[Python-checkins] gh-102856: Tokenize performance improvement (#104731)

[pablogsal]

https://github.com/python/cpython/commit/8817886ae571f5b5ce4e2e6cfd2458622d0efac1
commit: 8817886ae571f5b5ce4e2e6cfd2458622d0efac1
branch: main
author: Marta Gómez Macías <mgmacias at google.com>
committer: pablogsal <Pablogsal at gmail.com>
date: 2023-05-22T00:29:04Z
summary:

gh-102856: Tokenize performance improvement (#104731)

files:
M Lib/tokenize.py
M Python/Python-tokenize.c

diff --git a/Lib/tokenize.py b/Lib/tokenize.py
index cef2773feac2..911f0f12f9bb 100644
--- a/Lib/tokenize.py
+++ b/Lib/tokenize.py
@@ -449,16 +449,6 @@ def _tokenize(rl_gen, encoding):
     source = b"".join(rl_gen).decode(encoding)
     token = None
     for token in _generate_tokens_from_c_tokenizer(source, extra_tokens=True):
-        # TODO: Marta -> limpiar esto
-        if 6 < token.type <= 54:
-            token = token._replace(type=OP)
-        if token.type in {ASYNC, AWAIT}:
-            token = token._replace(type=NAME)
-        if token.type == NEWLINE:
-            l_start, c_start = token.start
-            l_end, c_end = token.end
-            token = token._replace(string='\n', start=(l_start, c_start), end=(l_end, c_end+1))
-
         yield token
     if token is not None:
         last_line, _ = token.start
@@ -550,8 +540,7 @@ def _generate_tokens_from_c_tokenizer(source, extra_tokens=False):
     """Tokenize a source reading Python code as unicode strings using the internal C tokenizer"""
     import _tokenize as c_tokenizer
     for info in c_tokenizer.TokenizerIter(source, extra_tokens=extra_tokens):
-        tok, type, lineno, end_lineno, col_off, end_col_off, line = info
-        yield TokenInfo(type, tok, (lineno, col_off), (end_lineno, end_col_off), line)
+        yield TokenInfo._make(info)
 
 
 if __name__ == "__main__":
diff --git a/Python/Python-tokenize.c b/Python/Python-tokenize.c
index ece238672e34..43b44be94583 100644
--- a/Python/Python-tokenize.c
+++ b/Python/Python-tokenize.c
@@ -207,7 +207,22 @@ tokenizeriter_next(tokenizeriterobject *it)
         end_col_offset = _PyPegen_byte_offset_to_character_offset(line, token.end - it->tok->line_start);
     }
 
-    result = Py_BuildValue("(NinnnnN)", str, type, lineno, end_lineno, col_offset, end_col_offset, line);
+    if (it->tok->tok_extra_tokens) {
+        // Necessary adjustments to match the original Python tokenize
+        // implementation
+        if (type > DEDENT && type < OP) {
+            type = OP;
+        }
+        else if (type == ASYNC || type == AWAIT) {
+            type = NAME;
+        }
+        else if (type == NEWLINE) {
+            str = PyUnicode_FromString("\n");
+            end_col_offset++;
+        }
+    }
+
+    result = Py_BuildValue("(iN(nn)(nn)N)", type, str, lineno, col_offset, end_lineno, end_col_offset, line);
 exit:
     _PyToken_Free(&token);
     return result;