diff options
Diffstat (limited to 'Lib/lib2to3/refactor.py')
-rw-r--r-- | Lib/lib2to3/refactor.py | 137 |
1 files changed, 100 insertions, 37 deletions
diff --git a/Lib/lib2to3/refactor.py b/Lib/lib2to3/refactor.py index 339b94f..5edf584 100644 --- a/Lib/lib2to3/refactor.py +++ b/Lib/lib2to3/refactor.py @@ -14,14 +14,15 @@ __author__ = "Guido van Rossum <guido@python.org>" # Python imports import os import sys -import difflib import logging import operator -from collections import defaultdict +import collections +import io +import warnings from itertools import chain # Local imports -from .pgen2 import driver, tokenize +from .pgen2 import driver, tokenize, token from . import pytree, pygram @@ -37,7 +38,12 @@ def get_all_fix_names(fixer_pkg, remove_prefix=True): fix_names.append(name[:-3]) return fix_names -def get_head_types(pat): + +class _EveryNode(Exception): + pass + + +def _get_head_types(pat): """ Accepts a pytree Pattern Node and returns a set of the pattern types which will match first. """ @@ -45,34 +51,50 @@ def get_head_types(pat): # NodePatters must either have no type and no content # or a type and content -- so they don't get any farther # Always return leafs + if pat.type is None: + raise _EveryNode return set([pat.type]) if isinstance(pat, pytree.NegatedPattern): if pat.content: - return get_head_types(pat.content) - return set([None]) # Negated Patterns don't have a type + return _get_head_types(pat.content) + raise _EveryNode # Negated Patterns don't have a type if isinstance(pat, pytree.WildcardPattern): # Recurse on each node in content r = set() for p in pat.content: for x in p: - r.update(get_head_types(x)) + r.update(_get_head_types(x)) return r raise Exception("Oh no! I don't understand pattern %s" %(pat)) -def get_headnode_dict(fixer_list): + +def _get_headnode_dict(fixer_list): """ Accepts a list of fixers and returns a dictionary of head node type --> fixer list. """ - head_nodes = defaultdict(list) + head_nodes = collections.defaultdict(list) + every = [] for fixer in fixer_list: - if not fixer.pattern: - head_nodes[None].append(fixer) - continue - for t in get_head_types(fixer.pattern): - head_nodes[t].append(fixer) - return head_nodes + if fixer.pattern: + try: + heads = _get_head_types(fixer.pattern) + except _EveryNode: + every.append(fixer) + else: + for node_type in heads: + head_nodes[node_type].append(fixer) + else: + if fixer._accept_type is not None: + head_nodes[fixer._accept_type].append(fixer) + else: + every.append(fixer) + for node_type in chain(pygram.python_grammar.symbol2number.values(), + pygram.python_grammar.tokens): + head_nodes[node_type].extend(every) + return dict(head_nodes) + def get_fixers_from_package(pkg_name): """ @@ -101,13 +123,56 @@ else: _to_system_newlines = _identity +def _detect_future_print(source): + have_docstring = False + gen = tokenize.generate_tokens(io.StringIO(source).readline) + def advance(): + tok = next(gen) + return tok[0], tok[1] + ignore = frozenset((token.NEWLINE, tokenize.NL, token.COMMENT)) + try: + while True: + tp, value = advance() + if tp in ignore: + continue + elif tp == token.STRING: + if have_docstring: + break + have_docstring = True + elif tp == token.NAME: + if value == "from": + tp, value = advance() + if tp != token.NAME and value != "__future__": + break + tp, value = advance() + if tp != token.NAME and value != "import": + break + tp, value = advance() + if tp == token.OP and value == "(": + tp, value = advance() + while tp == token.NAME: + if value == "print_function": + return True + tp, value = advance() + if tp != token.OP and value != ",": + break + tp, value = advance() + else: + break + else: + break + except StopIteration: + pass + return False + + class FixerError(Exception): """A fixer could not be loaded.""" class RefactoringTool(object): - _default_options = {"print_function": False} + _default_options = {} CLASS_PREFIX = "Fix" # The prefix for fixer classes FILE_PREFIX = "fix_" # The prefix for modules with a fixer within @@ -124,20 +189,21 @@ class RefactoringTool(object): self.explicit = explicit or [] self.options = self._default_options.copy() if options is not None: + if "print_function" in options: + warnings.warn("the 'print_function' option is deprecated", + DeprecationWarning) self.options.update(options) self.errors = [] self.logger = logging.getLogger("RefactoringTool") self.fixer_log = [] self.wrote = False - if self.options["print_function"]: - del pygram.python_grammar.keywords["print"] self.driver = driver.Driver(pygram.python_grammar, convert=pytree.convert, logger=self.logger) self.pre_order, self.post_order = self.get_fixers() - self.pre_order_heads = get_headnode_dict(self.pre_order) - self.post_order_heads = get_headnode_dict(self.post_order) + self.pre_order_heads = _get_headnode_dict(self.pre_order) + self.post_order_heads = _get_headnode_dict(self.post_order) self.files = [] # List of files that were or should be modified @@ -196,8 +262,9 @@ class RefactoringTool(object): msg = msg % args self.logger.debug(msg) - def print_output(self, lines): - """Called with lines of output to give to the user.""" + def print_output(self, old_text, new_text, filename, equal): + """Called with the old version, new version, and filename of a + refactored file.""" pass def refactor(self, items, write=False, doctests_only=False): @@ -220,7 +287,8 @@ class RefactoringTool(object): dirnames.sort() filenames.sort() for name in filenames: - if not name.startswith(".") and name.endswith("py"): + if not name.startswith(".") and \ + os.path.splitext(name)[1].endswith("py"): fullname = os.path.join(dirpath, name) self.refactor_file(fullname, write, doctests_only) # Modify dirnames in-place to remove subdirs with leading dots @@ -276,12 +344,16 @@ class RefactoringTool(object): An AST corresponding to the refactored input stream; None if there were errors during the parse. """ + if _detect_future_print(data): + self.driver.grammar = pygram.python_grammar_no_print_statement try: tree = self.driver.parse_string(data) except Exception as err: self.log_error("Can't parse %s: %s: %s", name, err.__class__.__name__, err) return + finally: + self.driver.grammar = pygram.python_grammar self.log_debug("Refactoring %s", name) self.refactor_tree(tree, name) return tree @@ -338,12 +410,11 @@ class RefactoringTool(object): if not fixers: return for node in traversal: - for fixer in fixers[node.type] + fixers[None]: + for fixer in fixers[node.type]: results = fixer.match(node) if results: new = fixer.transform(node, results) - if new is not None and (new != node or - str(new) != str(node)): + if new is not None: node.replace(new) node = new @@ -357,10 +428,11 @@ class RefactoringTool(object): old_text = self._read_python_source(filename)[0] if old_text is None: return - if old_text == new_text: + equal = old_text == new_text + self.print_output(old_text, new_text, filename, equal) + if equal: self.log_debug("No changes to %s", filename) return - self.print_output(diff_texts(old_text, new_text, filename)) if write: self.write_file(new_text, filename, old_text, encoding) else: @@ -582,12 +654,3 @@ class MultiprocessRefactoringTool(RefactoringTool): else: return super(MultiprocessRefactoringTool, self).refactor_file( *args, **kwargs) - - -def diff_texts(a, b, filename): - """Return a unified diff of two strings.""" - a = a.splitlines() - b = b.splitlines() - return difflib.unified_diff(a, b, filename, filename, - "(original)", "(refactored)", - lineterm="") |