chore: optimize side effects detect

.vscode/settings.json

@@ -5,9 +5,12 @@
   "prettier.enable": false,
   "editor.defaultFormatter": "biomejs.biome",
   "[typescript]": {
-    "editor.defaultFormatter": "biomejs.biome",
+    "editor.defaultFormatter": "biomejs.biome"
   },
   "[javascript]": {
-    "editor.defaultFormatter": "biomejs.biome",
+    "editor.defaultFormatter": "biomejs.biome"
   },
+  "[rust]": {
+    "editor.defaultFormatter": "rust-lang.rust-analyzer"
+  }
 }

crates/compiler/tests/fixtures/tree_shake/cjs/output.js

@@ -1,5 +1,5 @@
 //index.js:
- (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};r[e](i,i.exports,o,n);t[e]=i;return i.exports}o(e)})({"d2214aaa":function  (module, exports, farmRequire, farmDynamicRequire) {
+ (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};t[e]=i;r[e](i,i.exports,o,n);return i.exports}o(e)})({"d2214aaa":function  (module, exports, farmRequire, farmDynamicRequire) {
     console.log("runtime/index.js")(globalThis || window || global)["__farm_default_namespace__"].__farm_module_system__.setPlugins([]);
 }
 ,},"d2214aaa");(function(_){for(var r in _){_[r].__farm_resource_pot__='index_6889.js';(globalThis || window || global)['__farm_default_namespace__'].__farm_module_system__.register(r,_[r])}})({"3da733a3":function  (module, exports, farmRequire, farmDynamicRequire) {

crates/compiler/tests/fixtures/tree_shake/decl/self-ref/output.js

@@ -1,5 +1,5 @@
 //index.js:
- (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};r[e](i,i.exports,o,n);t[e]=i;return i.exports}o(e)})({"ec853507":function  (module, exports, farmRequire, farmDynamicRequire) {
+ (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};t[e]=i;r[e](i,i.exports,o,n);return i.exports}o(e)})({"ec853507":function  (module, exports, farmRequire, farmDynamicRequire) {
     console.log("runtime/index.js")(globalThis || window || global)["__farm_default_namespace__"].__farm_module_system__.setPlugins([]);
 }
 ,},"ec853507");(function(_){for(var r in _){_[r].__farm_resource_pot__='index_d7d4.js';(globalThis || window || global)['__farm_default_namespace__'].__farm_module_system__.register(r,_[r])}})({"3e3af5b6":function  (module, exports, farmRequire, farmDynamicRequire) {

crates/compiler/tests/fixtures/tree_shake/deep_dynamic_import/output.js

@@ -45,7 +45,7 @@
 ,});
 
 //index.js:
- (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};r[e](i,i.exports,o,n);t[e]=i;return i.exports}o(e)})({"d2214aaa":function  (module, exports, farmRequire, farmDynamicRequire) {
+ (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};t[e]=i;r[e](i,i.exports,o,n);return i.exports}o(e)})({"d2214aaa":function  (module, exports, farmRequire, farmDynamicRequire) {
     console.log("runtime/index.js")(globalThis || window || global)["__farm_default_namespace__"].__farm_module_system__.setPlugins([]);
 }
 ,},"d2214aaa");(function(_){for(var r in _){_[r].__farm_resource_pot__='index_5d9b.js';(globalThis || window || global)['__farm_default_namespace__'].__farm_module_system__.register(r,_[r])}})({"7c4a34c2":async function  (module, exports, farmRequire, farmDynamicRequire) {

crates/compiler/tests/fixtures/tree_shake/empty_module/output.js

@@ -1,5 +1,5 @@
 //index.js:
- (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};r[e](i,i.exports,o,n);t[e]=i;return i.exports}o(e)})({"d2214aaa":function  (module, exports, farmRequire, farmDynamicRequire) {
+ (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};t[e]=i;r[e](i,i.exports,o,n);return i.exports}o(e)})({"d2214aaa":function  (module, exports, farmRequire, farmDynamicRequire) {
     console.log("runtime/index.js")(globalThis || window || global)["__farm_default_namespace__"].__farm_module_system__.setPlugins([]);
 }
 ,},"d2214aaa");(function(_){for(var r in _){_[r].__farm_resource_pot__='index_6b9f.js';(globalThis || window || global)['__farm_default_namespace__'].__farm_module_system__.register(r,_[r])}})({"569704c1":function  (module, exports, farmRequire, farmDynamicRequire) {

crates/compiler/tests/fixtures/tree_shake/remove_module/output.js

@@ -1,5 +1,5 @@
 //index.js:
- (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};r[e](i,i.exports,o,n);t[e]=i;return i.exports}o(e)})({"d2214aaa":function  (module, exports, farmRequire, farmDynamicRequire) {
+ (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};t[e]=i;r[e](i,i.exports,o,n);return i.exports}o(e)})({"d2214aaa":function  (module, exports, farmRequire, farmDynamicRequire) {
     console.log("runtime/index.js")(globalThis || window || global)["__farm_default_namespace__"].__farm_module_system__.setPlugins([]);
 }
 ,},"d2214aaa");(function(_){for(var r in _){_[r].__farm_resource_pot__='index_ecb7.js';(globalThis || window || global)['__farm_default_namespace__'].__farm_module_system__.register(r,_[r])}})({"569704c1":function  (module, exports, farmRequire, farmDynamicRequire) {

crates/compiler/tests/fixtures/tree_shake/self-executed/decl/var_side_effect/output.js

@@ -1,5 +1,5 @@
 //index.js:
- (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};r[e](i,i.exports,o,n);t[e]=i;return i.exports}o(e)})({"0b3bded0":function  (module, exports, farmRequire, farmDynamicRequire) {
+ (globalThis || window || global)['__farm_default_namespace__'] = {__FARM_TARGET_ENV__: 'browser'};(function(r,e){var t={};function n(r){return Promise.resolve(o(r))}function o(e){if(t[e])return t[e].exports;var i={id:e,exports:{}};t[e]=i;r[e](i,i.exports,o,n);return i.exports}o(e)})({"0b3bded0":function  (module, exports, farmRequire, farmDynamicRequire) {
     console.log("runtime/index.js")(globalThis || window || global)["__farm_default_namespace__"].__farm_module_system__.setPlugins([]);
 }
 ,},"0b3bded0");(function(_){for(var r in _){_[r].__farm_resource_pot__='index_5314.js';(globalThis || window || global)['__farm_default_namespace__'].__farm_module_system__.register(r,_[r])}})({"b5d64806":function  (module, exports, farmRequire, farmDynamicRequire) {
@@ -31,10 +31,6 @@
         throw error;
     }
     var Context = React.createContext();
-    var LOADABLE_REQUIRED_CHUNKS_KEY = "__LOADABLE_REQUIRED_CHUNKS__";
-    function getRequiredChunkKey(namespace) {
-        return "" + namespace + LOADABLE_REQUIRED_CHUNKS_KEY;
-    }
     var sharedInternals = Object.freeze({
         __proto__: null,
         getRequiredChunkKey: getRequiredChunkKey,
@@ -329,7 +325,6 @@
             return null;
         }
     }), loadable$1 = _createLoadable$1.loadable, lazy$1 = _createLoadable$1.lazy;
-    var BROWSER = typeof window !== "undefined";
     function loadableReady(done, _temp) {
         if (done === void 0) {
             done = function done() {};

crates/core/src/module/mod.rs

@@ -323,6 +323,10 @@ impl ScriptModuleMetaData {
     self.ast = ast;
   }
 
+  pub fn take_comments(&mut self) -> CommentsMetaData {
+    std::mem::replace(&mut self.comments, CommentsMetaData::default())
+  }
+
   pub fn set_comments(&mut self, comments: CommentsMetaData) {
     self.comments = comments;
   }

crates/plugin_tree_shake/src/init_tree_shake_module_map.rs

@@ -3,29 +3,32 @@ use std::collections::HashMap;
 use farmfe_core::{
   module::{module_graph::ModuleGraph, ModuleId},
   parking_lot::Mutex,
-  rayon::iter::{IntoParallelRefIterator, ParallelIterator},
+  rayon::iter::{IntoParallelIterator, ParallelIterator},
   swc_common::GLOBALS,
 };
 
 use crate::module::TreeShakeModule;
 
 pub fn init_tree_shake_module_map(
-  module_graph: &ModuleGraph,
+  module_graph: &mut ModuleGraph,
   context: &std::sync::Arc<farmfe_core::context::CompilationContext>,
 ) -> HashMap<ModuleId, TreeShakeModule> {
   let tree_shake_modules_map =
     Mutex::new(std::collections::HashMap::<ModuleId, TreeShakeModule>::new());
-  module_graph.modules().par_iter().for_each(|module| {
-    if !module.module_type.is_script() || module.external {
-      return;
-    }
+  module_graph
+    .modules_mut()
+    .into_par_iter()
+    .for_each(|module| {
+      if !module.module_type.is_script() || module.external {
+        return;
+      }
 
-    GLOBALS.set(&context.meta.script.globals, || {
-      let tree_shake_module = TreeShakeModule::new(module);
-      tree_shake_modules_map
-        .lock()
-        .insert(module.id.clone(), tree_shake_module);
+      GLOBALS.set(&context.meta.script.globals, || {
+        let tree_shake_module = TreeShakeModule::new(module);
+        tree_shake_modules_map
+          .lock()
+          .insert(module.id.clone(), tree_shake_module);
+      });
     });
-  });
   tree_shake_modules_map.into_inner()
 }

crates/plugin_tree_shake/src/mark_initial_side_effects.rs

@@ -2,18 +2,18 @@ use std::collections::HashMap;
 
 use farmfe_core::module::{module_graph::ModuleGraph, ModuleId};
 
-use crate::module::TreeShakeModule;
+use crate::module::{TreeShakeModule, UsedExports};
 
 pub fn mark_initial_side_effects(
   module_graph: &ModuleGraph,
   tree_shake_modules_map: &mut HashMap<ModuleId, TreeShakeModule>,
 ) -> Vec<ModuleId> {
   let mut entry_module_ids = vec![];
 
-  // mark entry modules as side_effects
   for (entry_module_id, _) in module_graph.entries.clone() {
+    // mark entry modules as UsedExports::All
     if let Some(tree_shake_module) = tree_shake_modules_map.get_mut(&entry_module_id) {
-      tree_shake_module.side_effects = true;
+      tree_shake_module.pending_used_exports = UsedExports::All;
     }
 
     entry_module_ids.push(entry_module_id);
@@ -42,21 +42,21 @@ pub fn mark_initial_side_effects(
     }
   }
 
-  // update contains_self_executed_stmt for the tree_shake_modules
-  for tree_shake_module_id in module_graph.toposort().0 {
-    if let Some(tree_shake_module) = tree_shake_modules_map.get(&tree_shake_module_id) {
-      let contains_self_executed_stmt = tree_shake_module.contains_self_executed_stmt;
-      module_graph
-        .dependents_ids(&tree_shake_module_id)
-        .into_iter()
-        .for_each(|dept_id| {
-          if let Some(dept_tree_shake_module) = tree_shake_modules_map.get_mut(&dept_id) {
-            dept_tree_shake_module.contains_self_executed_stmt =
-              dept_tree_shake_module.contains_self_executed_stmt || contains_self_executed_stmt
-          }
-        });
-    }
-  }
+  // // update contains_self_executed_stmt for the tree_shake_modules
+  // for tree_shake_module_id in module_graph.toposort().0 {
+  //   if let Some(tree_shake_module) = tree_shake_modules_map.get(&tree_shake_module_id) {
+  //     let contains_self_executed_stmt = tree_shake_module.contains_self_executed_stmt;
+  //     module_graph
+  //       .dependents_ids(&tree_shake_module_id)
+  //       .into_iter()
+  //       .for_each(|dept_id| {
+  //         if let Some(dept_tree_shake_module) = tree_shake_modules_map.get_mut(&dept_id) {
+  //           dept_tree_shake_module.contains_self_executed_stmt =
+  //             dept_tree_shake_module.contains_self_executed_stmt || contains_self_executed_stmt
+  //         }
+  //       });
+  //   }
+  // }
 
   entry_module_ids
 }

crates/plugin_tree_shake/src/module.rs

@@ -5,7 +5,7 @@ use std::{
 
 use farmfe_core::{
   module::{Module, ModuleId, ModuleSystem},
-  swc_common::Mark,
+  swc_common::{comments::SingleThreadedComments, Mark},
 };
 
 use crate::statement_graph::{
@@ -67,6 +67,13 @@ impl Default for UsedExports {
 }
 
 impl UsedExports {
+  pub fn as_partial(&self) -> &HashSet<UsedExportsIdent> {
+    match self {
+      UsedExports::All => panic!("UsedExports is not Partial"),
+      UsedExports::Partial(res) => res,
+    }
+  }
+
   pub fn add_used_export(&mut self, used_export: UsedExportsIdent) {
     // All means all exports are used, only handle Partial here
     if let UsedExports::Partial(used_exports) = self {
@@ -122,23 +129,25 @@ pub struct TreeShakeModule {
 }
 
 impl TreeShakeModule {
-  pub fn new(module: &Module) -> Self {
+  pub fn new(module: &mut Module) -> Self {
     farmfe_core::farm_profile_function!(format!(
       "TreeShakeModule::new {:?}",
       module.id.to_string()
     ));
     let module_system = module.meta.as_script().module_system.clone();
 
     // 1. generate statement graph
+    let comments_meta = module.meta.as_script_mut().take_comments();
     let ast = &module.meta.as_script().ast;
     let unresolved_mark = Mark::from_u32(module.meta.as_script().unresolved_mark);
     let top_level_mark = Mark::from_u32(module.meta.as_script().top_level_mark);
-
+    let comments = SingleThreadedComments::from(comments_meta);
     let stmt_graph = if module_system == ModuleSystem::EsModule {
-      StatementGraph::new(ast, unresolved_mark, top_level_mark)
+      StatementGraph::new(ast, unresolved_mark, top_level_mark, &comments)
     } else {
       StatementGraph::empty()
     };
+    module.meta.as_script_mut().set_comments(comments.into());
 
     // 2. set default used exports
     let handled_used_exports = if module.side_effects {
@@ -150,6 +159,8 @@ impl TreeShakeModule {
     Self {
       module_id: module.id.clone(),
       contains_self_executed_stmt: module.side_effects
+        || !matches!(module_system, ModuleSystem::EsModule)
+        || stmt_graph.contains_bare_import_stmt()
         || !stmt_graph.preserved_side_effects_stmts().is_empty(),
       stmt_graph,
       pending_used_exports: handled_used_exports.clone(),
@@ -355,7 +366,10 @@ impl TreeShakeModule {
               // skip default for export * from 'xxx'
               if ident != *"default" {
                 for export_all_stmt_id in export_all_stmt_id {
-                  used_idents.push((UsedStatementIdent::InExportAll(ident), export_all_stmt_id));
+                  used_idents.push((
+                    UsedStatementIdent::InExportAll(ident.clone()),
+                    export_all_stmt_id,
+                  ));
                 }
               }
             }

crates/plugin_tree_shake/src/statement_graph.rs

@@ -2,6 +2,7 @@ use std::collections::VecDeque;
 use std::collections::{HashMap, HashSet};
 
 use farmfe_core::petgraph::Direction;
+use farmfe_core::swc_common::comments::SingleThreadedComments;
 use farmfe_core::swc_common::Mark;
 use farmfe_core::swc_ecma_ast::{Id, ImportSpecifier, ModuleDecl, ModuleExportName};
 use farmfe_core::{
@@ -93,13 +94,13 @@ pub enum StatementSideEffects {
   /// ```
   WriteTopLevelVar(HashSet<Id>),
 
-  /// Access global variable, it's always preserved, for example:
+  /// Maybe modify global variable, it's always preserved, for example:
   /// ```js
   /// console.log('123');
   /// window.b = 3;
   /// document.body.addEventListener('click', () =/*  */> {});
   /// ```
-  AccessGlobalVar, // TODO investigate how top level mark works
+  WriteOrCallGlobalVar,
 
   /// Unclassified default self executed statements are always treated as side effects. For example:
   /// ```js
@@ -114,9 +115,6 @@ pub enum StatementSideEffects {
   ///  console.log('123');
   /// }
   /// foo();
-  /// for (let i = 0; i < 10; i++) {
-  ///  window['a' + i] = i;
-  /// }
   /// ```
   /// They may be classified in the future to improve the accuracy of tree shaking
   UnclassifiedSelfExecuted,
@@ -130,7 +128,10 @@ pub enum StatementSideEffects {
 
 impl StatementSideEffects {
   pub fn is_preserved(&self) -> bool {
-    matches!(self, Self::AccessGlobalVar | Self::UnclassifiedSelfExecuted)
+    matches!(
+      self,
+      Self::WriteOrCallGlobalVar | Self::UnclassifiedSelfExecuted
+    )
   }
 
   pub fn merge_side_effects(&mut self, other: Self) {
@@ -141,8 +142,8 @@ impl StatementSideEffects {
         a.extend(b.iter().cloned());
       }
       (Self::NoSideEffects, _) => original_self_value = other,
-      (Self::AccessGlobalVar | Self::UnclassifiedSelfExecuted, _) => {}
-      (Self::WriteTopLevelVar(_), Self::AccessGlobalVar | Self::UnclassifiedSelfExecuted) => {
+      (Self::WriteOrCallGlobalVar | Self::UnclassifiedSelfExecuted, _) => {}
+      (Self::WriteTopLevelVar(_), Self::WriteOrCallGlobalVar | Self::UnclassifiedSelfExecuted) => {
         original_self_value = other;
       }
       (_, Self::NoSideEffects) => {}
@@ -204,6 +205,7 @@ impl Statement {
     stmt: &ModuleItem,
     unresolved_mark: Mark,
     top_level_mark: Mark,
+    comments: &SingleThreadedComments,
   ) -> Self {
     // 1. analyze all import, export and defined idents of the ModuleItem
     let AnalyzedStatementInfo {
@@ -217,6 +219,7 @@ impl Statement {
       stmt,
       unresolved_mark,
       top_level_mark,
+      comments,
     );
 
     Self {
@@ -265,14 +268,19 @@ pub struct StatementGraph {
 }
 
 impl StatementGraph {
-  pub fn new(module: &SwcModule, unresolved_mark: Mark, top_level_mark: Mark) -> Self {
+  pub fn new(
+    module: &SwcModule,
+    unresolved_mark: Mark,
+    top_level_mark: Mark,
+    comments: &SingleThreadedComments,
+  ) -> Self {
     let mut g = petgraph::graph::Graph::new();
     let mut id_index_map = HashMap::new();
 
     let mut reverse_defined_idents_map = HashMap::new();
     // 1. analyze all defined idents of each statement
     for (index, item) in module.body.iter().enumerate() {
-      let stmt = Statement::new(index, item, unresolved_mark, top_level_mark);
+      let stmt = Statement::new(index, item, unresolved_mark, top_level_mark, comments);
 
       // export named does not define any idents
       if !matches!(item, ModuleItem::ModuleDecl(ModuleDecl::ExportNamed(_))) {
@@ -330,6 +338,24 @@ impl StatementGraph {
       .collect()
   }
 
+  pub fn contains_bare_import_stmt(&self) -> bool {
+    self
+      .stmt_ids()
+      .into_iter()
+      .any(|stmt_id| self.is_bare_import_stmt(stmt_id))
+  }
+
+  /// true if stmt is import './xxx'. (without specifiers)
+  pub fn is_bare_import_stmt(&self, stmt_id: StatementId) -> bool {
+    let stmt = self.stmt(&stmt_id);
+
+    if let Some(import_info) = &stmt.import_info {
+      return import_info.specifiers.is_empty();
+    }
+
+    false
+  }
+
   pub fn add_edge(&mut self, from: StatementId, to: StatementId, edge_weight: StatementGraphEdge) {
     let from_node = self.id_index_map.get(&from).unwrap();
     let to_node = self.id_index_map.get(&to).unwrap();
@@ -599,7 +625,7 @@ impl StatementGraph {
               all_dept_used_idents.extend(dept_used_idents);
             }
           }
-          StatementSideEffects::AccessGlobalVar
+          StatementSideEffects::WriteOrCallGlobalVar
           | StatementSideEffects::UnclassifiedSelfExecuted
           | StatementSideEffects::NoSideEffects => { /* These 3 types are handled already */ }
         }