Fix a string merging/split issue caused by standalone comments. (#3227)

[etc/vim.git] / src / black / parsing.py

diff --git a/src/black/parsing.py b/src/black/parsing.py
index 6b63368871c80c29a4e353d4c126d141235d0f61..64c0b1e30183fda1e711c27bda48264cba4e000d 100644 (file)
--- a/src/black/parsing.py
+++ b/src/black/parsing.py
@@ -11,16 +11,14 @@ if sys.version_info < (3, 8):
 else:
     from typing import Final
 
-# lib2to3 fork
-from blib2to3.pytree import Node, Leaf
+from black.mode import Feature, TargetVersion, supports_feature
+from black.nodes import syms
 from blib2to3 import pygram
 from blib2to3.pgen2 import driver
 from blib2to3.pgen2.grammar import Grammar
 from blib2to3.pgen2.parse import ParseError
 from blib2to3.pgen2.tokenize import TokenError
-
-from black.mode import TargetVersion, Feature, supports_feature
-from black.nodes import syms
+from blib2to3.pytree import Leaf, Node
 
 ast3: Any
 
@@ -42,7 +40,6 @@ except ImportError:
         ast3 = ast
 
 
-PY310_HINT: Final = "Consider using --target-version py310 to parse Python 3.10 code."
 PY2_HINT: Final = "Python 2 support was removed in version 22.0."
 
 
@@ -58,12 +55,11 @@ def get_grammars(target_versions: Set[TargetVersion]) -> List[Grammar]:
             pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords,
             # Python 3.0-3.6
             pygram.python_grammar_no_print_statement_no_exec_statement,
+            # Python 3.10+
+            pygram.python_grammar_soft_keywords,
         ]
 
     grammars = []
-    if supports_feature(target_versions, Feature.PATTERN_MATCHING):
-        # Python 3.10+
-        grammars.append(pygram.python_grammar_soft_keywords)
     # If we have to parse both, try to parse async as a keyword first
     if not supports_feature(
         target_versions, Feature.ASYNC_IDENTIFIERS
@@ -75,6 +71,10 @@ def get_grammars(target_versions: Set[TargetVersion]) -> List[Grammar]:
     if not supports_feature(target_versions, Feature.ASYNC_KEYWORDS):
         # Python 3.0-3.6
         grammars.append(pygram.python_grammar_no_print_statement_no_exec_statement)
+    if supports_feature(target_versions, Feature.PATTERN_MATCHING):
+        # Python 3.10+
+        grammars.append(pygram.python_grammar_soft_keywords)
+
     # At least one of the above branches must have been taken, because every Python
     # version has exactly one of the two 'ASYNC_*' flags
     return grammars
@@ -86,6 +86,7 @@ def lib2to3_parse(src_txt: str, target_versions: Iterable[TargetVersion] = ()) -
         src_txt += "\n"
 
     grammars = get_grammars(set(target_versions))
+    errors = {}
     for grammar in grammars:
         drv = driver.Driver(grammar)
         try:
@@ -99,20 +100,21 @@ def lib2to3_parse(src_txt: str, target_versions: Iterable[TargetVersion] = ()) -
                 faulty_line = lines[lineno - 1]
             except IndexError:
                 faulty_line = "<line number missing in source>"
-            exc = InvalidInput(f"Cannot parse: {lineno}:{column}: {faulty_line}")
+            errors[grammar.version] = InvalidInput(
+                f"Cannot parse: {lineno}:{column}: {faulty_line}"
+            )
 
         except TokenError as te:
             # In edge cases these are raised; and typically don't have a "faulty_line".
             lineno, column = te.args[1]
-            exc = InvalidInput(f"Cannot parse: {lineno}:{column}: {te.args[0]}")
+            errors[grammar.version] = InvalidInput(
+                f"Cannot parse: {lineno}:{column}: {te.args[0]}"
+            )
 
     else:
-        if pygram.python_grammar_soft_keywords not in grammars and matches_grammar(
-            src_txt, pygram.python_grammar_soft_keywords
-        ):
-            original_msg = exc.args[0]
-            msg = f"{original_msg}\n{PY310_HINT}"
-            raise InvalidInput(msg) from None
+        # Choose the latest version when raising the actual parsing error.
+        assert len(errors) >= 1
+        exc = errors[max(errors)]
 
         if matches_grammar(src_txt, pygram.python_grammar) or matches_grammar(
             src_txt, pygram.python_grammar_no_print_statement
@@ -148,14 +150,22 @@ def parse_single_version(
     src: str, version: Tuple[int, int]
 ) -> Union[ast.AST, ast3.AST]:
     filename = "<unknown>"
-    # typed_ast is needed because of feature version limitations in the builtin ast
+    # typed-ast is needed because of feature version limitations in the builtin ast 3.8>
     if sys.version_info >= (3, 8) and version >= (3,):
-        return ast.parse(src, filename, feature_version=version)
-    elif version >= (3,):
-        if _IS_PYPY:
-            return ast3.parse(src, filename)
+        return ast.parse(src, filename, feature_version=version, type_comments=True)
+
+    if _IS_PYPY:
+        # PyPy 3.7 doesn't support type comment tracking which is not ideal, but there's
+        # not much we can do as typed-ast won't work either.
+        if sys.version_info >= (3, 8):
+            return ast3.parse(src, filename, type_comments=True)
         else:
-            return ast3.parse(src, filename, feature_version=version[1])
+            return ast3.parse(src, filename)
+    else:
+        # Typed-ast is guaranteed to be used here and automatically tracks type
+        # comments separately.
+        return ast3.parse(src, filename, feature_version=version[1])
+
     raise AssertionError("INTERNAL ERROR: Tried parsing unsupported Python version!")
 
 
@@ -221,8 +231,8 @@ def stringify_ast(node: Union[ast.AST, ast3.AST], depth: int = 0) -> Iterator[st
                     and isinstance(node, (ast.Delete, ast3.Delete))
                     and isinstance(item, (ast.Tuple, ast3.Tuple))
                 ):
-                    for item in item.elts:
-                        yield from stringify_ast(item, depth + 2)
+                    for elt in item.elts:
+                        yield from stringify_ast(elt, depth + 2)
 
                 elif isinstance(item, (ast.AST, ast3.AST)):
                     yield from stringify_ast(item, depth + 2)