Skip to content

perf: optimize ESM linking passes #14758

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
JSerFeng merged 1 commit into from
Jul 14, 2026
Merged

perf: optimize ESM linking passes #14758

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
100 changes: 25 additions & 75 deletions crates/rspack_plugin_esm_library/src/link.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,12 +11,12 @@ use rspack_core::{
ChunkUkey, CodeGenerationPublicPathAutoReplace, Compilation, ConcatenatedModuleIdent,
ConditionalInitFragment, DependencyType, ExportInfo, ExportMode, ExportProvided,
ExportsInfoArtifact, ExportsType, FindTargetResult, ImportSpec, InitFragmentKey, ModuleGraph,
ModuleGraphCacheArtifact, ModuleIdentifier, ModuleInfo, NAMESPACE_OBJECT_EXPORT, PathData,
RuntimeGlobals, SideEffectsStateArtifact, SourceType, URLStaticMode, UsageState, UsedName,
UsedNameItem, collect_ident, escape_name_atom_ref, find_new_name, find_target,
get_cached_readable_identifier, get_js_chunk_filename_template, get_module_directives,
get_module_hashbang, property_access, property_name, reserved_names::RESERVED_NAMES_ATOM_SET,
rspack_sources::ReplaceSource, split_readable_identifier, to_normal_comment,
ModuleGraphCacheArtifact, ModuleIdentifier, ModuleInfo, NAMESPACE_OBJECT_EXPORT, RuntimeGlobals,
SideEffectsStateArtifact, SourceType, URLStaticMode, UsageState, UsedName, UsedNameItem,
collect_ident, escape_name_atom_ref, find_new_name, find_target, get_cached_readable_identifier,
get_module_directives, get_module_hashbang, property_access, property_name,
reserved_names::RESERVED_NAMES_ATOM_SET, rspack_sources::ReplaceSource,
split_readable_identifier, to_normal_comment,
};
use rspack_error::{Diagnostic, Error, Result};
use rspack_plugin_javascript::{
Expand Down Expand Up @@ -621,25 +621,25 @@ impl EsmLibraryPlugin {
let mut namespace_object_sources: IdentifierMap<String> = IdentifierMap::default();
let mut namespace_re_export_star_cache = IdentifierMap::default();
for (ukey, mut needed_namespace_objects) in needed_namespace_objects_by_ukey {
let mut visited = FxHashSet::default();

let chunk_link = link.get_mut_unwrap(&ukey);

// webpack require iterate the needed_namespace_objects and mutate `needed_namespace_objects`
// at the same time, https://github.com/webpack/webpack/blob/1f99ad6367f2b8a6ef17cce0e058f7a67fb7db18/lib/optimize/ConcatenatedModule.js#L1514
// Which is impossible in rust, using a fixed point algorithm to reach the same goal.
// Process the insertion-ordered set as a worklist. Newly discovered namespace objects are
// handled in the next round without cloning all previously visited modules.
let mut next_namespace_object = 0;
loop {
let mut changed = false;
let round_end = needed_namespace_objects.len();
if next_namespace_object == round_end {
break;
}

// using the previous round snapshot `needed_namespace_objects` to iterate, and modify the
// original `needed_namespace_objects` during the iteration,
// if there is no new id inserted into `needed_namespace_objects`, break the outer loop
for module_info_id in needed_namespace_objects.clone().iter() {
if visited.contains(module_info_id) {
continue;
}
visited.insert(*module_info_id);
changed = true;
for index in next_namespace_object..round_end {
let module_info_id = needed_namespace_objects
.get_index(index)
.copied()
.expect("index is within the current worklist");
let module_info_id = &module_info_id;

let module_info = concate_modules_map[module_info_id].as_concatenated();
let mut runtime_template = compilation.runtime_template.create_module_code_template();
Expand Down Expand Up @@ -903,9 +903,7 @@ var {} = {{}};
.runtime_requirements
.insert(*runtime_template.runtime_requirements());
}
if !changed {
break;
}
next_namespace_object = round_end;
}
}

Expand Down Expand Up @@ -1377,54 +1375,6 @@ var {} = {{}};
external_module_init_fragments: &mut IdentifierMap<ChunkInitFragments>,
) -> Result<()> {
let runtime_template = compilation.runtime_template.create_chunk_code_template();
let mut outputs = FxHashMap::<ChunkUkey, String>::default();
let module_keys: Vec<ModuleIdentifier> = orig_concate_modules_map.keys().copied().collect();
for m in &module_keys {
if compilation
.build_chunk_graph_artifact
.chunk_graph
.get_module_chunks(*m)
.is_empty()
{
// orphan module
continue;
}

let chunk_ukey = Self::get_module_chunk(*m, compilation)?;
let chunk = compilation
.build_chunk_graph_artifact
.chunk_by_ukey
.expect_get(&chunk_ukey);
let filename_template = get_js_chunk_filename_template(
chunk,
&compilation.options.output,
&compilation.build_chunk_graph_artifact.chunk_group_by_ukey,
);

let output_path = compilation
.get_path_with_info(
&filename_template,
PathData::default()
.chunk(chunk.ukey(), compilation)
.chunk_hash_optional(chunk.rendered_hash(
&compilation.chunk_hashes_artifact,
compilation.options.output.hash_digest_length,
))
.chunk_id_optional(chunk.id().map(|id| id.as_str()))
.chunk_name_optional(chunk.name_for_filename_template())
.content_hash_optional(chunk.rendered_content_hash_by_source_type(
&compilation.chunk_hashes_artifact,
&SourceType::JavaScript,
compilation.options.output.hash_digest_length,
))
.runtime(chunk.runtime().as_str()),
&mut Default::default(),
)
.await
.expect("should have output path");
outputs.insert(chunk_ukey, output_path);
}

let concate_modules_map = std::mem::take(orig_concate_modules_map);
let map = rspack_parallel::scope::<_, _>(|token| {
for (m, info) in concate_modules_map {
Expand Down Expand Up @@ -2649,9 +2599,9 @@ var {} = {{}};
// foo; // access foo
// foo; // access foo again, but no require call
// ```
for m in chunk_link.hoisted_modules.clone() {
for m in &chunk_link.hoisted_modules {
let module = module_graph
.module_by_identifier(&m)
.module_by_identifier(m)
.expect("should have module");

// make sure all side-effect modules are rendered
Expand Down Expand Up @@ -2718,7 +2668,7 @@ var {} = {{}};

Self::add_require(
ref_module,
Some(m),
Some(*m),
/*
do not specify the symbol now, if it really needs to be used, it will be
modified later by get_binding
Expand All @@ -2730,7 +2680,7 @@ var {} = {{}};
}
}

let codegen_res = compilation.code_generation_results.get(&m, None);
let codegen_res = compilation.code_generation_results.get(m, None);
let concatenation_scope = codegen_res
.concatenation_scope
.as_ref()
Expand All @@ -2744,7 +2694,7 @@ var {} = {{}};
}

let Some(binding) = Self::get_binding(
Some(m),
Some(*m),
module_graph,
&compilation.module_graph_cache_artifact,
&compilation.exports_info_artifact,
Expand Down
Loading