From d1d0607158ab40463d1b123fed52cc526eba8385 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Donny/=EA=B0=95=EB=8F=99=EC=9C=A4?= <kdy1997.dev@gmail.com>
Date: Sat, 11 Mar 2023 12:43:06 +0900
Subject: [PATCH] perf(es/lexer): Make lexer faster (#7052)

---
 crates/swc_common/src/input.rs            | 49 +++++++++++++----------
 crates/swc_ecma_parser/src/lexer/jsx.rs   |  2 +-
 crates/swc_ecma_parser/src/lexer/mod.rs   | 15 ++++---
 crates/swc_ecma_parser/src/lexer/state.rs |  4 +-
 crates/swc_ecma_parser/src/lexer/util.rs  |  6 +--
 5 files changed, 44 insertions(+), 32 deletions(-)

diff --git a/crates/swc_common/src/input.rs b/crates/swc_common/src/input.rs
index 47949744728..246a98b1d94 100644
--- a/crates/swc_common/src/input.rs
+++ b/crates/swc_common/src/input.rs
@@ -77,20 +77,6 @@ impl<'a> Input for StringInput<'a> {
         self.iter.clone().nth(2).map(|i| i.1)
     }
 
-    #[inline]
-    fn is_str(&self, s: &str) -> bool {
-        let mut s_iter = s.as_bytes().iter();
-        let mut p_iter = self.iter.clone().map(|i| i.1);
-
-        while let (Some(expected), Some(actual)) = (s_iter.next(), p_iter.next()) {
-            if *expected as char != actual {
-                return false;
-            }
-        }
-
-        s_iter.next().is_none()
-    }
-
     #[inline]
     fn bump(&mut self) {
         if let Some((i, c)) = self.iter.next() {
@@ -102,6 +88,16 @@ impl<'a> Input for StringInput<'a> {
         }
     }
 
+    #[inline]
+    fn cur_as_ascii(&mut self) -> Option<u8> {
+        let first_byte = *self.as_str().as_bytes().first()?;
+        if first_byte <= 0x7f {
+            Some(first_byte)
+        } else {
+            None
+        }
+    }
+
     #[inline]
     fn is_at_start(&self) -> bool {
         self.orig_start == self.last_pos
@@ -120,15 +116,17 @@ impl<'a> Input for StringInput<'a> {
 
     #[inline]
     fn slice(&mut self, start: BytePos, end: BytePos) -> &str {
-        assert!(start <= end, "Cannot slice {:?}..{:?}", start, end);
+        debug_assert!(start <= end, "Cannot slice {:?}..{:?}", start, end);
         let s = self.orig;
 
         let start_idx = (start - self.orig_start).0 as usize;
         let end_idx = (end - self.orig_start).0 as usize;
 
-        let ret = &s[start_idx..end_idx];
+        debug_assert!(end_idx <= s.len());
 
-        self.iter = s[end_idx..].char_indices();
+        let ret = unsafe { s.get_unchecked(start_idx..end_idx) };
+
+        self.iter = unsafe { s.get_unchecked(end_idx..) }.char_indices();
         self.last_pos = end;
         self.start_pos_of_iter = end;
 
@@ -149,11 +147,12 @@ impl<'a> Input for StringInput<'a> {
                 break;
             }
         }
-        let ret = &s[..last];
+        debug_assert!(last <= s.len());
+        let ret = unsafe { s.get_unchecked(..last) };
 
         self.last_pos = self.last_pos + BytePos(last as _);
         self.start_pos_of_iter = self.last_pos;
-        self.iter = s[last..].char_indices();
+        self.iter = unsafe { s.get_unchecked(last..) }.char_indices();
 
         ret
     }
@@ -175,9 +174,11 @@ impl<'a> Input for StringInput<'a> {
             return None;
         }
 
+        debug_assert!(last <= s.len());
+
         self.last_pos = self.last_pos + BytePos(last as _);
         self.start_pos_of_iter = self.last_pos;
-        self.iter = s[last..].char_indices();
+        self.iter = unsafe { s.get_unchecked(last..) }.char_indices();
 
         Some(self.last_pos)
     }
@@ -187,7 +188,8 @@ impl<'a> Input for StringInput<'a> {
         let orig = self.orig;
         let idx = (to - self.orig_start).0 as usize;
 
-        let s = &orig[idx..];
+        debug_assert!(idx <= orig.len());
+        let s = unsafe { orig.get_unchecked(idx..) };
         self.iter = s.char_indices();
         self.start_pos_of_iter = to;
         self.last_pos = to;
@@ -203,6 +205,11 @@ impl<'a> Input for StringInput<'a> {
         }
     }
 
+    #[inline]
+    fn is_str(&self, s: &str) -> bool {
+        self.as_str().starts_with(s)
+    }
+
     #[inline]
     fn eat_byte(&mut self, c: u8) -> bool {
         if self.is_byte(c) {
diff --git a/crates/swc_ecma_parser/src/lexer/jsx.rs b/crates/swc_ecma_parser/src/lexer/jsx.rs
index aab1ae833d0..ea64b0c188b 100644
--- a/crates/swc_ecma_parser/src/lexer/jsx.rs
+++ b/crates/swc_ecma_parser/src/lexer/jsx.rs
@@ -26,7 +26,7 @@ impl<'a> Lexer<'a> {
 
                     self.emit_error_span(span, SyntaxError::TS1185);
                     self.skip_line_comment(6);
-                    self.skip_space(true)?;
+                    self.skip_space::<true>()?;
                     return self.read_token();
                 }
                 '<' | '{' => {
diff --git a/crates/swc_ecma_parser/src/lexer/mod.rs b/crates/swc_ecma_parser/src/lexer/mod.rs
index 4341e74dda8..11efbc5d313 100644
--- a/crates/swc_ecma_parser/src/lexer/mod.rs
+++ b/crates/swc_ecma_parser/src/lexer/mod.rs
@@ -243,7 +243,7 @@ impl<'a> Lexer<'a> {
                                     SyntaxError::LegacyCommentInModule,
                                 );
                                 self.skip_line_comment(0);
-                                self.skip_space(true)?;
+                                self.skip_space::<true>()?;
                                 return self.read_token();
                             }
 
@@ -282,7 +282,7 @@ impl<'a> Lexer<'a> {
                                             SyntaxError::TS1185,
                                         );
                                         self.skip_line_comment(4);
-                                        self.skip_space(true)?;
+                                        self.skip_space::<true>()?;
                                         return self.read_token();
                                     }
 
@@ -303,6 +303,11 @@ impl<'a> Lexer<'a> {
                             AssignOp(Assign)
                         }));
                     }
+
+                    b'a'..=b'z' | b'A'..=b'Z' | b'$' | b'_' | b'\\' => {
+                        // Fast path for ascii identifiers.
+                        return self.read_ident_or_keyword().map(Some);
+                    }
                     _ => {}
                 }
             }
@@ -504,7 +509,7 @@ impl<'a> Lexer<'a> {
                 let span = fixed_len_span(start, 7);
                 self.emit_error_span(span, SyntaxError::TS1185);
                 self.skip_line_comment(5);
-                self.skip_space(true)?;
+                self.skip_space::<true>()?;
                 return self.error_span(span, SyntaxError::TS1185);
             }
 
@@ -719,7 +724,7 @@ impl<'a> Lexer<'a> {
         // XML style comment. `<!--`
         if c == '<' && self.is(b'!') && self.peek() == Some('-') && self.peek_ahead() == Some('-') {
             self.skip_line_comment(3);
-            self.skip_space(true)?;
+            self.skip_space::<true>()?;
             self.emit_module_mode_error(start, SyntaxError::LegacyCommentInModule);
 
             return self.read_token();
@@ -767,7 +772,7 @@ impl<'a> Lexer<'a> {
         {
             self.emit_error_span(fixed_len_span(start, 7), SyntaxError::TS1185);
             self.skip_line_comment(5);
-            self.skip_space(true)?;
+            self.skip_space::<true>()?;
             return self.read_token();
         }
 
diff --git a/crates/swc_ecma_parser/src/lexer/state.rs b/crates/swc_ecma_parser/src/lexer/state.rs
index 6ab4fd5c507..357162e661f 100644
--- a/crates/swc_ecma_parser/src/lexer/state.rs
+++ b/crates/swc_ecma_parser/src/lexer/state.rs
@@ -214,7 +214,7 @@ impl<'a> Iterator for Lexer<'a> {
 
             // skip spaces before getting next character, if we are allowed to.
             if self.state.can_skip_space() {
-                self.skip_space(true)?;
+                self.skip_space::<true>()?;
                 start = self.input.cur_pos();
             };
 
@@ -309,7 +309,7 @@ impl<'a> Iterator for Lexer<'a> {
 
                             self.emit_error_span(span, SyntaxError::TS1185);
                             self.skip_line_comment(6);
-                            self.skip_space(true)?;
+                            self.skip_space::<true>()?;
                             return self.read_token();
                         }
 
diff --git a/crates/swc_ecma_parser/src/lexer/util.rs b/crates/swc_ecma_parser/src/lexer/util.rs
index d01beb5792d..7365d4b7243 100644
--- a/crates/swc_ecma_parser/src/lexer/util.rs
+++ b/crates/swc_ecma_parser/src/lexer/util.rs
@@ -182,7 +182,7 @@ impl<'a> Lexer<'a> {
     /// Skip comments or whitespaces.
     ///
     /// See https://tc39.github.io/ecma262/#sec-white-space
-    pub(super) fn skip_space(&mut self, lex_comments: bool) -> LexResult<()> {
+    pub(super) fn skip_space<const LEX_COMMENTS: bool>(&mut self) -> LexResult<()> {
         loop {
             let cur_b = self.input.cur_as_ascii();
 
@@ -197,7 +197,7 @@ impl<'a> Lexer<'a> {
                 continue;
             }
 
-            if lex_comments && self.input.is_byte(b'/') {
+            if LEX_COMMENTS && self.input.is_byte(b'/') {
                 if self.peek() == Some('/') {
                     self.skip_line_comment(2);
                     continue;
@@ -308,7 +308,7 @@ impl<'a> Lexer<'a> {
 
                 let end = self.cur_pos();
 
-                self.skip_space(false)?;
+                self.skip_space::<false>()?;
 
                 if self.input.is_byte(b';') {
                     is_for_next = false;