• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# Copyright 2006 Google, Inc. All Rights Reserved.
2# Licensed to PSF under a Contributor Agreement.
3
4"""Refactoring framework.
5
6Used as a main program, this can refactor any number of files and/or
7recursively descend down directories.  Imported as a module, this
8provides infrastructure to write your own refactoring tool.
9"""
10
11from __future__ import with_statement
12
13__author__ = "Guido van Rossum <guido@python.org>"
14
15
16# Python imports
17import os
18import sys
19import logging
20import operator
21import collections
22import StringIO
23from itertools import chain
24
25# Local imports
26from .pgen2 import driver, tokenize, token
27from .fixer_util import find_root
28from . import pytree, pygram
29from . import btm_utils as bu
30from . import btm_matcher as bm
31
32
33def get_all_fix_names(fixer_pkg, remove_prefix=True):
34    """Return a sorted list of all available fix names in the given package."""
35    pkg = __import__(fixer_pkg, [], [], ["*"])
36    fixer_dir = os.path.dirname(pkg.__file__)
37    fix_names = []
38    for name in sorted(os.listdir(fixer_dir)):
39        if name.startswith("fix_") and name.endswith(".py"):
40            if remove_prefix:
41                name = name[4:]
42            fix_names.append(name[:-3])
43    return fix_names
44
45
46class _EveryNode(Exception):
47    pass
48
49
50def _get_head_types(pat):
51    """ Accepts a pytree Pattern Node and returns a set
52        of the pattern types which will match first. """
53
54    if isinstance(pat, (pytree.NodePattern, pytree.LeafPattern)):
55        # NodePatters must either have no type and no content
56        #   or a type and content -- so they don't get any farther
57        # Always return leafs
58        if pat.type is None:
59            raise _EveryNode
60        return set([pat.type])
61
62    if isinstance(pat, pytree.NegatedPattern):
63        if pat.content:
64            return _get_head_types(pat.content)
65        raise _EveryNode # Negated Patterns don't have a type
66
67    if isinstance(pat, pytree.WildcardPattern):
68        # Recurse on each node in content
69        r = set()
70        for p in pat.content:
71            for x in p:
72                r.update(_get_head_types(x))
73        return r
74
75    raise Exception("Oh no! I don't understand pattern %s" %(pat))
76
77
78def _get_headnode_dict(fixer_list):
79    """ Accepts a list of fixers and returns a dictionary
80        of head node type --> fixer list.  """
81    head_nodes = collections.defaultdict(list)
82    every = []
83    for fixer in fixer_list:
84        if fixer.pattern:
85            try:
86                heads = _get_head_types(fixer.pattern)
87            except _EveryNode:
88                every.append(fixer)
89            else:
90                for node_type in heads:
91                    head_nodes[node_type].append(fixer)
92        else:
93            if fixer._accept_type is not None:
94                head_nodes[fixer._accept_type].append(fixer)
95            else:
96                every.append(fixer)
97    for node_type in chain(pygram.python_grammar.symbol2number.itervalues(),
98                           pygram.python_grammar.tokens):
99        head_nodes[node_type].extend(every)
100    return dict(head_nodes)
101
102
103def get_fixers_from_package(pkg_name):
104    """
105    Return the fully qualified names for fixers in the package pkg_name.
106    """
107    return [pkg_name + "." + fix_name
108            for fix_name in get_all_fix_names(pkg_name, False)]
109
110def _identity(obj):
111    return obj
112
113if sys.version_info < (3, 0):
114    import codecs
115    _open_with_encoding = codecs.open
116    # codecs.open doesn't translate newlines sadly.
117    def _from_system_newlines(input):
118        return input.replace(u"\r\n", u"\n")
119    def _to_system_newlines(input):
120        if os.linesep != "\n":
121            return input.replace(u"\n", os.linesep)
122        else:
123            return input
124else:
125    _open_with_encoding = open
126    _from_system_newlines = _identity
127    _to_system_newlines = _identity
128
129
130def _detect_future_features(source):
131    have_docstring = False
132    gen = tokenize.generate_tokens(StringIO.StringIO(source).readline)
133    def advance():
134        tok = gen.next()
135        return tok[0], tok[1]
136    ignore = frozenset((token.NEWLINE, tokenize.NL, token.COMMENT))
137    features = set()
138    try:
139        while True:
140            tp, value = advance()
141            if tp in ignore:
142                continue
143            elif tp == token.STRING:
144                if have_docstring:
145                    break
146                have_docstring = True
147            elif tp == token.NAME and value == u"from":
148                tp, value = advance()
149                if tp != token.NAME or value != u"__future__":
150                    break
151                tp, value = advance()
152                if tp != token.NAME or value != u"import":
153                    break
154                tp, value = advance()
155                if tp == token.OP and value == u"(":
156                    tp, value = advance()
157                while tp == token.NAME:
158                    features.add(value)
159                    tp, value = advance()
160                    if tp != token.OP or value != u",":
161                        break
162                    tp, value = advance()
163            else:
164                break
165    except StopIteration:
166        pass
167    return frozenset(features)
168
169
170class FixerError(Exception):
171    """A fixer could not be loaded."""
172
173
174class RefactoringTool(object):
175
176    _default_options = {"print_function" : False,
177                        "write_unchanged_files" : False}
178
179    CLASS_PREFIX = "Fix" # The prefix for fixer classes
180    FILE_PREFIX = "fix_" # The prefix for modules with a fixer within
181
182    def __init__(self, fixer_names, options=None, explicit=None):
183        """Initializer.
184
185        Args:
186            fixer_names: a list of fixers to import
187            options: a dict with configuration.
188            explicit: a list of fixers to run even if they are explicit.
189        """
190        self.fixers = fixer_names
191        self.explicit = explicit or []
192        self.options = self._default_options.copy()
193        if options is not None:
194            self.options.update(options)
195        if self.options["print_function"]:
196            self.grammar = pygram.python_grammar_no_print_statement
197        else:
198            self.grammar = pygram.python_grammar
199        # When this is True, the refactor*() methods will call write_file() for
200        # files processed even if they were not changed during refactoring. If
201        # and only if the refactor method's write parameter was True.
202        self.write_unchanged_files = self.options.get("write_unchanged_files")
203        self.errors = []
204        self.logger = logging.getLogger("RefactoringTool")
205        self.fixer_log = []
206        self.wrote = False
207        self.driver = driver.Driver(self.grammar,
208                                    convert=pytree.convert,
209                                    logger=self.logger)
210        self.pre_order, self.post_order = self.get_fixers()
211
212
213        self.files = []  # List of files that were or should be modified
214
215        self.BM = bm.BottomMatcher()
216        self.bmi_pre_order = [] # Bottom Matcher incompatible fixers
217        self.bmi_post_order = []
218
219        for fixer in chain(self.post_order, self.pre_order):
220            if fixer.BM_compatible:
221                self.BM.add_fixer(fixer)
222                # remove fixers that will be handled by the bottom-up
223                # matcher
224            elif fixer in self.pre_order:
225                self.bmi_pre_order.append(fixer)
226            elif fixer in self.post_order:
227                self.bmi_post_order.append(fixer)
228
229        self.bmi_pre_order_heads = _get_headnode_dict(self.bmi_pre_order)
230        self.bmi_post_order_heads = _get_headnode_dict(self.bmi_post_order)
231
232
233
234    def get_fixers(self):
235        """Inspects the options to load the requested patterns and handlers.
236
237        Returns:
238          (pre_order, post_order), where pre_order is the list of fixers that
239          want a pre-order AST traversal, and post_order is the list that want
240          post-order traversal.
241        """
242        pre_order_fixers = []
243        post_order_fixers = []
244        for fix_mod_path in self.fixers:
245            mod = __import__(fix_mod_path, {}, {}, ["*"])
246            fix_name = fix_mod_path.rsplit(".", 1)[-1]
247            if fix_name.startswith(self.FILE_PREFIX):
248                fix_name = fix_name[len(self.FILE_PREFIX):]
249            parts = fix_name.split("_")
250            class_name = self.CLASS_PREFIX + "".join([p.title() for p in parts])
251            try:
252                fix_class = getattr(mod, class_name)
253            except AttributeError:
254                raise FixerError("Can't find %s.%s" % (fix_name, class_name))
255            fixer = fix_class(self.options, self.fixer_log)
256            if fixer.explicit and self.explicit is not True and \
257                    fix_mod_path not in self.explicit:
258                self.log_message("Skipping optional fixer: %s", fix_name)
259                continue
260
261            self.log_debug("Adding transformation: %s", fix_name)
262            if fixer.order == "pre":
263                pre_order_fixers.append(fixer)
264            elif fixer.order == "post":
265                post_order_fixers.append(fixer)
266            else:
267                raise FixerError("Illegal fixer order: %r" % fixer.order)
268
269        key_func = operator.attrgetter("run_order")
270        pre_order_fixers.sort(key=key_func)
271        post_order_fixers.sort(key=key_func)
272        return (pre_order_fixers, post_order_fixers)
273
274    def log_error(self, msg, *args, **kwds):
275        """Called when an error occurs."""
276        raise
277
278    def log_message(self, msg, *args):
279        """Hook to log a message."""
280        if args:
281            msg = msg % args
282        self.logger.info(msg)
283
284    def log_debug(self, msg, *args):
285        if args:
286            msg = msg % args
287        self.logger.debug(msg)
288
289    def print_output(self, old_text, new_text, filename, equal):
290        """Called with the old version, new version, and filename of a
291        refactored file."""
292        pass
293
294    def refactor(self, items, write=False, doctests_only=False):
295        """Refactor a list of files and directories."""
296
297        for dir_or_file in items:
298            if os.path.isdir(dir_or_file):
299                self.refactor_dir(dir_or_file, write, doctests_only)
300            else:
301                self.refactor_file(dir_or_file, write, doctests_only)
302
303    def refactor_dir(self, dir_name, write=False, doctests_only=False):
304        """Descends down a directory and refactor every Python file found.
305
306        Python files are assumed to have a .py extension.
307
308        Files and subdirectories starting with '.' are skipped.
309        """
310        py_ext = os.extsep + "py"
311        for dirpath, dirnames, filenames in os.walk(dir_name):
312            self.log_debug("Descending into %s", dirpath)
313            dirnames.sort()
314            filenames.sort()
315            for name in filenames:
316                if (not name.startswith(".") and
317                    os.path.splitext(name)[1] == py_ext):
318                    fullname = os.path.join(dirpath, name)
319                    self.refactor_file(fullname, write, doctests_only)
320            # Modify dirnames in-place to remove subdirs with leading dots
321            dirnames[:] = [dn for dn in dirnames if not dn.startswith(".")]
322
323    def _read_python_source(self, filename):
324        """
325        Do our best to decode a Python source file correctly.
326        """
327        try:
328            f = open(filename, "rb")
329        except IOError as err:
330            self.log_error("Can't open %s: %s", filename, err)
331            return None, None
332        try:
333            encoding = tokenize.detect_encoding(f.readline)[0]
334        finally:
335            f.close()
336        with _open_with_encoding(filename, "r", encoding=encoding) as f:
337            return _from_system_newlines(f.read()), encoding
338
339    def refactor_file(self, filename, write=False, doctests_only=False):
340        """Refactors a file."""
341        input, encoding = self._read_python_source(filename)
342        if input is None:
343            # Reading the file failed.
344            return
345        input += u"\n" # Silence certain parse errors
346        if doctests_only:
347            self.log_debug("Refactoring doctests in %s", filename)
348            output = self.refactor_docstring(input, filename)
349            if self.write_unchanged_files or output != input:
350                self.processed_file(output, filename, input, write, encoding)
351            else:
352                self.log_debug("No doctest changes in %s", filename)
353        else:
354            tree = self.refactor_string(input, filename)
355            if self.write_unchanged_files or (tree and tree.was_changed):
356                # The [:-1] is to take off the \n we added earlier
357                self.processed_file(unicode(tree)[:-1], filename,
358                                    write=write, encoding=encoding)
359            else:
360                self.log_debug("No changes in %s", filename)
361
362    def refactor_string(self, data, name):
363        """Refactor a given input string.
364
365        Args:
366            data: a string holding the code to be refactored.
367            name: a human-readable name for use in error/log messages.
368
369        Returns:
370            An AST corresponding to the refactored input stream; None if
371            there were errors during the parse.
372        """
373        features = _detect_future_features(data)
374        if "print_function" in features:
375            self.driver.grammar = pygram.python_grammar_no_print_statement
376        try:
377            tree = self.driver.parse_string(data)
378        except Exception as err:
379            self.log_error("Can't parse %s: %s: %s",
380                           name, err.__class__.__name__, err)
381            return
382        finally:
383            self.driver.grammar = self.grammar
384        tree.future_features = features
385        self.log_debug("Refactoring %s", name)
386        self.refactor_tree(tree, name)
387        return tree
388
389    def refactor_stdin(self, doctests_only=False):
390        input = sys.stdin.read()
391        if doctests_only:
392            self.log_debug("Refactoring doctests in stdin")
393            output = self.refactor_docstring(input, "<stdin>")
394            if self.write_unchanged_files or output != input:
395                self.processed_file(output, "<stdin>", input)
396            else:
397                self.log_debug("No doctest changes in stdin")
398        else:
399            tree = self.refactor_string(input, "<stdin>")
400            if self.write_unchanged_files or (tree and tree.was_changed):
401                self.processed_file(unicode(tree), "<stdin>", input)
402            else:
403                self.log_debug("No changes in stdin")
404
405    def refactor_tree(self, tree, name):
406        """Refactors a parse tree (modifying the tree in place).
407
408        For compatible patterns the bottom matcher module is
409        used. Otherwise the tree is traversed node-to-node for
410        matches.
411
412        Args:
413            tree: a pytree.Node instance representing the root of the tree
414                  to be refactored.
415            name: a human-readable name for this tree.
416
417        Returns:
418            True if the tree was modified, False otherwise.
419        """
420
421        for fixer in chain(self.pre_order, self.post_order):
422            fixer.start_tree(tree, name)
423
424        #use traditional matching for the incompatible fixers
425        self.traverse_by(self.bmi_pre_order_heads, tree.pre_order())
426        self.traverse_by(self.bmi_post_order_heads, tree.post_order())
427
428        # obtain a set of candidate nodes
429        match_set = self.BM.run(tree.leaves())
430
431        while any(match_set.values()):
432            for fixer in self.BM.fixers:
433                if fixer in match_set and match_set[fixer]:
434                    #sort by depth; apply fixers from bottom(of the AST) to top
435                    match_set[fixer].sort(key=pytree.Base.depth, reverse=True)
436
437                    if fixer.keep_line_order:
438                        #some fixers(eg fix_imports) must be applied
439                        #with the original file's line order
440                        match_set[fixer].sort(key=pytree.Base.get_lineno)
441
442                    for node in list(match_set[fixer]):
443                        if node in match_set[fixer]:
444                            match_set[fixer].remove(node)
445
446                        try:
447                            find_root(node)
448                        except ValueError:
449                            # this node has been cut off from a
450                            # previous transformation ; skip
451                            continue
452
453                        if node.fixers_applied and fixer in node.fixers_applied:
454                            # do not apply the same fixer again
455                            continue
456
457                        results = fixer.match(node)
458
459                        if results:
460                            new = fixer.transform(node, results)
461                            if new is not None:
462                                node.replace(new)
463                                #new.fixers_applied.append(fixer)
464                                for node in new.post_order():
465                                    # do not apply the fixer again to
466                                    # this or any subnode
467                                    if not node.fixers_applied:
468                                        node.fixers_applied = []
469                                    node.fixers_applied.append(fixer)
470
471                                # update the original match set for
472                                # the added code
473                                new_matches = self.BM.run(new.leaves())
474                                for fxr in new_matches:
475                                    if not fxr in match_set:
476                                        match_set[fxr]=[]
477
478                                    match_set[fxr].extend(new_matches[fxr])
479
480        for fixer in chain(self.pre_order, self.post_order):
481            fixer.finish_tree(tree, name)
482        return tree.was_changed
483
484    def traverse_by(self, fixers, traversal):
485        """Traverse an AST, applying a set of fixers to each node.
486
487        This is a helper method for refactor_tree().
488
489        Args:
490            fixers: a list of fixer instances.
491            traversal: a generator that yields AST nodes.
492
493        Returns:
494            None
495        """
496        if not fixers:
497            return
498        for node in traversal:
499            for fixer in fixers[node.type]:
500                results = fixer.match(node)
501                if results:
502                    new = fixer.transform(node, results)
503                    if new is not None:
504                        node.replace(new)
505                        node = new
506
507    def processed_file(self, new_text, filename, old_text=None, write=False,
508                       encoding=None):
509        """
510        Called when a file has been refactored and there may be changes.
511        """
512        self.files.append(filename)
513        if old_text is None:
514            old_text = self._read_python_source(filename)[0]
515            if old_text is None:
516                return
517        equal = old_text == new_text
518        self.print_output(old_text, new_text, filename, equal)
519        if equal:
520            self.log_debug("No changes to %s", filename)
521            if not self.write_unchanged_files:
522                return
523        if write:
524            self.write_file(new_text, filename, old_text, encoding)
525        else:
526            self.log_debug("Not writing changes to %s", filename)
527
528    def write_file(self, new_text, filename, old_text, encoding=None):
529        """Writes a string to a file.
530
531        It first shows a unified diff between the old text and the new text, and
532        then rewrites the file; the latter is only done if the write option is
533        set.
534        """
535        try:
536            f = _open_with_encoding(filename, "w", encoding=encoding)
537        except os.error as err:
538            self.log_error("Can't create %s: %s", filename, err)
539            return
540        try:
541            f.write(_to_system_newlines(new_text))
542        except os.error as err:
543            self.log_error("Can't write %s: %s", filename, err)
544        finally:
545            f.close()
546        self.log_debug("Wrote changes to %s", filename)
547        self.wrote = True
548
549    PS1 = ">>> "
550    PS2 = "... "
551
552    def refactor_docstring(self, input, filename):
553        """Refactors a docstring, looking for doctests.
554
555        This returns a modified version of the input string.  It looks
556        for doctests, which start with a ">>>" prompt, and may be
557        continued with "..." prompts, as long as the "..." is indented
558        the same as the ">>>".
559
560        (Unfortunately we can't use the doctest module's parser,
561        since, like most parsers, it is not geared towards preserving
562        the original source.)
563        """
564        result = []
565        block = None
566        block_lineno = None
567        indent = None
568        lineno = 0
569        for line in input.splitlines(True):
570            lineno += 1
571            if line.lstrip().startswith(self.PS1):
572                if block is not None:
573                    result.extend(self.refactor_doctest(block, block_lineno,
574                                                        indent, filename))
575                block_lineno = lineno
576                block = [line]
577                i = line.find(self.PS1)
578                indent = line[:i]
579            elif (indent is not None and
580                  (line.startswith(indent + self.PS2) or
581                   line == indent + self.PS2.rstrip() + u"\n")):
582                block.append(line)
583            else:
584                if block is not None:
585                    result.extend(self.refactor_doctest(block, block_lineno,
586                                                        indent, filename))
587                block = None
588                indent = None
589                result.append(line)
590        if block is not None:
591            result.extend(self.refactor_doctest(block, block_lineno,
592                                                indent, filename))
593        return u"".join(result)
594
595    def refactor_doctest(self, block, lineno, indent, filename):
596        """Refactors one doctest.
597
598        A doctest is given as a block of lines, the first of which starts
599        with ">>>" (possibly indented), while the remaining lines start
600        with "..." (identically indented).
601
602        """
603        try:
604            tree = self.parse_block(block, lineno, indent)
605        except Exception as err:
606            if self.logger.isEnabledFor(logging.DEBUG):
607                for line in block:
608                    self.log_debug("Source: %s", line.rstrip(u"\n"))
609            self.log_error("Can't parse docstring in %s line %s: %s: %s",
610                           filename, lineno, err.__class__.__name__, err)
611            return block
612        if self.refactor_tree(tree, filename):
613            new = unicode(tree).splitlines(True)
614            # Undo the adjustment of the line numbers in wrap_toks() below.
615            clipped, new = new[:lineno-1], new[lineno-1:]
616            assert clipped == [u"\n"] * (lineno-1), clipped
617            if not new[-1].endswith(u"\n"):
618                new[-1] += u"\n"
619            block = [indent + self.PS1 + new.pop(0)]
620            if new:
621                block += [indent + self.PS2 + line for line in new]
622        return block
623
624    def summarize(self):
625        if self.wrote:
626            were = "were"
627        else:
628            were = "need to be"
629        if not self.files:
630            self.log_message("No files %s modified.", were)
631        else:
632            self.log_message("Files that %s modified:", were)
633            for file in self.files:
634                self.log_message(file)
635        if self.fixer_log:
636            self.log_message("Warnings/messages while refactoring:")
637            for message in self.fixer_log:
638                self.log_message(message)
639        if self.errors:
640            if len(self.errors) == 1:
641                self.log_message("There was 1 error:")
642            else:
643                self.log_message("There were %d errors:", len(self.errors))
644            for msg, args, kwds in self.errors:
645                self.log_message(msg, *args, **kwds)
646
647    def parse_block(self, block, lineno, indent):
648        """Parses a block into a tree.
649
650        This is necessary to get correct line number / offset information
651        in the parser diagnostics and embedded into the parse tree.
652        """
653        tree = self.driver.parse_tokens(self.wrap_toks(block, lineno, indent))
654        tree.future_features = frozenset()
655        return tree
656
657    def wrap_toks(self, block, lineno, indent):
658        """Wraps a tokenize stream to systematically modify start/end."""
659        tokens = tokenize.generate_tokens(self.gen_lines(block, indent).next)
660        for type, value, (line0, col0), (line1, col1), line_text in tokens:
661            line0 += lineno - 1
662            line1 += lineno - 1
663            # Don't bother updating the columns; this is too complicated
664            # since line_text would also have to be updated and it would
665            # still break for tokens spanning lines.  Let the user guess
666            # that the column numbers for doctests are relative to the
667            # end of the prompt string (PS1 or PS2).
668            yield type, value, (line0, col0), (line1, col1), line_text
669
670
671    def gen_lines(self, block, indent):
672        """Generates lines as expected by tokenize from a list of lines.
673
674        This strips the first len(indent + self.PS1) characters off each line.
675        """
676        prefix1 = indent + self.PS1
677        prefix2 = indent + self.PS2
678        prefix = prefix1
679        for line in block:
680            if line.startswith(prefix):
681                yield line[len(prefix):]
682            elif line == prefix.rstrip() + u"\n":
683                yield u"\n"
684            else:
685                raise AssertionError("line=%r, prefix=%r" % (line, prefix))
686            prefix = prefix2
687        while True:
688            yield ""
689
690
691class MultiprocessingUnsupported(Exception):
692    pass
693
694
695class MultiprocessRefactoringTool(RefactoringTool):
696
697    def __init__(self, *args, **kwargs):
698        super(MultiprocessRefactoringTool, self).__init__(*args, **kwargs)
699        self.queue = None
700        self.output_lock = None
701
702    def refactor(self, items, write=False, doctests_only=False,
703                 num_processes=1):
704        if num_processes == 1:
705            return super(MultiprocessRefactoringTool, self).refactor(
706                items, write, doctests_only)
707        try:
708            import multiprocessing
709        except ImportError:
710            raise MultiprocessingUnsupported
711        if self.queue is not None:
712            raise RuntimeError("already doing multiple processes")
713        self.queue = multiprocessing.JoinableQueue()
714        self.output_lock = multiprocessing.Lock()
715        processes = [multiprocessing.Process(target=self._child)
716                     for i in xrange(num_processes)]
717        try:
718            for p in processes:
719                p.start()
720            super(MultiprocessRefactoringTool, self).refactor(items, write,
721                                                              doctests_only)
722        finally:
723            self.queue.join()
724            for i in xrange(num_processes):
725                self.queue.put(None)
726            for p in processes:
727                if p.is_alive():
728                    p.join()
729            self.queue = None
730
731    def _child(self):
732        task = self.queue.get()
733        while task is not None:
734            args, kwargs = task
735            try:
736                super(MultiprocessRefactoringTool, self).refactor_file(
737                    *args, **kwargs)
738            finally:
739                self.queue.task_done()
740            task = self.queue.get()
741
742    def refactor_file(self, *args, **kwargs):
743        if self.queue is not None:
744            self.queue.put((args, kwargs))
745        else:
746            return super(MultiprocessRefactoringTool, self).refactor_file(
747                *args, **kwargs)
748