Fix parser bug where "type" was misinterpreted as a keyword inside a match (#3950)

[etc/vim.git] / src / blib2to3 / pgen2 / parse.py

diff --git a/src/blib2to3/pgen2/parse.py b/src/blib2to3/pgen2/parse.py
index a9dc11f39ce15deedbcb57cbef8d5774c238d4ea..ad51a3dad08dfbbe1fee7cc9b7845eeb7cb263fd 100644 (file)
--- a/src/blib2to3/pgen2/parse.py
+++ b/src/blib2to3/pgen2/parse.py
@@ -9,33 +9,32 @@ See Parser/parser.c in the Python distribution for additional info on
 how this parsing engine works.
 
 """
-import copy
 from contextlib import contextmanager
-
-# Local imports
-from . import grammar, token, tokenize
 from typing import (
-    cast,
+    TYPE_CHECKING,
     Any,
-    Optional,
-    Text,
-    Union,
-    Tuple,
+    Callable,
     Dict,
-    List,
     Iterator,
-    Callable,
+    List,
+    Optional,
     Set,
-    TYPE_CHECKING,
+    Tuple,
+    Union,
+    cast,
 )
+
 from blib2to3.pgen2.grammar import Grammar
-from blib2to3.pytree import convert, NL, Context, RawNode, Leaf, Node
+from blib2to3.pytree import NL, Context, Leaf, Node, RawNode, convert
+
+# Local imports
+from . import grammar, token, tokenize
 
 if TYPE_CHECKING:
-    from blib2to3.driver import TokenProxy
+    from blib2to3.pgen2.driver import TokenProxy
 
 
-Results = Dict[Text, NL]
+Results = Dict[str, NL]
 Convert = Callable[[Grammar, RawNode], Union[Node, Leaf]]
 DFA = List[List[Tuple[int, int]]]
 DFAS = Tuple[DFA, Dict[int, int]]
@@ -100,7 +99,7 @@ class Recorder:
         finally:
             self.parser.is_backtracking = is_backtracking
 
-    def add_token(self, tok_type: int, tok_val: Text, raw: bool = False) -> None:
+    def add_token(self, tok_type: int, tok_val: str, raw: bool = False) -> None:
         func: Callable[..., Any]
         if raw:
             func = self.parser._addtoken
@@ -114,7 +113,9 @@ class Recorder:
                     args.insert(0, ilabel)
                 func(*args)
 
-    def determine_route(self, value: Text = None, force: bool = False) -> Optional[int]:
+    def determine_route(
+        self, value: Optional[str] = None, force: bool = False
+    ) -> Optional[int]:
         alive_ilabels = self.ilabels
         if len(alive_ilabels) == 0:
             *_, most_successful_ilabel = self._dead_ilabels
@@ -131,10 +132,10 @@ class ParseError(Exception):
     """Exception to signal the parser is stuck."""
 
     def __init__(
-        self, msg: Text, type: Optional[int], value: Optional[Text], context: Context
+        self, msg: str, type: Optional[int], value: Optional[str], context: Context
     ) -> None:
         Exception.__init__(
-            self, "%s: type=%r, value=%r, context=%r" % (msg, type, value, context)
+            self, f"{msg}: type={type!r}, value={value!r}, context={context!r}"
         )
         self.msg = msg
         self.type = type
@@ -142,7 +143,7 @@ class ParseError(Exception):
         self.context = context
 
 
-class Parser(object):
+class Parser:
     """Parser engine.
 
     The proper usage sequence is:
@@ -210,6 +211,7 @@ class Parser(object):
         # See note in docstring above. TL;DR this is ignored.
         self.convert = convert or lam_sub
         self.is_backtracking = False
+        self.last_token: Optional[int] = None
 
     def setup(self, proxy: "TokenProxy", start: Optional[int] = None) -> None:
         """Prepare for parsing.
@@ -235,8 +237,9 @@ class Parser(object):
         self.rootnode: Optional[NL] = None
         self.used_names: Set[str] = set()
         self.proxy = proxy
+        self.last_token = None
 
-    def addtoken(self, type: int, value: Text, context: Context) -> bool:
+    def addtoken(self, type: int, value: str, context: Context) -> bool:
         """Add a token; return True iff this is the end of the program."""
         # Map from token to label
         ilabels = self.classify(type, value, context)
@@ -284,7 +287,7 @@ class Parser(object):
 
         return self._addtoken(ilabel, type, value, context)
 
-    def _addtoken(self, ilabel: int, type: int, value: Text, context: Context) -> bool:
+    def _addtoken(self, ilabel: int, type: int, value: str, context: Context) -> bool:
         # Loop until the token is shifted; may raise exceptions
         while True:
             dfa, state, node = self.stack[-1]
@@ -316,6 +319,7 @@ class Parser(object):
                         dfa, state, node = self.stack[-1]
                         states, first = dfa
                     # Done with this token
+                    self.last_token = type
                     return False
 
             else:
@@ -329,7 +333,7 @@ class Parser(object):
                     # No success finding a transition
                     raise ParseError("bad input", type, value, context)
 
-    def classify(self, type: int, value: Text, context: Context) -> List[int]:
+    def classify(self, type: int, value: str, context: Context) -> List[int]:
         """Turn a token into a label.  (Internal)
 
         Depending on whether the value is a soft-keyword or not,
@@ -342,9 +346,23 @@ class Parser(object):
                 return [self.grammar.keywords[value]]
             elif value in self.grammar.soft_keywords:
                 assert type in self.grammar.tokens
+                # Current soft keywords (match, case, type) can only appear at the
+                # beginning of a statement. So as a shortcut, don't try to treat them
+                # like keywords in any other context.
+                # ('_' is also a soft keyword in the real grammar, but for our grammar
+                # it's just an expression, so we don't need to treat it specially.)
+                if self.last_token not in (
+                    None,
+                    token.INDENT,
+                    token.DEDENT,
+                    token.NEWLINE,
+                    token.SEMI,
+                    token.COLON,
+                ):
+                    return [self.grammar.tokens[type]]
                 return [
-                    self.grammar.soft_keywords[value],
                     self.grammar.tokens[type],
+                    self.grammar.soft_keywords[value],
                 ]
 
         ilabel = self.grammar.tokens.get(type)
@@ -352,7 +370,7 @@ class Parser(object):
             raise ParseError("bad token", type, value, context)
         return [ilabel]
 
-    def shift(self, type: int, value: Text, newstate: int, context: Context) -> None:
+    def shift(self, type: int, value: str, newstate: int, context: Context) -> None:
         """Shift a token.  (Internal)"""
         if self.is_backtracking:
             dfa, state, _ = self.stack[-1]