cpython/Lib/lib2to3/patcomp.py
Benjamin Peterson f37eb3a184 Merged revisions 83852-83853,83857,84042,84216,84274-84276,84375,85388,85478,85506-85508 via svnmerge from
svn+ssh://pythondev@svn.python.org/sandbox/trunk/2to3/lib2to3

........
  r83852 | benjamin.peterson | 2010-08-08 15:45:44 -0500 (Sun, 08 Aug 2010) | 1 line

  wrap with parens
........
  r83853 | benjamin.peterson | 2010-08-08 15:46:31 -0500 (Sun, 08 Aug 2010) | 1 line

  use parens
........
  r83857 | benjamin.peterson | 2010-08-08 15:59:49 -0500 (Sun, 08 Aug 2010) | 1 line

  things which use touch_import should be pre order
........
  r84042 | george.boutsioukis | 2010-08-14 16:10:19 -0500 (Sat, 14 Aug 2010) | 2 lines

  This revision incorporates into the 2to3 tool the new, faster, tree matching algorithm developed during a GSOC project. The algorithm resides in the two added modules, btm_matcher and btm_utils. New code has been added to drive the new matching process in refactor.py and a few minor changes were made in other modules. A BM_compatible flag(False by default) has been added in fixer_base and it is set to True in most of the current fixers.
........
  r84216 | benjamin.peterson | 2010-08-19 16:44:05 -0500 (Thu, 19 Aug 2010) | 1 line

  allow star_expr in testlist_gexp
........
  r84274 | benjamin.peterson | 2010-08-22 18:40:46 -0500 (Sun, 22 Aug 2010) | 1 line

  wrap long line
........
  r84275 | benjamin.peterson | 2010-08-22 18:42:22 -0500 (Sun, 22 Aug 2010) | 1 line

  cleanup
........
  r84276 | benjamin.peterson | 2010-08-22 18:51:01 -0500 (Sun, 22 Aug 2010) | 1 line

  when there's a None value and a traceback, don't call type with it #9661
........
  r84375 | george.boutsioukis | 2010-08-31 08:38:53 -0500 (Tue, 31 Aug 2010) | 3 lines

  Idiomatic code changes & stylistic issues fixed in the BottomMatcher module. Thanks to Benjamin Peterson for taking the time to review the code.
........
  r85388 | benjamin.peterson | 2010-10-12 17:27:44 -0500 (Tue, 12 Oct 2010) | 1 line

  fix urllib fixer with multiple as imports on a line #10069
........
  r85478 | benjamin.peterson | 2010-10-14 08:09:56 -0500 (Thu, 14 Oct 2010) | 1 line

  stop abusing docstrings
........
  r85506 | benjamin.peterson | 2010-10-14 17:45:19 -0500 (Thu, 14 Oct 2010) | 1 line

  kill sibling import
........
  r85507 | benjamin.peterson | 2010-10-14 17:54:15 -0500 (Thu, 14 Oct 2010) | 1 line

  remove trailing whitespace
........
  r85508 | benjamin.peterson | 2010-10-14 17:55:28 -0500 (Thu, 14 Oct 2010) | 1 line

  typo
........
2010-10-14 23:00:04 +00:00

205 lines
6.9 KiB
Python

# Copyright 2006 Google, Inc. All Rights Reserved.
# Licensed to PSF under a Contributor Agreement.
"""Pattern compiler.
The grammer is taken from PatternGrammar.txt.
The compiler compiles a pattern to a pytree.*Pattern instance.
"""
__author__ = "Guido van Rossum <guido@python.org>"
# Python imports
import os
# Fairly local imports
from .pgen2 import driver, literals, token, tokenize, parse, grammar
# Really local imports
from . import pytree
from . import pygram
# The pattern grammar file
_PATTERN_GRAMMAR_FILE = os.path.join(os.path.dirname(__file__),
"PatternGrammar.txt")
class PatternSyntaxError(Exception):
pass
def tokenize_wrapper(input):
"""Tokenizes a string suppressing significant whitespace."""
skip = set((token.NEWLINE, token.INDENT, token.DEDENT))
tokens = tokenize.generate_tokens(driver.generate_lines(input).__next__)
for quintuple in tokens:
type, value, start, end, line_text = quintuple
if type not in skip:
yield quintuple
class PatternCompiler(object):
def __init__(self, grammar_file=_PATTERN_GRAMMAR_FILE):
"""Initializer.
Takes an optional alternative filename for the pattern grammar.
"""
self.grammar = driver.load_grammar(grammar_file)
self.syms = pygram.Symbols(self.grammar)
self.pygrammar = pygram.python_grammar
self.pysyms = pygram.python_symbols
self.driver = driver.Driver(self.grammar, convert=pattern_convert)
def compile_pattern(self, input, debug=False, with_tree=False):
"""Compiles a pattern string to a nested pytree.*Pattern object."""
tokens = tokenize_wrapper(input)
try:
root = self.driver.parse_tokens(tokens, debug=debug)
except parse.ParseError as e:
raise PatternSyntaxError(str(e))
if with_tree:
return self.compile_node(root), root
else:
return self.compile_node(root)
def compile_node(self, node):
"""Compiles a node, recursively.
This is one big switch on the node type.
"""
# XXX Optimize certain Wildcard-containing-Wildcard patterns
# that can be merged
if node.type == self.syms.Matcher:
node = node.children[0] # Avoid unneeded recursion
if node.type == self.syms.Alternatives:
# Skip the odd children since they are just '|' tokens
alts = [self.compile_node(ch) for ch in node.children[::2]]
if len(alts) == 1:
return alts[0]
p = pytree.WildcardPattern([[a] for a in alts], min=1, max=1)
return p.optimize()
if node.type == self.syms.Alternative:
units = [self.compile_node(ch) for ch in node.children]
if len(units) == 1:
return units[0]
p = pytree.WildcardPattern([units], min=1, max=1)
return p.optimize()
if node.type == self.syms.NegatedUnit:
pattern = self.compile_basic(node.children[1:])
p = pytree.NegatedPattern(pattern)
return p.optimize()
assert node.type == self.syms.Unit
name = None
nodes = node.children
if len(nodes) >= 3 and nodes[1].type == token.EQUAL:
name = nodes[0].value
nodes = nodes[2:]
repeat = None
if len(nodes) >= 2 and nodes[-1].type == self.syms.Repeater:
repeat = nodes[-1]
nodes = nodes[:-1]
# Now we've reduced it to: STRING | NAME [Details] | (...) | [...]
pattern = self.compile_basic(nodes, repeat)
if repeat is not None:
assert repeat.type == self.syms.Repeater
children = repeat.children
child = children[0]
if child.type == token.STAR:
min = 0
max = pytree.HUGE
elif child.type == token.PLUS:
min = 1
max = pytree.HUGE
elif child.type == token.LBRACE:
assert children[-1].type == token.RBRACE
assert len(children) in (3, 5)
min = max = self.get_int(children[1])
if len(children) == 5:
max = self.get_int(children[3])
else:
assert False
if min != 1 or max != 1:
pattern = pattern.optimize()
pattern = pytree.WildcardPattern([[pattern]], min=min, max=max)
if name is not None:
pattern.name = name
return pattern.optimize()
def compile_basic(self, nodes, repeat=None):
# Compile STRING | NAME [Details] | (...) | [...]
assert len(nodes) >= 1
node = nodes[0]
if node.type == token.STRING:
value = str(literals.evalString(node.value))
return pytree.LeafPattern(_type_of_literal(value), value)
elif node.type == token.NAME:
value = node.value
if value.isupper():
if value not in TOKEN_MAP:
raise PatternSyntaxError("Invalid token: %r" % value)
if nodes[1:]:
raise PatternSyntaxError("Can't have details for token")
return pytree.LeafPattern(TOKEN_MAP[value])
else:
if value == "any":
type = None
elif not value.startswith("_"):
type = getattr(self.pysyms, value, None)
if type is None:
raise PatternSyntaxError("Invalid symbol: %r" % value)
if nodes[1:]: # Details present
content = [self.compile_node(nodes[1].children[1])]
else:
content = None
return pytree.NodePattern(type, content)
elif node.value == "(":
return self.compile_node(nodes[1])
elif node.value == "[":
assert repeat is None
subpattern = self.compile_node(nodes[1])
return pytree.WildcardPattern([[subpattern]], min=0, max=1)
assert False, node
def get_int(self, node):
assert node.type == token.NUMBER
return int(node.value)
# Map named tokens to the type value for a LeafPattern
TOKEN_MAP = {"NAME": token.NAME,
"STRING": token.STRING,
"NUMBER": token.NUMBER,
"TOKEN": None}
def _type_of_literal(value):
if value[0].isalpha():
return token.NAME
elif value in grammar.opmap:
return grammar.opmap[value]
else:
return None
def pattern_convert(grammar, raw_node_info):
"""Converts raw node information to a Node or Leaf instance."""
type, value, context, children = raw_node_info
if children or type in grammar.number2symbol:
return pytree.Node(type, children, context=context)
else:
return pytree.Leaf(type, value, context=context)
def compile_pattern(pattern):
return PatternCompiler().compile_pattern(pattern)