feat(parser)!: Build ModuleRecord directly in parser (#7546)

This has the benefit of:

* expose dynamic import / import meta info from parser
* 1 less ast shallow in semantic builder
* no ast walk in oxc's module lexer
* some more benefits coming soon

crates/oxc/src/compiler.rs

@@ -143,7 +143,7 @@ pub trait CompilerInterface {
 
         /* Semantic */
 
-        let mut semantic_return = self.semantic(&program, source_path);
+        let mut semantic_return = self.semantic(&program);
         if !semantic_return.errors.is_empty() {
             self.handle_errors(semantic_return.errors);
             return;
@@ -231,7 +231,7 @@ pub trait CompilerInterface {
         Parser::new(allocator, source_text, source_type).with_options(self.parse_options()).parse()
     }
 
-    fn semantic<'a>(&self, program: &Program<'a>, source_path: &Path) -> SemanticBuilderReturn<'a> {
+    fn semantic<'a>(&self, program: &Program<'a>) -> SemanticBuilderReturn<'a> {
         let mut builder = SemanticBuilder::new();
 
         if self.transform_options().is_some() {
@@ -242,7 +242,6 @@ pub trait CompilerInterface {
         builder
             .with_check_syntax_error(self.check_semantic_error())
             .with_scope_tree_child_ids(self.semantic_child_scope_ids())
-            .build_module_record(source_path, program)
             .build(program)
     }
 

crates/oxc_linter/src/frameworks.rs

@@ -1,7 +1,7 @@
 use std::{hash, path::Path};
 
 use bitflags::bitflags;
-use oxc_semantic::ModuleRecord;
+use oxc_syntax::module_record::ModuleRecord;
 
 bitflags! {
     #[derive(Debug, Clone, Copy, PartialEq, Eq)]

crates/oxc_linter/src/rules/import/export.rs

@@ -1,10 +1,11 @@
 use std::path::PathBuf;
 
+use rustc_hash::{FxHashMap, FxHashSet};
+
 use oxc_diagnostics::{LabeledSpan, OxcDiagnostic};
 use oxc_macros::declare_oxc_lint;
-use oxc_semantic::ModuleRecord;
 use oxc_span::{CompactStr, Span};
-use rustc_hash::{FxHashMap, FxHashSet};
+use oxc_syntax::module_record::ModuleRecord;
 
 use crate::{context::LintContext, rule::Rule};
 

crates/oxc_linter/src/rules/import/namespace.rs

@@ -6,9 +6,11 @@ use oxc_ast::{
 };
 use oxc_diagnostics::OxcDiagnostic;
 use oxc_macros::declare_oxc_lint;
-use oxc_semantic::{AstNode, ModuleRecord};
+use oxc_semantic::AstNode;
 use oxc_span::{GetSpan, Span};
-use oxc_syntax::module_record::{ExportExportName, ExportImportName, ImportImportName};
+use oxc_syntax::module_record::{
+    ExportExportName, ExportImportName, ImportImportName, ModuleRecord,
+};
 
 use crate::{context::LintContext, rule::Rule};
 

crates/oxc_linter/src/rules/oxc/no_barrel_file.rs

@@ -1,7 +1,9 @@
 use oxc_diagnostics::{LabeledSpan, OxcDiagnostic};
 use oxc_macros::declare_oxc_lint;
-use oxc_semantic::ModuleRecord;
-use oxc_syntax::module_graph_visitor::{ModuleGraphVisitorBuilder, VisitFoldWhile};
+use oxc_syntax::{
+    module_graph_visitor::{ModuleGraphVisitorBuilder, VisitFoldWhile},
+    module_record::ModuleRecord,
+};
 
 use crate::{context::LintContext, rule::Rule};
 

crates/oxc_linter/src/service/module_cache.rs

@@ -1,12 +1,13 @@
 use std::{
+    num::NonZeroUsize,
     path::Path,
     sync::{Arc, Condvar, Mutex},
 };
 
 use dashmap::{mapref::one::Ref, DashMap};
-use oxc_semantic::ModuleRecord;
 use rustc_hash::{FxBuildHasher, FxHashMap};
-use std::num::NonZeroUsize;
+
+use oxc_syntax::module_record::ModuleRecord;
 
 type FxDashMap<K, V> = DashMap<K, V, FxBuildHasher>;
 

crates/oxc_linter/src/service/runtime.rs

@@ -11,8 +11,9 @@ use oxc_allocator::Allocator;
 use oxc_diagnostics::{DiagnosticSender, DiagnosticService, Error, OxcDiagnostic};
 use oxc_parser::{ParseOptions, Parser};
 use oxc_resolver::Resolver;
-use oxc_semantic::{ModuleRecord, SemanticBuilder};
+use oxc_semantic::SemanticBuilder;
 use oxc_span::{SourceType, VALID_EXTENSIONS};
+use oxc_syntax::module_record::ModuleRecord;
 use rayon::{iter::ParallelBridge, prelude::ParallelIterator};
 use rustc_hash::FxHashSet;
 
@@ -215,9 +216,11 @@ impl Runtime {
             .with_cfg(true)
             .with_scope_tree_child_ids(true)
             .with_build_jsdoc(true)
-            .with_check_syntax_error(check_syntax_errors)
-            .build_module_record(path, &ret.program);
-        let module_record = semantic_builder.module_record();
+            .with_check_syntax_error(check_syntax_errors);
+
+        let mut module_record = ret.module_record;
+        module_record.resolved_absolute_path = path.to_path_buf();
+        let module_record = Arc::new(module_record);
 
         if self.resolver.is_some() {
             self.modules.add_resolved_module(path, Arc::clone(&module_record));
@@ -294,6 +297,7 @@ impl Runtime {
         };
 
         let mut semantic = semantic_ret.semantic;
+        semantic.set_module_record(&module_record);
         semantic.set_irregular_whitespaces(ret.irregular_whitespaces);
         self.linter.run(path, Rc::new(semantic))
     }

crates/oxc_module_lexer/tests/integration/main.rs

@@ -71,7 +71,6 @@ pub fn parse(source: &str) -> ModuleLexer {
     let allocator = Allocator::default();
     let source_type = SourceType::mjs();
     let ret = Parser::new(&allocator, source, source_type).parse();
-    assert!(ret.errors.is_empty(), "{source} should not produce errors.\n{:?}", ret.errors);
     let module_lexer = oxc_module_lexer::ModuleLexer::new().build(&ret.program);
     // Copy data over because `ModuleLexer<'a>` can't be returned
     ModuleLexer {

crates/oxc_parser/src/diagnostics.rs

@@ -448,6 +448,14 @@ pub fn cover_initialized_name(span: Span) -> OxcDiagnostic {
 .with_label(span)
 }
 
+#[cold]
+pub fn duplicate_export(x0: &str, span1: Span, span2: Span) -> OxcDiagnostic {
+    OxcDiagnostic::error(format!("Duplicated export '{x0}'")).with_labels([
+        span1.label("Export has already been declared here"),
+        span2.label("It cannot be redeclared here"),
+    ])
+}
+
 // ================================= MODIFIERS =================================
 
 #[cold]

crates/oxc_parser/src/js/statement.rs

@@ -42,8 +42,11 @@ impl<'a> ParserImpl<'a> {
             }
             let stmt = self.parse_statement_list_item(StatementContext::StatementList)?;
 
-            if is_top_level && stmt.is_module_declaration() {
-                self.set_source_type_to_module_if_unambiguous();
+            if is_top_level {
+                if let Some(module_decl) = stmt.as_module_declaration() {
+                    self.set_source_type_to_module_if_unambiguous();
+                    self.module_record_builder.visit_module_declaration(module_decl);
+                }
             }
 
             // Section 11.2.1 Directive Prologue

crates/oxc_parser/src/lib.rs

@@ -69,6 +69,7 @@
 mod context;
 mod cursor;
 mod modifiers;
+mod module_record;
 mod state;
 
 mod js;
@@ -84,17 +85,19 @@ mod lexer;
 #[doc(hidden)]
 pub mod lexer;
 
-use context::{Context, StatementContext};
 use oxc_allocator::{Allocator, Box as ArenaBox};
 use oxc_ast::{
     ast::{Expression, Program},
     AstBuilder,
 };
 use oxc_diagnostics::{OxcDiagnostic, Result};
 use oxc_span::{ModuleKind, SourceType, Span};
+use oxc_syntax::module_record::ModuleRecord;
 
 use crate::{
+    context::{Context, StatementContext},
     lexer::{Kind, Lexer, Token},
+    module_record::ModuleRecordBuilder,
     state::ParserState,
 };
 
@@ -150,6 +153,9 @@ pub struct ParserReturn<'a> {
     /// semantic analysis with syntax error checking enabled.
     pub program: Program<'a>,
 
+    /// See <https://tc39.es/ecma262/#sec-abstract-module-records>
+    pub module_record: ModuleRecord,
+
     /// Syntax errors encountered while parsing.
     ///
     /// This list is not comprehensive. Oxc offloads more-expensive checks to [semantic
@@ -363,6 +369,9 @@ struct ParserImpl<'a> {
     /// Ast builder for creating AST nodes
     ast: AstBuilder<'a>,
 
+    /// Module Record Builder
+    module_record_builder: ModuleRecordBuilder,
+
     /// Precomputed typescript detection
     is_ts: bool,
 }
@@ -380,6 +389,8 @@ impl<'a> ParserImpl<'a> {
         options: ParseOptions,
         unique: UniquePromise,
     ) -> Self {
+        let mut module_record_builder = ModuleRecordBuilder::default();
+        module_record_builder.module_record.not_esm = true;
         Self {
             options,
             lexer: Lexer::new(allocator, source_text, source_type, unique),
@@ -391,6 +402,7 @@ impl<'a> ParserImpl<'a> {
             state: ParserState::default(),
             ctx: Self::default_context(source_type, options),
             ast: AstBuilder::new(allocator),
+            module_record_builder,
             is_ts: source_type.is_typescript(),
         }
     }
@@ -432,10 +444,21 @@ impl<'a> ParserImpl<'a> {
             errors.reserve(self.lexer.errors.len() + self.errors.len());
             errors.extend(self.lexer.errors);
             errors.extend(self.errors);
+            // Skip checking for exports in TypeScript {
+            if !self.source_type.is_typescript() {
+                errors.extend(self.module_record_builder.errors());
+            }
         }
         let irregular_whitespaces =
             self.lexer.trivia_builder.irregular_whitespaces.into_boxed_slice();
-        ParserReturn { program, errors, irregular_whitespaces, panicked, is_flow_language }
+        ParserReturn {
+            program,
+            module_record: self.module_record_builder.build(),
+            errors,
+            irregular_whitespaces,
+            panicked,
+            is_flow_language,
+        }
     }
 
     pub fn parse_expression(mut self) -> std::result::Result<Expression<'a>, Vec<OxcDiagnostic>> {