Replace the blib2to3 tokenizer with pytokens (#4536)

Author: tusharsadhwani

.pre-commit-config.yaml

@@ -51,6 +51,7 @@ repos:
           - click >= 8.1.0, != 8.1.4, != 8.1.5
           - packaging >= 22.0
           - platformdirs >= 2.1.0
+          - pytokens >= 0.1.10
           - pytest
           - hypothesis
           - aiohttp >= 3.7.4

CHANGES.md

@@ -29,6 +29,7 @@
 
 <!-- Changes to the parser or to version autodetection -->
 
+- Rewrite tokenizer to improve performance and compliance (#4536)
 - Fix bug where certain unusual expressions (e.g., lambdas) were not accepted
   in type parameter bounds and defaults. (#4602)
 

pyproject.toml

@@ -69,6 +69,7 @@ dependencies = [
   "packaging>=22.0",
   "pathspec>=0.9.0",
   "platformdirs>=2",
+  "pytokens>=0.1.10",
   "tomli>=1.1.0; python_version < '3.11'",
   "typing_extensions>=4.0.1; python_version < '3.11'",
 ]

src/blib2to3/pgen2/driver.py

@@ -28,7 +28,7 @@
 from typing import IO, Any, Optional, Union, cast
 
 from blib2to3.pgen2.grammar import Grammar
-from blib2to3.pgen2.tokenize import GoodTokenInfo
+from blib2to3.pgen2.tokenize import TokenInfo
 from blib2to3.pytree import NL
 
 # Pgen imports
@@ -112,7 +112,7 @@ def __init__(self, grammar: Grammar, logger: Optional[Logger] = None) -> None:
             logger = logging.getLogger(__name__)
         self.logger = logger
 
-    def parse_tokens(self, tokens: Iterable[GoodTokenInfo], debug: bool = False) -> NL:
+    def parse_tokens(self, tokens: Iterable[TokenInfo], debug: bool = False) -> NL:
         """Parse a series of tokens and return the syntax tree."""
         # XXX Move the prefix computation into a wrapper around tokenize.
         proxy = TokenProxy(tokens)
@@ -180,27 +180,17 @@ def parse_tokens(self, tokens: Iterable[GoodTokenInfo], debug: bool = False) ->
         assert p.rootnode is not None
         return p.rootnode
 
-    def parse_stream_raw(self, stream: IO[str], debug: bool = False) -> NL:
-        """Parse a stream and return the syntax tree."""
-        tokens = tokenize.generate_tokens(stream.readline, grammar=self.grammar)
-        return self.parse_tokens(tokens, debug)
-
-    def parse_stream(self, stream: IO[str], debug: bool = False) -> NL:
-        """Parse a stream and return the syntax tree."""
-        return self.parse_stream_raw(stream, debug)
-
     def parse_file(
         self, filename: Path, encoding: Optional[str] = None, debug: bool = False
     ) -> NL:
         """Parse a file and return the syntax tree."""
         with open(filename, encoding=encoding) as stream:
-            return self.parse_stream(stream, debug)
+            text = stream.read()
+        return self.parse_string(text, debug)
 
     def parse_string(self, text: str, debug: bool = False) -> NL:
         """Parse a string and return the syntax tree."""
-        tokens = tokenize.generate_tokens(
-            io.StringIO(text).readline, grammar=self.grammar
-        )
+        tokens = tokenize.tokenize(text, grammar=self.grammar)
         return self.parse_tokens(tokens, debug)
 
     def _partially_consume_prefix(self, prefix: str, column: int) -> tuple[str, str]:

src/blib2to3/pgen2/pgen.py

@@ -6,7 +6,7 @@
 from typing import IO, Any, NoReturn, Optional, Union
 
 from blib2to3.pgen2 import grammar, token, tokenize
-from blib2to3.pgen2.tokenize import GoodTokenInfo
+from blib2to3.pgen2.tokenize import TokenInfo
 
 Path = Union[str, "os.PathLike[str]"]
 
@@ -18,7 +18,7 @@ class PgenGrammar(grammar.Grammar):
 class ParserGenerator:
     filename: Path
     stream: IO[str]
-    generator: Iterator[GoodTokenInfo]
+    generator: Iterator[TokenInfo]
     first: dict[str, Optional[dict[str, int]]]
 
     def __init__(self, filename: Path, stream: Optional[IO[str]] = None) -> None:
@@ -27,8 +27,7 @@ def __init__(self, filename: Path, stream: Optional[IO[str]] = None) -> None:
             stream = open(filename, encoding="utf-8")
             close_stream = stream.close
         self.filename = filename
-        self.stream = stream
-        self.generator = tokenize.generate_tokens(stream.readline)
+        self.generator = tokenize.tokenize(stream.read())
         self.gettoken()  # Initialize lookahead
         self.dfas, self.startsymbol = self.parse()
         if close_stream is not None: