut imported_stylesheets and that visit_html_element’s implementation no longer
expects or uses the stylesheet parameter.

crates/biome_service/src/workspace/server.rs (2)

164-167: Gate this test-only accessor behind the testing feature.

The doc says “for use in tests”, but Lines 164-167 are exposed in non-testing builds too. Keeping it feature-gated avoids widening the public surface by accident.

Suggested patch

     /// Returns the module graph, for use in tests.
+    #[cfg(feature = "testing")]
     pub fn module_graph(&self) -> Arc<ModuleGraph> {
         self.module_graph.clone()
     }

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_service/src/workspace/server.rs` around lines 164 - 167, Gate
the test-only accessor module_graph behind the testing feature by adding a
conditional compilation attribute to the method (e.g., annotate the pub fn
module_graph(&self) with #[cfg(feature = "testing")]) so it is only compiled
when that feature is enabled; optionally add #[cfg_attr(docsrs, doc(cfg(feature
= "testing")))] above the function to make the feature requirement visible in
docs and keep the signature (pub fn module_graph(&self) -> Arc<ModuleGraph>)
unchanged.

---

178-196: Don’t swallow indexing failures in the test helper.

Line 184 currently drops the Result. If parsing/indexing fails, tests can silently proceed with a partial graph.

Suggested patch

     #[cfg(feature = "testing")]
     pub fn index_files_for_test(
         &self,
         project_key: ProjectKey,
         files: impl IntoIterator<Item = (BiomePath, DocumentFileSource)>,
-    ) {
+    ) -> Result<(), WorkspaceError> {
         for (path, document_file_source) in files {
-            let _ = self.open_file_internal(
+            self.open_file_internal(
                 OpenFileReason::Index(IndexTrigger::InitialScan),
                 OpenFileParams {
                     project_key,
                     path,
                     content: FileContent::FromServer,
                     document_file_source: Some(document_file_source),
                     persist_node_cache: false,
                     inline_config: None,
                 },
-            );
+            )?;
         }
+        Ok(())
     }

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_service/src/workspace/server.rs` around lines 178 - 196, The
helper index_files_for_test currently drops the Result from open_file_internal
which hides parsing/indexing failures; update index_files_for_test to surface
errors by changing its signature to return a Result (e.g., Result<(), E> or
anyhow::Result<()>) and propagate the open_file_internal result using the ?
operator (or collect and return the first Err), or alternatively call
.expect/.unwrap with a clear message to fail fast in tests; ensure you reference
the open_file_internal call inside index_files_for_test and propagate errors
originating from OpenFileReason::Index / OpenFileParams so tests don't continue
on partial indexing.

crates/biome_module_graph/src/html_module_info/mod.rs (1)

96-103: Rustdoc for style_classes is stale after the type change.

The comment still says “Each TokenText…”, but the field is now IndexSet<CssClassDefinition>.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_module_graph/src/html_module_info/mod.rs` around lines 96 - 103,
The rustdoc for the field style_classes is out of date: it still refers to
TokenText even though the field type is now IndexSet<CssClassDefinition>. Update
the comment for the style_classes field to describe that it stores
CssClassDefinition entries (not TokenText), mention that these are collected
from CssClassSelector nodes in embedded CSS ASTs (no re-parsing), and briefly
note what a CssClassDefinition represents (e.g., a class name plus any
associated metadata like source/span) so readers know why the richer type is
used.

crates/biome_html_analyze/src/lint/nursery/no_undeclared_classes.rs (1)

139-148: Build the import tree once per attribute, not once per missing class.

Lines 139-148 recompute the same tree for every missing class token. Cache it lazily in run and reuse it for all emitted signals.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_html_analyze/src/lint/nursery/no_undeclared_classes.rs` around
lines 139 - 148, The code currently calls
module_graph.build_import_tree_for_html(file_path) for every missing class
inside the loop (when !found_class), which recomputes the same import tree
repeatedly; change this by computing and caching the import tree once per
attribute-run: add an Option or lazy variable in run (e.g., let mut
cached_import_tree: Option<ImportTree> = None) and inside the loop, when you
first need it call module_graph.build_import_tree_for_html(file_path) and store
it in cached_import_tree, then reuse cached_import_tree.as_ref() for all
subsequent UndeclaredClass signals instead of calling build_import_tree_for_html
repeatedly; update references in the block that constructs UndeclaredClass (the
import_tree field) to use the cached value.

crates/biome_test_utils/src/lib.rs (1)

374-388: Make recursive file collection deterministic for stable snapshots.

read_dir ordering is filesystem-dependent; sorting entries here reduces snapshot churn and “works on my machine” noise.

Suggested patch

 fn collect_all_files_in_dir(dir: &Utf8Path, out: &mut Vec<Utf8PathBuf>) {
     let Ok(entries) = std::fs::read_dir(dir.as_std_path()) else {
         return;
     };
-    for entry in entries.flatten() {
+    let mut entries: Vec<_> = entries.flatten().collect();
+    entries.sort_by_key(|entry| entry.path());
+    for entry in entries {
         let Ok(path) = Utf8PathBuf::try_from(entry.path()) else {
             continue;
         };
         if path.is_dir() {
             collect_all_files_in_dir(&path, out);

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_test_utils/src/lib.rs` around lines 374 - 388, The recursive
collector collect_all_files_in_dir uses std::fs::read_dir which yields
filesystem-dependent ordering; make it deterministic by collecting entries into
a Vec, mapping/filtering to Utf8PathBufs (using the existing try_from logic),
sorting that Vec by a stable key (e.g., path.as_str() or path.to_string()) and
then iterating the sorted list to recurse or push to out so snapshots are stable
across machines; preserve the existing error/continue behavior for entries that
fail conversion.

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

crates/biome_html_syntax/src/element_ext.rs (1)

297-299: Please add a brief rustdoc for is_style_scoped.

Line 297 adds a public helper; a short doc comment would clarify intended attribute names (scoped, global, framework-specific variants) for callers.

Suggested tweak

+    /// Returns `true` when this is a `<style>` tag and `attribute_name` is present
+    /// on its opening tag (for example, `scoped` or framework-specific style flags).
     pub fn is_style_scoped(&self, attribute_name: &str) -> bool {
         self.is_style_tag() && self.find_attribute_by_name(attribute_name).is_some()
     }

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_html_syntax/src/element_ext.rs` around lines 297 - 299, Add a
short Rust doc comment for the public method is_style_scoped explaining its
purpose and the expected attribute names it checks (e.g., "scoped", "global",
and any framework-specific variants), and mention that it returns true only for
style tags when find_attribute_by_name(attribute_name) is Some. Update the doc
above pub fn is_style_scoped(&self, attribute_name: &str) -> bool to reference
related helpers is_style_tag and find_attribute_by_name for context.

crates/biome_module_graph/tests/snap/mod.rs (1)

65-69: Stabilise file ordering before rendering snapshots.

Snapshot output currently depends on incoming files order. Sorting once here will prevent order-based flakes across callers and environments.

Suggested patch

-        let files: Vec<_> = self.files.clone();
+        let mut files: Vec<_> = self.files.iter().collect();
+        files.sort_unstable_by(|(left, _), (right, _)| left.cmp(right));

         let dependency_data = self.module_graph.data();
-        for (file_name, source_code) in &files {
+        for (file_name, source_code) in files {
             let file_name = Utf8PathBuf::from(file_name.as_str());

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_module_graph/tests/snap/mod.rs` around lines 65 - 69, The
snapshot flakiness is caused by relying on the incoming order of self.files; fix
it by sorting the cloned files Vec before rendering. After let files: Vec<_> =
self.files.clone(); sort the files collection (e.g., using
files.sort_unstable_by_key(|(name, _)| name.clone()) or a stable sort if needed)
so the subsequent loop for (file_name, source_code) in &files produces
deterministic ordering for snapshots.

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

crates/biome_module_graph/src/html_module_info/mod.rs (1)

96-103: Doc comment is stale after the type change.

The text still says each entry is a TokenText, but the field now stores CssClassDefinition (name + applicability).

Suggested doc tweak

-    /// Each `TokenText` represents a single class name (e.g., "header" from `.header`).
+    /// Each `CssClassDefinition` stores a class name (e.g., "header" from `.header`)
+    /// plus its applicability (Local/Global/Unknown).

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_module_graph/src/html_module_info/mod.rs` around lines 96 - 103,
The doc comment for the field style_classes is stale: it still claims each entry
is a TokenText but the field type is now IndexSet<CssClassDefinition>. Update
the comment on style_classes to describe that each entry is a CssClassDefinition
containing the class name and applicability (or similar metadata) rather than
TokenText, and remove or adjust the references to TokenText so the documentation
matches the CssClassDefinition structure.

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

crates/biome_module_graph/tests/spec_tests.rs (1)

2899-3230: ⚠️ Potential issue | 🟡 Minor

Doc/test wording still disagrees with actual behaviour.

Line 2899 says plain HTML <style> is “always Global”, but Line 2962 asserts Unknown. Likewise, names like “are_global/hidden” in this block conflict with assertions that local classes appear in same-file traversal (for example Lines 3079, 3131, 3161, 3224). Please align names/comments with what is asserted.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_module_graph/tests/spec_tests.rs` around lines 2899 - 3230, The
docstrings/test names/comments misstate behavior: plain HTML inline style
applicability is EmbeddingStyleApplicability::Unknown (not "always Global") and
many tests that say "hidden" actually assert Local classes still appear in
same-file traversal; update the wording to match assertions. Edit the
html_css_source doc comment and the test name/comment for
test_html_inline_style_classes_are_global to reflect "Unknown" applicability (or
rename to test_html_inline_style_classes_are_unknown), and change comments/names
that claim "hidden"/"are_hidden" (e.g.,
test_astro_local_style_classes_are_hidden,
test_svelte_local_style_classes_are_hidden,
test_vue_scoped_style_classes_are_local_and_hidden) to indicate "Local (hidden
from cross-file traversal but present in same-file traversal)" or similar
consistent phrasing so the test names and inline comments match the assertions
using EmbeddingStyleApplicability and traverse_import_tree_for_html_classes
behavior.

crates/biome_module_graph/src/html_module_info/visitor.rs (1)

145-168: Consider a shared class-attribute collector for both element kinds.

The class extraction logic is duplicated across visit_html_element and visit_self_closing_element; extracting one helper would reduce drift and prevent this bug class from reappearing.

Also applies to: 183-201

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_module_graph/src/html_module_info/visitor.rs` around lines 145 -
168, Duplicate class-attribute extraction appears in visit_html_element and
visit_self_closing_element; refactor by extracting a helper (e.g.,
collect_class_attrs_from_opening) that takes the opening node (the value used in
opening.attributes()/opening) plus &self.file_path and referenced_classes and
encapsulates the logic currently inside both loops (checking
attr.as_html_attribute(), getting name_token via
attr.name().ok().and_then(|name| name.value_token().ok()), comparing
name_text.eq_ignore_ascii_case("class"), then calling
collect_class_attribute_reference on the initializer/value_node). Replace the
duplicated blocks in visit_html_element and visit_self_closing_element with a
single call to this helper.

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

crates/biome_module_graph/src/html_module_info/visitor.rs (1)

135-139: Consider removing the unused _imported_stylesheets parameter.

This parameter is never used because <link> elements are self-closing and handled in visit_self_closing_element. If kept for API symmetry, a brief comment explaining the rationale would help future readers.

♻️ Suggested simplification

     fn visit_html_element(
         &self,
         element: HtmlElement,
         referenced_classes: &mut Vec<CssClassReference>,
-        _imported_stylesheets: &mut Vec<ResolvedPath>,
     ) {

And update the call site at line 73-77:

                 self.visit_html_element(
                     element,
                     &mut referenced_classes,
-                    &mut imported_stylesheets,
                 );

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@crates/biome_module_graph/src/html_module_info/visitor.rs` around lines 135 -
139, The visit_html_element method currently accepts an unused parameter
_imported_stylesheets; remove this parameter from the visit_html_element
signature in visitor.rs and update all call sites (where
Visitor::visit_html_element is invoked) to no longer pass the
imported_stylesheets argument, or alternatively if you prefer API symmetry keep
the parameter but rename it to _imported_stylesheets and add a short comment
inside visit_html_element explaining that imported stylesheets are handled in
visit_self_closing_element (e.g., "Imported <link> stylesheets are handled in
visit_self_closing_element; parameter kept for symmetry"), and ensure any
compiler warnings are resolved by using the underscore-prefixed name.

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro