From 05f592b834e32870711566815996dd0ece5f2882 Mon Sep 17 00:00:00 2001
From: camchenry <1514176+camchenry@users.noreply.github.com>
Date: Sat, 21 Sep 2024 18:15:07 +0000
Subject: [PATCH] refactor(linter): Use parsed patterns in
 `unicorn/prefer-string-starts-ends-with` (#5949)

- part of https://github.com/oxc-project/oxc/issues/5416

This change enhances the accuracy of the `prefer_string_starts_ends_with` rule by using the parsed regex patterns for analysis. It allows for more precise detection of patterns that can be replaced with `startsWith()` and `endsWith()` methods, reducing false positives and improving the overall effectiveness of the linter.

### What changed?

- Replaced the simple string-based regex analysis with a more robust AST-based approach.
- Removed the `is_simple_string` function as it's no longer needed.
---
 .../unicorn/prefer_string_starts_ends_with.rs | 36 ++++++++++++-------
 1 file changed, 23 insertions(+), 13 deletions(-)

diff --git a/crates/oxc_linter/src/rules/unicorn/prefer_string_starts_ends_with.rs b/crates/oxc_linter/src/rules/unicorn/prefer_string_starts_ends_with.rs
index 0986578567417..359ed4cbc1167 100644
--- a/crates/oxc_linter/src/rules/unicorn/prefer_string_starts_ends_with.rs
+++ b/crates/oxc_linter/src/rules/unicorn/prefer_string_starts_ends_with.rs
@@ -4,6 +4,7 @@ use oxc_ast::{
 };
 use oxc_diagnostics::OxcDiagnostic;
 use oxc_macros::declare_oxc_lint;
+use oxc_regular_expression::ast::{BoundaryAssertionKind, Term};
 use oxc_span::{GetSpan, Span};
 
 use crate::{
@@ -146,24 +147,33 @@ fn check_regex(regexp_lit: &RegExpLiteral, pattern_text: &str) -> Option<ErrorKi
         return None;
     }
 
-    if pattern_text.starts_with('^')
-        && is_simple_string(&pattern_text[1..regexp_lit.regex.pattern.len()])
-    {
-        return Some(ErrorKind::StartsWith);
+    let alternatives = regexp_lit.regex.pattern.as_pattern().map(|pattern| &pattern.body.body)?;
+    // Must not be something with multiple alternatives like `/^a|b/`
+    if alternatives.len() > 1 {
+        return None;
     }
+    let pattern_terms = alternatives.first().map(|it| &it.body)?;
 
-    if pattern_text.ends_with('$')
-        && is_simple_string(&pattern_text[0..regexp_lit.regex.pattern.len() - 1])
-    {
-        return Some(ErrorKind::EndsWith);
+    if let Some(Term::BoundaryAssertion(boundary_assert)) = pattern_terms.first() {
+        if boundary_assert.kind == BoundaryAssertionKind::Start
+            && pattern_terms.iter().skip(1).all(|term| matches!(term, Term::Character(_)))
+        {
+            return Some(ErrorKind::StartsWith);
+        }
     }
 
-    None
-}
+    if let Some(Term::BoundaryAssertion(boundary_assert)) = pattern_terms.last() {
+        if boundary_assert.kind == BoundaryAssertionKind::End
+            && pattern_terms
+                .iter()
+                .take(pattern_terms.len() - 1)
+                .all(|term| matches!(term, Term::Character(_)))
+        {
+            return Some(ErrorKind::EndsWith);
+        }
+    }
 
-fn is_simple_string(str: &str) -> bool {
-    str.chars()
-        .all(|c| !matches!(c, '^' | '$' | '+' | '[' | '{' | '(' | '\\' | '.' | '?' | '*' | '|'))
+    None
 }
 
 // `/^#/i` => `true` (the `i` flag is useless)