fix(ast, codegen, linter): panics in fixers. (#5431)

Closes #5434

https://github.com/oxc-project/oxlint-ecosystem-ci/actions/runs/10685877915/job/29619941099

crates/oxc_ast/src/ast_impl/literal.rs

@@ -6,6 +6,7 @@
 use crate::ast::*;
 
 use std::{
+    borrow::Cow,
     fmt,
     hash::{Hash, Hasher},
 };
@@ -132,10 +133,11 @@ impl<'a> RegExpPattern<'a> {
         self.len() == 0
     }
 
-    pub fn source_text(&self, source_text: &'a str) -> &'a str {
+    pub fn source_text(&self, source_text: &'a str) -> Cow<str> {
         match self {
-            Self::Raw(raw) | Self::Invalid(raw) => raw,
-            Self::Pattern(pat) => pat.span.source_text(source_text),
+            Self::Raw(raw) | Self::Invalid(raw) => Cow::Borrowed(raw),
+            Self::Pattern(pat) if pat.span.is_unspanned() => Cow::Owned(pat.to_string()),
+            Self::Pattern(pat) => Cow::Borrowed(pat.span.source_text(source_text)),
         }
     }
 

crates/oxc_codegen/src/annotation_comment.rs

@@ -54,6 +54,7 @@ impl<'a> Codegen<'a> {
             return vec![];
         }
         let mut latest_comment_start = node_start;
+        let source_text = self.source_text.unwrap_or_default();
         let mut ret = self
             .get_leading_comments(self.latest_consumed_comment_end, node_start)
             .rev()
@@ -61,15 +62,14 @@ impl<'a> Codegen<'a> {
             .take_while(|comment| {
                 let comment_end = comment.real_span_end();
                 let range_content =
-                    &self.source_text[comment_end as usize..latest_comment_start as usize];
+                    &source_text[comment_end as usize..latest_comment_start as usize];
                 let all_whitespace = range_content.chars().all(char::is_whitespace);
                 latest_comment_start = comment.real_span_start();
                 all_whitespace
             })
             .filter_map(|comment| {
-                let source_code = self.source_text;
                 let comment_content =
-                    &source_code[comment.span.start as usize..comment.span.end as usize];
+                    &source_text[comment.span.start as usize..comment.span.end as usize];
                 if let Some(m) = MATCHER.find_iter(&comment_content).next() {
                     let annotation_kind = match m.value() {
                         0 | 1 => AnnotationKind::NO_SIDE_EFFECTS,

crates/oxc_codegen/src/gen.rs

@@ -1,4 +1,4 @@
-use std::ops::Not;
+use std::{borrow::Cow, ops::Not};
 
 use oxc_allocator::{Box, Vec};
 #[allow(clippy::wildcard_imports)]
@@ -1234,7 +1234,10 @@ impl<'a> Gen for RegExpLiteral<'a> {
     fn gen(&self, p: &mut Codegen, _ctx: Context) {
         p.add_source_mapping(self.span.start);
         let last = p.peek_nth(0);
-        let pattern_text = self.regex.pattern.source_text(p.source_text);
+        let pattern_text = p.source_text.map_or_else(
+            || Cow::Owned(self.regex.pattern.to_string()),
+            |src| self.regex.pattern.source_text(src),
+        );
         // Avoid forming a single-line comment or "</script" sequence
         if Some('/') == last
             || (Some('<') == last && pattern_text.to_lowercase().starts_with("script"))
@@ -1242,7 +1245,7 @@ impl<'a> Gen for RegExpLiteral<'a> {
             p.print_hard_space();
         }
         p.print_char(b'/');
-        p.print_str(pattern_text);
+        p.print_str(pattern_text.as_ref());
         p.print_char(b'/');
         p.print_str(self.regex.flags.to_string().as_str());
         p.prev_reg_exp_end = p.code().len();

crates/oxc_codegen/src/lib.rs

@@ -63,7 +63,8 @@ pub struct Codegen<'a> {
     options: CodegenOptions,
     comment_options: CommentOptions,
 
-    source_text: &'a str,
+    /// Original source code of the AST
+    source_text: Option<&'a str>,
 
     trivias: Trivias,
 
@@ -131,7 +132,7 @@ impl<'a> Codegen<'a> {
         Self {
             options: CodegenOptions::default(),
             comment_options: CommentOptions::default(),
-            source_text: "",
+            source_text: None,
             trivias: Trivias::default(),
             mangler: None,
             code: vec![],
@@ -169,6 +170,15 @@ impl<'a> Codegen<'a> {
         self
     }
 
+    /// Adds the source text of the original AST, It is used with comments or for improving the
+    /// generated output.
+    #[must_use]
+    pub fn with_source_text(mut self, source_text: &'a str) -> Self {
+        self.source_text = Some(source_text);
+        self
+    }
+
+    /// Also sets the [Self::with_source_text]
     #[must_use]
     pub fn enable_comment(
         mut self,
@@ -176,10 +186,9 @@ impl<'a> Codegen<'a> {
         trivias: Trivias,
         options: CommentOptions,
     ) -> Self {
-        self.source_text = source_text;
         self.trivias = trivias;
         self.comment_options = options;
-        self
+        self.with_source_text(source_text)
     }
 
     #[must_use]
@@ -539,8 +548,12 @@ impl<'a> Codegen<'a> {
     /// Avoid issue related to rustc borrow checker .
     /// Since if you want to print a range of source code, you need to borrow the source code
     /// as immutable first, and call the [Self::print_str] which is a mutable borrow.
+    ///
+    /// # Panics
+    /// If `self.source_text` isn't set.
     fn print_range_of_source_code(&mut self, range: Range<usize>) {
-        self.code.extend_from_slice(self.source_text[range].as_bytes());
+        let source_text = self.source_text.expect("expect `Codegen::source_text` to be set.");
+        self.code.extend_from_slice(source_text[range].as_bytes());
     }
 
     fn get_leading_comments(

crates/oxc_codegen/tests/integration/tester.rs

@@ -21,6 +21,17 @@ pub fn test(source_text: &str, expected: &str) {
     );
 }
 
+pub fn test_without_source(source_text: &str, expected: &str) {
+    let source_type = SourceType::default().with_module(true).with_jsx(true);
+    let allocator = Allocator::default();
+    let ret = Parser::new(&allocator, source_text, source_type).parse();
+    let result = CodeGenerator::new().build(&ret.program).source_text;
+    assert_eq!(
+        result, expected,
+        "\nfor source {source_text:?}\nexpect {expected:?}\ngot    {result:?}\nwithout providing the original code."
+    );
+}
+
 pub fn test_minify(source_text: &str, expected: &str) {
     let source_type = SourceType::default().with_module(true).with_jsx(true);
     let allocator = Allocator::default();

crates/oxc_codegen/tests/integration/unit.rs

@@ -1,4 +1,4 @@
-use crate::tester::{test, test_minify};
+use crate::tester::{test, test_minify, test_without_source};
 
 #[test]
 fn module_decl() {
@@ -63,6 +63,33 @@ fn unicode_escape() {
     test("console.log('🧑‍🤝‍🧑');", "console.log(\"🧑‍🤝‍🧑\");\n");
 }
 
+#[test]
+fn regex() {
+    fn test_all(source: &str, expect: &str, minify: &str) {
+        test(source, expect);
+        test_minify(source, minify);
+        test_without_source(source, expect);
+    }
+    test_all("/regex/giv", "/regex/giv;\n", "/regex/giv;");
+    test_all(
+        r"/(.)(.)(.)(.)(.)(.)(.)(.)\8\8/",
+        "/(.)(.)(.)(.)(.)(.)(.)(.)\\8\\8/;\n",
+        "/(.)(.)(.)(.)(.)(.)(.)(.)\\8\\8/;",
+    );
+
+    test_all(
+        r"/\n\cM\0\x41\u{1f600}\./u",
+        "/\\n\\cM\\0\\x41\\u{1f600}\\./u;\n",
+        "/\\n\\cM\\0\\x41\\u{1f600}\\./u;",
+    );
+    test_all(r"/\n\cM\0\x41\./u", "/\\n\\cM\\0\\x41\\./u;\n", "/\\n\\cM\\0\\x41\\./u;");
+    test_all(
+        r"/\n\cM\0\x41\u1234\./",
+        "/\\n\\cM\\0\\x41\\u1234\\./;\n",
+        "/\\n\\cM\\0\\x41\\u1234\\./;",
+    );
+}
+
 #[test]
 fn comma() {
     test_minify("1, 2, 3", "1,2,3;");

crates/oxc_linter/src/fixer/mod.rs

@@ -171,6 +171,7 @@ impl<'c, 'a: 'c> RuleFixer<'c, 'a> {
     #[allow(clippy::unused_self)]
     pub fn codegen(self) -> CodeGenerator<'a> {
         CodeGenerator::new()
+            .with_source_text(self.source_text())
             .with_options(CodegenOptions { single_quote: true, ..CodegenOptions::default() })
     }
 

crates/oxc_linter/src/rules/eslint/no_control_regex.rs

@@ -68,7 +68,7 @@ impl Rule for NoControlRegex {
             let mut violations: Vec<&str> = Vec::new();
             let pattern = pattern.as_ref();
             let pattern_text = pattern.source_text(context.source_text());
-            for matched_ctl_pattern in control_patterns(pattern_text) {
+            for matched_ctl_pattern in control_patterns(pattern_text.as_ref()) {
                 let ctl = matched_ctl_pattern.as_str();
 
                 // check for an even number of backslashes, since these will

crates/oxc_linter/src/rules/eslint/no_empty_character_class.rs

@@ -50,7 +50,7 @@ impl Rule for NoEmptyCharacterClass {
 
         if let AstKind::RegExpLiteral(lit) = node.kind() {
             if !NO_EMPTY_CLASS_REGEX_PATTERN
-                .is_match(lit.regex.pattern.source_text(ctx.source_text()))
+                .is_match(lit.regex.pattern.source_text(ctx.source_text()).as_ref())
             {
                 ctx.diagnostic(no_empty_character_class_diagnostic(lit.span));
             }

crates/oxc_linter/src/rules/eslint/no_regex_spaces.rs

@@ -69,6 +69,7 @@ impl Rule for NoRegexSpaces {
 impl NoRegexSpaces {
     fn find_literal_to_report(literal: &RegExpLiteral, ctx: &LintContext) -> Option<Span> {
         let pattern_text = literal.regex.pattern.source_text(ctx.source_text());
+        let pattern_text = pattern_text.as_ref();
         if Self::has_exempted_char_class(pattern_text) {
             return None;
         }

crates/oxc_linter/src/rules/unicorn/no_hex_escape.rs

@@ -86,7 +86,7 @@ impl Rule for NoHexEscape {
             }
             AstKind::RegExpLiteral(regex) => {
                 if let Some(fixed) =
-                    check_escape(regex.regex.pattern.source_text(ctx.source_text()))
+                    check_escape(regex.regex.pattern.source_text(ctx.source_text()).as_ref())
                 {
                     #[allow(clippy::cast_possible_truncation)]
                     ctx.diagnostic_with_fix(no_hex_escape_diagnostic(regex.span), |fixer| {

crates/oxc_linter/src/rules/unicorn/prefer_string_replace_all.rs

@@ -127,6 +127,7 @@ fn get_pattern_replacement<'a>(
     }
 
     let pattern_text = reg_exp_literal.regex.pattern.source_text(ctx.source_text());
+    let pattern_text = pattern_text.as_ref();
     if !is_simple_string(pattern_text) {
         return None;
     }

crates/oxc_linter/src/rules/unicorn/prefer_string_starts_ends_with.rs

@@ -75,6 +75,7 @@ impl Rule for PreferStringStartsEndsWith {
         };
 
         let pattern_text = regex.regex.pattern.source_text(ctx.source_text());
+        let pattern_text = pattern_text.as_ref();
 
         let Some(err_kind) = check_regex(regex, pattern_text) else {
             return;

crates/oxc_prettier/src/format/mod.rs

@@ -1416,7 +1416,7 @@ impl<'a> Format<'a> for RegExpLiteral<'a> {
     fn format(&self, p: &mut Prettier<'a>) -> Doc<'a> {
         let mut parts = p.vec();
         parts.push(ss!("/"));
-        parts.push(p.str(self.regex.pattern.source_text(p.source_text)));
+        parts.push(p.str(self.regex.pattern.source_text(p.source_text).as_ref()));
         parts.push(ss!("/"));
         parts.push(format!(p, self.regex.flags));
         Doc::Array(parts)