'`' testlist1 '`' |
NAME | NUMBER | STRING+ | '.' '.' '.')
listmaker: (test|star_expr) ( comp_for | (',' (test|star_expr))* [','] )
-testlist_gexp: test ( comp_for | (',' (test|star_expr))* [','] )
+testlist_gexp: (test|star_expr) ( comp_for | (',' (test|star_expr))* [','] )
lambdef: 'lambda' [varargslist] ':' test
trailer: '(' [arglist] ')' | '[' subscriptlist ']' | '.' NAME
subscriptlist: subscript (',' subscript)* [',']
--- /dev/null
+"""A bottom-up tree matching algorithm implementation meant to speed
+up 2to3's matching process. After the tree patterns are reduced to
+their rarest linear path, a linear Aho-Corasick automaton is
+created. The linear automaton traverses the linear paths from the
+leaves to the root of the AST and returns a set of nodes for further
+matching. This reduces significantly the number of candidate nodes."""
+
+__author__ = "George Boutsioukis <gboutsioukis@gmail.com>"
+
+import logging
+import itertools
+from collections import defaultdict
+
+from . import pytree
+from .btm_utils import reduce_tree
+
+class BMNode(object):
+ """Class for a node of the Aho-Corasick automaton used in matching"""
+ count = itertools.count()
+ def __init__(self):
+ self.transition_table = {}
+ self.fixers = []
+ self.id = next(BMNode.count)
+ self.content = ''
+
+class BottomMatcher(object):
+ """The main matcher class. After instantiating the patterns should
+ be added using the add_fixer method"""
+
+ def __init__(self):
+ self.match = set()
+ self.root = BMNode()
+ self.nodes = [self.root]
+ self.fixers = []
+ self.logger = logging.getLogger("RefactoringTool")
+
+ def add_fixer(self, fixer):
+ """Reduces a fixer's pattern tree to a linear path and adds it
+ to the matcher(a common Aho-Corasick automaton). The fixer is
+ appended on the matching states and called when they are
+ reached"""
+ self.fixers.append(fixer)
+ tree = reduce_tree(fixer.pattern_tree)
+ linear = tree.get_linear_subpattern()
+ match_nodes = self.add(linear, start=self.root)
+ for match_node in match_nodes:
+ match_node.fixers.append(fixer)
+
+ def add(self, pattern, start):
+ "Recursively adds a linear pattern to the AC automaton"
+ #print("adding pattern", pattern, "to", start)
+ if not pattern:
+ #print("empty pattern")
+ return [start]
+ if isinstance(pattern[0], tuple):
+ #alternatives
+ #print("alternatives")
+ match_nodes = []
+ for alternative in pattern[0]:
+ #add all alternatives, and add the rest of the pattern
+ #to each end node
+ end_nodes = self.add(alternative, start=start)
+ for end in end_nodes:
+ match_nodes.extend(self.add(pattern[1:], end))
+ return match_nodes
+ else:
+ #single token
+ #not last
+ if pattern[0] not in start.transition_table:
+ #transition did not exist, create new
+ next_node = BMNode()
+ start.transition_table[pattern[0]] = next_node
+ else:
+ #transition exists already, follow
+ next_node = start.transition_table[pattern[0]]
+
+ if pattern[1:]:
+ end_nodes = self.add(pattern[1:], start=next_node)
+ else:
+ end_nodes = [next_node]
+ return end_nodes
+
+ def run(self, leaves):
+ """The main interface with the bottom matcher. The tree is
+ traversed from the bottom using the constructed
+ automaton. Nodes are only checked once as the tree is
+ retraversed. When the automaton fails, we give it one more
+ shot(in case the above tree matches as a whole with the
+ rejected leaf), then we break for the next leaf. There is the
+ special case of multiple arguments(see code comments) where we
+ recheck the nodes
+
+ Args:
+ The leaves of the AST tree to be matched
+
+ Returns:
+ A dictionary of node matches with fixers as the keys
+ """
+ current_ac_node = self.root
+ results = defaultdict(list)
+ for leaf in leaves:
+ current_ast_node = leaf
+ while current_ast_node:
+ current_ast_node.was_checked = True
+ for child in current_ast_node.children:
+ # multiple statements, recheck
+ if isinstance(child, pytree.Leaf) and child.value == u";":
+ current_ast_node.was_checked = False
+ break
+ if current_ast_node.type == 1:
+ #name
+ node_token = current_ast_node.value
+ else:
+ node_token = current_ast_node.type
+
+ if node_token in current_ac_node.transition_table:
+ #token matches
+ current_ac_node = current_ac_node.transition_table[node_token]
+ for fixer in current_ac_node.fixers:
+ if not fixer in results:
+ results[fixer] = []
+ results[fixer].append(current_ast_node)
+
+ else:
+ #matching failed, reset automaton
+ current_ac_node = self.root
+ if (current_ast_node.parent is not None
+ and current_ast_node.parent.was_checked):
+ #the rest of the tree upwards has been checked, next leaf
+ break
+
+ #recheck the rejected node once from the root
+ if node_token in current_ac_node.transition_table:
+ #token matches
+ current_ac_node = current_ac_node.transition_table[node_token]
+ for fixer in current_ac_node.fixers:
+ if not fixer in results.keys():
+ results[fixer] = []
+ results[fixer].append(current_ast_node)
+
+ current_ast_node = current_ast_node.parent
+ return results
+
+ def print_ac(self):
+ "Prints a graphviz diagram of the BM automaton(for debugging)"
+ print("digraph g{")
+ def print_node(node):
+ for subnode_key in node.transition_table.keys():
+ subnode = node.transition_table[subnode_key]
+ print("%d -> %d [label=%s] //%s" %
+ (node.id, subnode.id, type_repr(subnode_key), str(subnode.fixers)))
+ if subnode_key == 1:
+ print(subnode.content)
+ print_node(subnode)
+ print_node(self.root)
+ print("}")
+
+# taken from pytree.py for debugging; only used by print_ac
+_type_reprs = {}
+def type_repr(type_num):
+ global _type_reprs
+ if not _type_reprs:
+ from .pygram import python_symbols
+ # printing tokens is possible but not as useful
+ # from .pgen2 import token // token.__dict__.items():
+ for name, val in python_symbols.__dict__.items():
+ if type(val) == int: _type_reprs[val] = name
+ return _type_reprs.setdefault(type_num, type_num)
--- /dev/null
+"Utility functions used by the btm_matcher module"
+
+from . import pytree
+from .pgen2 import grammar, token
+from .pygram import pattern_symbols, python_symbols
+
+syms = pattern_symbols
+pysyms = python_symbols
+tokens = grammar.opmap
+token_labels = token
+
+TYPE_ANY = -1
+TYPE_ALTERNATIVES = -2
+TYPE_GROUP = -3
+
+class MinNode(object):
+ """This class serves as an intermediate representation of the
+ pattern tree during the conversion to sets of leaf-to-root
+ subpatterns"""
+
+ def __init__(self, type=None, name=None):
+ self.type = type
+ self.name = name
+ self.children = []
+ self.leaf = False
+ self.parent = None
+ self.alternatives = []
+ self.group = []
+
+ def __repr__(self):
+ return str(self.type) + ' ' + str(self.name)
+
+ def leaf_to_root(self):
+ """Internal method. Returns a characteristic path of the
+ pattern tree. This method must be run for all leaves until the
+ linear subpatterns are merged into a single"""
+ node = self
+ subp = []
+ while node:
+ if node.type == TYPE_ALTERNATIVES:
+ node.alternatives.append(subp)
+ if len(node.alternatives) == len(node.children):
+ #last alternative
+ subp = [tuple(node.alternatives)]
+ node.alternatives = []
+ node = node.parent
+ continue
+ else:
+ node = node.parent
+ subp = None
+ break
+
+ if node.type == TYPE_GROUP:
+ node.group.append(subp)
+ #probably should check the number of leaves
+ if len(node.group) == len(node.children):
+ subp = get_characteristic_subpattern(node.group)
+ node.group = []
+ node = node.parent
+ continue
+ else:
+ node = node.parent
+ subp = None
+ break
+
+ if node.type == token_labels.NAME and node.name:
+ #in case of type=name, use the name instead
+ subp.append(node.name)
+ else:
+ subp.append(node.type)
+
+ node = node.parent
+ return subp
+
+ def get_linear_subpattern(self):
+ """Drives the leaf_to_root method. The reason that
+ leaf_to_root must be run multiple times is because we need to
+ reject 'group' matches; for example the alternative form
+ (a | b c) creates a group [b c] that needs to be matched. Since
+ matching multiple linear patterns overcomes the automaton's
+ capabilities, leaf_to_root merges each group into a single
+ choice based on 'characteristic'ity,
+
+ i.e. (a|b c) -> (a|b) if b more characteristic than c
+
+ Returns: The most 'characteristic'(as defined by
+ get_characteristic_subpattern) path for the compiled pattern
+ tree.
+ """
+
+ for l in self.leaves():
+ subp = l.leaf_to_root()
+ if subp:
+ return subp
+
+ def leaves(self):
+ "Generator that returns the leaves of the tree"
+ for child in self.children:
+ for x in child.leaves():
+ yield x
+ if not self.children:
+ yield self
+
+def reduce_tree(node, parent=None):
+ """
+ Internal function. Reduces a compiled pattern tree to an
+ intermediate representation suitable for feeding the
+ automaton. This also trims off any optional pattern elements(like
+ [a], a*).
+ """
+
+ new_node = None
+ #switch on the node type
+ if node.type == syms.Matcher:
+ #skip
+ node = node.children[0]
+
+ if node.type == syms.Alternatives :
+ #2 cases
+ if len(node.children) <= 2:
+ #just a single 'Alternative', skip this node
+ new_node = reduce_tree(node.children[0], parent)
+ else:
+ #real alternatives
+ new_node = MinNode(type=TYPE_ALTERNATIVES)
+ #skip odd children('|' tokens)
+ for child in node.children:
+ if node.children.index(child)%2:
+ continue
+ reduced = reduce_tree(child, new_node)
+ if reduced is not None:
+ new_node.children.append(reduced)
+ elif node.type == syms.Alternative:
+ if len(node.children) > 1:
+
+ new_node = MinNode(type=TYPE_GROUP)
+ for child in node.children:
+ reduced = reduce_tree(child, new_node)
+ if reduced:
+ new_node.children.append(reduced)
+ if not new_node.children:
+ # delete the group if all of the children were reduced to None
+ new_node = None
+
+ else:
+ new_node = reduce_tree(node.children[0], parent)
+
+ elif node.type == syms.Unit:
+ if (isinstance(node.children[0], pytree.Leaf) and
+ node.children[0].value == '('):
+ #skip parentheses
+ return reduce_tree(node.children[1], parent)
+ if ((isinstance(node.children[0], pytree.Leaf) and
+ node.children[0].value == '[')
+ or
+ (len(node.children)>1 and
+ hasattr(node.children[1], "value") and
+ node.children[1].value == '[')):
+ #skip whole unit if its optional
+ return None
+
+ leaf = True
+ details_node = None
+ alternatives_node = None
+ has_repeater = False
+ repeater_node = None
+ has_variable_name = False
+
+ for child in node.children:
+ if child.type == syms.Details:
+ leaf = False
+ details_node = child
+ elif child.type == syms.Repeater:
+ has_repeater = True
+ repeater_node = child
+ elif child.type == syms.Alternatives:
+ alternatives_node = child
+ if hasattr(child, 'value') and child.value == '=': # variable name
+ has_variable_name = True
+
+ #skip variable name
+ if has_variable_name:
+ #skip variable name, '='
+ name_leaf = node.children[2]
+ if hasattr(name_leaf, 'value') and name_leaf.value == '(':
+ # skip parenthesis
+ name_leaf = node.children[3]
+ else:
+ name_leaf = node.children[0]
+
+ #set node type
+ if name_leaf.type == token_labels.NAME:
+ #(python) non-name or wildcard
+ if name_leaf.value == 'any':
+ new_node = MinNode(type=TYPE_ANY)
+ else:
+ if hasattr(token_labels, name_leaf.value):
+ new_node = MinNode(type=getattr(token_labels, name_leaf.value))
+ else:
+ new_node = MinNode(type=getattr(pysyms, name_leaf.value))
+
+ elif name_leaf.type == token_labels.STRING:
+ #(python) name or character; remove the apostrophes from
+ #the string value
+ name = name_leaf.value.strip("'")
+ if name in tokens:
+ new_node = MinNode(type=tokens[name])
+ else:
+ new_node = MinNode(type=token_labels.NAME, name=name)
+ elif name_leaf.type == syms.Alternatives:
+ new_node = reduce_tree(alternatives_node, parent)
+
+ #handle repeaters
+ if has_repeater:
+ if repeater_node.children[0].value == '*':
+ #reduce to None
+ new_node = None
+ elif repeater_node.children[0].value == '+':
+ #reduce to a single occurence i.e. do nothing
+ pass
+ else:
+ #TODO: handle {min, max} repeaters
+ raise NotImplementedError
+ pass
+
+ #add children
+ if details_node and new_node is not None:
+ for child in details_node.children[1:-1]:
+ #skip '<', '>' markers
+ reduced = reduce_tree(child, new_node)
+ if reduced is not None:
+ new_node.children.append(reduced)
+ if new_node:
+ new_node.parent = parent
+ return new_node
+
+
+def get_characteristic_subpattern(subpatterns):
+ """Picks the most characteristic from a list of linear patterns
+ Current order used is:
+ names > common_names > common_chars
+ """
+ if not isinstance(subpatterns, list):
+ return subpatterns
+ if len(subpatterns)==1:
+ return subpatterns[0]
+
+ # first pick out the ones containing variable names
+ subpatterns_with_names = []
+ subpatterns_with_common_names = []
+ common_names = ['in', 'for', 'if' , 'not', 'None']
+ subpatterns_with_common_chars = []
+ common_chars = "[]().,:"
+ for subpattern in subpatterns:
+ if any(rec_test(subpattern, lambda x: type(x) is str)):
+ if any(rec_test(subpattern,
+ lambda x: isinstance(x, str) and x in common_chars)):
+ subpatterns_with_common_chars.append(subpattern)
+ elif any(rec_test(subpattern,
+ lambda x: isinstance(x, str) and x in common_names)):
+ subpatterns_with_common_names.append(subpattern)
+
+ else:
+ subpatterns_with_names.append(subpattern)
+
+ if subpatterns_with_names:
+ subpatterns = subpatterns_with_names
+ elif subpatterns_with_common_names:
+ subpatterns = subpatterns_with_common_names
+ elif subpatterns_with_common_chars:
+ subpatterns = subpatterns_with_common_chars
+ # of the remaining subpatterns pick out the longest one
+ return max(subpatterns, key=len)
+
+def rec_test(sequence, test_func):
+ """Tests test_func on all items of sequence and items of included
+ sub-iterables"""
+ for x in sequence:
+ if isinstance(x, (list, tuple)):
+ for y in rec_test(x, test_func):
+ yield y
+ else:
+ yield test_func(x)
PATTERN = None # Most subclasses should override with a string literal
pattern = None # Compiled pattern, set by compile_pattern()
+ pattern_tree = None # Tree representation of the pattern
options = None # Options object passed to initializer
filename = None # The filename (set by set_filename)
logger = None # A logger (set by set_filename)
_accept_type = None # [Advanced and not public] This tells RefactoringTool
# which node type to accept when there's not a pattern.
+ keep_line_order = False # For the bottom matcher: match with the
+ # original line order
+ BM_compatible = False # Compatibility with the bottom matching
+ # module; every fixer should set this
+ # manually
+
# Shortcut for access to Python grammar symbols
syms = pygram.python_symbols
self.{pattern,PATTERN} in .match().
"""
if self.PATTERN is not None:
- self.pattern = PatternCompiler().compile_pattern(self.PATTERN)
+ PC = PatternCompiler()
+ self.pattern, self.pattern_tree = PC.compile_pattern(self.PATTERN,
+ with_tree=True)
def set_filename(self, filename):
"""Set the filename, and a logger derived from it.
""" Works like `does_tree_import` but adds an import statement
if it was not imported. """
def is_import_stmt(node):
- return node.type == syms.simple_stmt and node.children and \
- is_import(node.children[0])
+ return (node.type == syms.simple_stmt and node.children and
+ is_import(node.children[0]))
root = find_root(node)
# if that also fails, we stick to the beginning of the file
if insert_pos == 0:
for idx, node in enumerate(root.children):
- if node.type == syms.simple_stmt and node.children and \
- node.children[0].type == token.STRING:
+ if (node.type == syms.simple_stmt and node.children and
+ node.children[0].type == token.STRING):
insert_pos = idx + 1
break
from ..fixer_util import Call, Comma, parenthesize
class FixApply(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
power< 'apply'
from ..fixer_util import Name
class FixBasestring(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = "'basestring'"
class FixBuffer(fixer_base.BaseFix):
+ BM_compatible = True
explicit = True # The user must ask for this fixer
from lib2to3.fixer_util import Call, Name, String, Attr, touch_import
class FixCallable(fixer_base.BaseFix):
+ BM_compatible = True
+
+ order = "pre"
# Ignore callable(*args) or use of keywords.
# Either could be a hint that the builtin callable() is not being used.
class FixDict(fixer_base.BaseFix):
+ BM_compatible = True
+
PATTERN = """
power< head=any+
trailer< '.' method=('keys'|'items'|'values'|
yield (n, nodes[i+2])
class FixExcept(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
try_stmt< 'try' ':' (simple_stmt | suite)
class FixExec(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
exec_stmt< 'exec' a=any 'in' b=any [',' c=any] >
class FixExecfile(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
power< 'execfile' trailer< '(' arglist< filename=any [',' globals=any [',' locals=any ] ] > ')' > >
class FixExitfunc(fixer_base.BaseFix):
+ keep_line_order = True
+ BM_compatible = True
PATTERN = """
(
from ..fixer_util import Name, Call, ListComp, in_special_context
class FixFilter(fixer_base.ConditionalFix):
+ BM_compatible = True
PATTERN = """
filter_lambda=power<
class FixFuncattrs(fixer_base.BaseFix):
+ BM_compatible = True
+
PATTERN = """
power< any+ trailer< '.' attr=('func_closure' | 'func_doc' | 'func_globals'
| 'func_name' | 'func_defaults' | 'func_code'
from ..fixer_util import BlankLine
class FixFuture(fixer_base.BaseFix):
+ BM_compatible = True
+
PATTERN = """import_from< 'from' module_name="__future__" 'import' any >"""
# This should be run last -- some things check for the import
from ..fixer_util import Name
class FixGetcwdu(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
power< 'os' trailer< dot='.' name='getcwdu' > any* >
class FixHasKey(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
anchor=power<
TYPE = "power< 'type' trailer< '(' x=any ')' > >"
class FixIdioms(fixer_base.BaseFix):
-
explicit = True # The user must ask for this fixer
PATTERN = r"""
class FixImport(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
import_from< 'from' imp=any 'import' ['('] any [')'] >
class FixImports(fixer_base.BaseFix):
+ BM_compatible = True
+ keep_line_order = True
# This is overridden in fix_imports2.
mapping = MAPPING
class FixInput(fixer_base.BaseFix):
-
+ BM_compatible = True
PATTERN = """
power< 'input' args=trailer< '(' [any] ')' > >
"""
class FixIntern(fixer_base.BaseFix):
+ BM_compatible = True
+ order = "pre"
PATTERN = """
power< 'intern'
class FixIsinstance(fixer_base.BaseFix):
-
+ BM_compatible = True
PATTERN = """
power<
'isinstance'
from ..fixer_util import Name
class FixItertools(fixer_base.BaseFix):
+ BM_compatible = True
it_funcs = "('imap'|'ifilter'|'izip'|'ifilterfalse')"
PATTERN = """
power< it='itertools'
class FixItertoolsImports(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
import_from< 'from' 'itertools' 'import' imports=any >
""" %(locals())
class FixLong(fixer_base.BaseFix):
-
+ BM_compatible = True
PATTERN = "'long'"
def transform(self, node, results):
from ..pygram import python_symbols as syms
class FixMap(fixer_base.ConditionalFix):
+ BM_compatible = True
PATTERN = """
map_none=power<
class FixMetaclass(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
classdef<any*>
}
class FixMethodattrs(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
power< any+ trailer< '.' attr=('im_func' | 'im_self' | 'im_class') > any* >
"""
class FixNext(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
power< base=any+ trailer< '.' attr='next' > trailer< '(' ')' > >
|
from ..fixer_util import Name, syms
class FixNonzero(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
classdef< 'class' any+ ':'
suite< any*
from lib2to3.fixer_util import Call, Name, String, touch_import
+def invocation(s):
+ def dec(f):
+ f.invocation = s
+ return f
+ return dec
+
+
class FixOperator(fixer_base.BaseFix):
+ BM_compatible = True
+ order = "pre"
methods = """
method=('isCallable'|'sequenceIncludes'
if method is not None:
return method(node, results)
+ @invocation("operator.contains(%s)")
def _sequenceIncludes(self, node, results):
- """operator.contains(%s)"""
return self._handle_rename(node, results, u"contains")
+ @invocation("hasattr(%s, '__call__')")
def _isCallable(self, node, results):
- """hasattr(%s, '__call__')"""
obj = results["obj"]
args = [obj.clone(), String(u", "), String(u"'__call__'")]
return Call(Name(u"hasattr"), args, prefix=node.prefix)
+ @invocation("operator.mul(%s)")
def _repeat(self, node, results):
- """operator.mul(%s)"""
return self._handle_rename(node, results, u"mul")
+ @invocation("operator.imul(%s)")
def _irepeat(self, node, results):
- """operator.imul(%s)"""
return self._handle_rename(node, results, u"imul")
+ @invocation("isinstance(%s, collections.Sequence)")
def _isSequenceType(self, node, results):
- """isinstance(%s, collections.Sequence)"""
return self._handle_type2abc(node, results, u"collections", u"Sequence")
+ @invocation("isinstance(%s, collections.Mapping)")
def _isMappingType(self, node, results):
- """isinstance(%s, collections.Mapping)"""
return self._handle_type2abc(node, results, u"collections", u"Mapping")
+ @invocation("isinstance(%s, numbers.Number)")
def _isNumberType(self, node, results):
- """isinstance(%s, numbers.Number)"""
return self._handle_type2abc(node, results, u"numbers", u"Number")
def _handle_rename(self, node, results, name):
return method
else:
sub = (unicode(results["obj"]),)
- invocation_str = unicode(method.__doc__) % sub
+ invocation_str = unicode(method.invocation) % sub
self.warning(node, u"You should use '%s' here." % invocation_str)
return None
# XXX This doesn't support nested for loops like [x for x in 1, 2 for x in 1, 2]
class FixParen(fixer_base.BaseFix):
+ BM_compatible = True
+
PATTERN = """
atom< ('[' | '(')
(listmaker< any
class FixPrint(fixer_base.BaseFix):
+ BM_compatible = True
+
PATTERN = """
simple_stmt< any* bare='print' any* > | print_stmt
"""
raise E -> raise E
raise E, V -> raise E(V)
raise E, V, T -> raise E(V).with_traceback(T)
+raise E, None, T -> raise E.with_traceback(T)
raise (((E, E'), E''), E'''), V -> raise E(V)
raise "foo", V, T -> warns about string exceptions
class FixRaise(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
raise_stmt< 'raise' exc=any [',' val=any [',' tb=any]] >
"""
syms = self.syms
exc = results["exc"].clone()
- if exc.type is token.STRING:
- self.cannot_convert(node, "Python 3 does not support string exceptions")
+ if exc.type == token.STRING:
+ msg = "Python 3 does not support string exceptions"
+ self.cannot_convert(node, msg)
return
# Python 2 supports
# exc.children[1:-1] is the unparenthesized tuple
# exc.children[1].children[0] is the first element of the tuple
exc = exc.children[1].children[0].clone()
- exc.prefix = " "
+ exc.prefix = u" "
if "val" not in results:
# One-argument raise
tb = results["tb"].clone()
tb.prefix = u""
- e = Call(exc, args)
+ e = exc
+ # If there's a traceback and None is passed as the value, then don't
+ # add a call, since the user probably just wants to add a
+ # traceback. See issue #9661.
+ if val.type != token.NAME or val.value != u"None":
+ e = Call(exc, args)
with_tb = Attr(e, Name(u'with_traceback')) + [ArgList([tb])]
new = pytree.Node(syms.simple_stmt, [Name(u"raise")] + with_tb)
new.prefix = node.prefix
class FixRawInput(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
power< name='raw_input' trailer< '(' [any] ')' > any* >
"""
class FixReduce(fixer_base.BaseFix):
+ BM_compatible = True
+ order = "pre"
+
PATTERN = """
power< 'reduce'
trailer< '('
class FixRenames(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = "|".join(build_pattern())
order = "pre" # Pre-order tree traversal
class FixRepr(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
atom < '`' expr=any '`' >
"""
class FixSetLiteral(fixer_base.BaseFix):
+ BM_compatible = True
explicit = True
PATTERN = """power< 'set' trailer< '('
class FixStandarderror(fixer_base.BaseFix):
-
+ BM_compatible = True
PATTERN = """
'StandardError'
"""
class FixSysExc(fixer_base.BaseFix):
# This order matches the ordering of sys.exc_info().
exc_info = [u"exc_type", u"exc_value", u"exc_traceback"]
+ BM_compatible = True
PATTERN = """
power< 'sys' trailer< dot='.' attribute=(%s) > >
""" % '|'.join("'%s'" % e for e in exc_info)
from ..fixer_util import Name, Call, ArgList, Attr, is_tuple
class FixThrow(fixer_base.BaseFix):
-
+ BM_compatible = True
PATTERN = """
power< any trailer< '.' 'throw' >
trailer< '(' args=arglist< exc=any ',' val=any [',' tb=any] > ')' >
stmt.children[0].type == token.STRING
class FixTupleParams(fixer_base.BaseFix):
+ run_order = 4 #use a lower order since lambda is part of other
+ #patterns
+ BM_compatible = True
+
PATTERN = """
funcdef< 'def' any parameters< '(' args=any ')' >
['->' any] ':' suite=any+ >
_pats = ["power< 'types' trailer< '.' name='%s' > >" % t for t in _TYPE_MAPPING]
class FixTypes(fixer_base.BaseFix):
-
+ BM_compatible = True
PATTERN = '|'.join(_pats)
def transform(self, node, results):
_literal_re = re.compile(ur"[uU][rR]?[\'\"]")
class FixUnicode(fixer_base.BaseFix):
-
+ BM_compatible = True
PATTERN = "STRING | 'unicode' | 'unichr'"
def transform(self, node, results):
from lib2to3.fixes.fix_imports import alternates, FixImports
from lib2to3 import fixer_base
from lib2to3.fixer_util import (Name, Comma, FromImport, Newline,
- find_indentation)
+ find_indentation, Node, syms)
MAPPING = {"urllib": [
("urllib.request",
mod_dict = {}
members = results["members"]
for member in members:
- member = member.value
# we only care about the actual members
- if member != u",":
+ if member.type == syms.import_as_name:
+ as_name = member.children[2].value
+ member_name = member.children[0].value
+ else:
+ member_name = member.value
+ as_name = None
+ if member_name != u",":
for change in MAPPING[mod_member.value]:
- if member in change[1]:
- if change[0] in mod_dict:
- mod_dict[change[0]].append(member)
- else:
- mod_dict[change[0]] = [member]
+ if member_name in change[1]:
+ if change[0] not in mod_dict:
modules.append(change[0])
+ mod_dict.setdefault(change[0], []).append(member)
new_nodes = []
indentation = find_indentation(node)
first = True
+ def handle_name(name, prefix):
+ if name.type == syms.import_as_name:
+ kids = [Name(name.children[0].value, prefix=prefix),
+ name.children[1].clone(),
+ name.children[2].clone()]
+ return [Node(syms.import_as_name, kids)]
+ return [Name(name.value, prefix=prefix)]
for module in modules:
elts = mod_dict[module]
names = []
for elt in elts[:-1]:
- names.extend([Name(elt, prefix=pref), Comma()])
- names.append(Name(elts[-1], prefix=pref))
+ names.extend(handle_name(elt, pref))
+ names.append(Comma())
+ names.extend(handle_name(elts[-1], pref))
new = FromImport(module, names)
if not first or node.parent.prefix.endswith(indentation):
new.prefix = indentation
class FixXrange(fixer_base.BaseFix):
-
+ BM_compatible = True
PATTERN = """
power<
(name='range'|name='xrange') trailer< '(' args=any ')' >
class FixXreadlines(fixer_base.BaseFix):
+ BM_compatible = True
PATTERN = """
power< call=any+ trailer< '.' 'xreadlines' > trailer< '(' ')' > >
|
class FixZip(fixer_base.ConditionalFix):
+ BM_compatible = True
PATTERN = """
power< 'zip' args=trailer< '(' [any] ')' >
>
self.pysyms = pygram.python_symbols
self.driver = driver.Driver(self.grammar, convert=pattern_convert)
- def compile_pattern(self, input, debug=False):
+ 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, e:
+ except parse.ParseError as e:
raise PatternSyntaxError(str(e))
- return self.compile_node(root)
+ if with_tree:
+ return self.compile_node(root), root
+ else:
+ return self.compile_node(root)
def compile_node(self, node):
"""Compiles a node, recursively.
# The grammar file
_GRAMMAR_FILE = os.path.join(os.path.dirname(__file__), "Grammar.txt")
+_PATTERN_GRAMMAR_FILE = os.path.join(os.path.dirname(__file__),
+ "PatternGrammar.txt")
class Symbols(object):
python_grammar_no_print_statement = python_grammar.copy()
del python_grammar_no_print_statement.keywords["print"]
+
+pattern_grammar = driver.load_grammar(_PATTERN_GRAMMAR_FILE)
+pattern_symbols = Symbols(pattern_grammar)
import warnings
from StringIO import StringIO
-
HUGE = 0x7FFFFFFF # maximum repeat count, default max
_type_reprs = {}
if type(val) == int: _type_reprs[val] = name
return _type_reprs.setdefault(type_num, type_num)
-
class Base(object):
"""
parent = None # Parent node pointer, or None
children = () # Tuple of subnodes
was_changed = False
+ was_checked = False
def __new__(cls, *args, **kwds):
"""Constructor that prevents Base from being instantiated."""
return None
return self.parent.children[i-1]
+ def leaves(self):
+ for child in self.children:
+ for x in child.leaves():
+ yield x
+
+ def depth(self):
+ if self.parent is None:
+ return 0
+ return 1 + self.parent.depth()
+
def get_suffix(self):
"""
Return the string immediately following the invocant node. This is
def __str__(self):
return unicode(self).encode("ascii")
-
class Node(Base):
"""Concrete implementation for interior nodes."""
- def __init__(self, type, children, context=None, prefix=None):
+ def __init__(self,type, children,
+ context=None,
+ prefix=None,
+ fixers_applied=None):
"""
Initializer.
ch.parent = self
if prefix is not None:
self.prefix = prefix
+ if fixers_applied:
+ self.fixers_applied = fixers_applied[:]
+ else:
+ self.fixers_applied = None
def __repr__(self):
"""Return a canonical string representation."""
def clone(self):
"""Return a cloned (deep) copy of self."""
- return Node(self.type, [ch.clone() for ch in self.children])
+ return Node(self.type, [ch.clone() for ch in self.children],
+ fixers_applied=self.fixers_applied)
def post_order(self):
"""Return a post-order iterator for the tree."""
lineno = 0 # Line where this token starts in the input
column = 0 # Column where this token tarts in the input
- def __init__(self, type, value, context=None, prefix=None):
+ def __init__(self, type, value,
+ context=None,
+ prefix=None,
+ fixers_applied=[]):
"""
Initializer.
self.value = value
if prefix is not None:
self._prefix = prefix
+ self.fixers_applied = fixers_applied[:]
def __repr__(self):
"""Return a canonical string representation."""
def clone(self):
"""Return a cloned (deep) copy of self."""
return Leaf(self.type, self.value,
- (self.prefix, (self.lineno, self.column)))
+ (self.prefix, (self.lineno, self.column)),
+ fixers_applied=self.fixers_applied)
+
+ def leaves(self):
+ yield self
def post_order(self):
"""Return a post-order iterator for the tree."""
# Local imports
from .pgen2 import driver, tokenize, token
+from .fixer_util import find_root
from . import pytree, pygram
+from . import btm_utils as bu
+from . import btm_matcher as bm
def get_all_fix_names(fixer_pkg, remove_prefix=True):
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.files = [] # List of files that were or should be modified
+ self.BM = bm.BottomMatcher()
+ self.bmi_pre_order = [] # Bottom Matcher incompatible fixers
+ self.bmi_post_order = []
+
+ for fixer in chain(self.post_order, self.pre_order):
+ if fixer.BM_compatible:
+ self.BM.add_fixer(fixer)
+ # remove fixers that will be handled by the bottom-up
+ # matcher
+ elif fixer in self.pre_order:
+ self.bmi_pre_order.append(fixer)
+ elif fixer in self.post_order:
+ self.bmi_post_order.append(fixer)
+
+ self.bmi_pre_order_heads = _get_headnode_dict(self.bmi_pre_order)
+ self.bmi_post_order_heads = _get_headnode_dict(self.bmi_post_order)
+
+
+
def get_fixers(self):
"""Inspects the options to load the requested patterns and handlers.
def refactor(self, items, write=False, doctests_only=False):
"""Refactor a list of files and directories."""
+
for dir_or_file in items:
if os.path.isdir(dir_or_file):
self.refactor_dir(dir_or_file, write, doctests_only)
"""
try:
f = open(filename, "rb")
- except IOError, err:
+ except IOError as err:
self.log_error("Can't open %s: %s", filename, err)
return None, None
try:
self.driver.grammar = pygram.python_grammar_no_print_statement
try:
tree = self.driver.parse_string(data)
- except Exception, err:
+ except Exception as err:
self.log_error("Can't parse %s: %s: %s",
name, err.__class__.__name__, err)
return
def refactor_tree(self, tree, name):
"""Refactors a parse tree (modifying the tree in place).
+ For compatible patterns the bottom matcher module is
+ used. Otherwise the tree is traversed node-to-node for
+ matches.
+
Args:
tree: a pytree.Node instance representing the root of the tree
to be refactored.
Returns:
True if the tree was modified, False otherwise.
"""
+
for fixer in chain(self.pre_order, self.post_order):
fixer.start_tree(tree, name)
- self.traverse_by(self.pre_order_heads, tree.pre_order())
- self.traverse_by(self.post_order_heads, tree.post_order())
+ #use traditional matching for the incompatible fixers
+ self.traverse_by(self.bmi_pre_order_heads, tree.pre_order())
+ self.traverse_by(self.bmi_post_order_heads, tree.post_order())
+
+ # obtain a set of candidate nodes
+ match_set = self.BM.run(tree.leaves())
+
+ while any(match_set.values()):
+ for fixer in self.BM.fixers:
+ if fixer in match_set and match_set[fixer]:
+ #sort by depth; apply fixers from bottom(of the AST) to top
+ match_set[fixer].sort(key=pytree.Base.depth, reverse=True)
+
+ if fixer.keep_line_order:
+ #some fixers(eg fix_imports) must be applied
+ #with the original file's line order
+ match_set[fixer].sort(key=pytree.Base.get_lineno)
+
+ for node in list(match_set[fixer]):
+ if node in match_set[fixer]:
+ match_set[fixer].remove(node)
+
+ try:
+ find_root(node)
+ except AssertionError:
+ # this node has been cut off from a
+ # previous transformation ; skip
+ continue
+
+ if node.fixers_applied and fixer in node.fixers_applied:
+ # do not apply the same fixer again
+ continue
+
+ results = fixer.match(node)
+
+ if results:
+ new = fixer.transform(node, results)
+ if new is not None:
+ node.replace(new)
+ #new.fixers_applied.append(fixer)
+ for node in new.post_order():
+ # do not apply the fixer again to
+ # this or any subnode
+ if not node.fixers_applied:
+ node.fixers_applied = []
+ node.fixers_applied.append(fixer)
+
+ # update the original match set for
+ # the added code
+ new_matches = self.BM.run(new.leaves())
+ for fxr in new_matches:
+ if not fxr in match_set:
+ match_set[fxr]=[]
+
+ match_set[fxr].extend(new_matches[fxr])
for fixer in chain(self.pre_order, self.post_order):
fixer.finish_tree(tree, name)
"""
try:
f = _open_with_encoding(filename, "w", encoding=encoding)
- except os.error, err:
+ except os.error as err:
self.log_error("Can't create %s: %s", filename, err)
return
try:
f.write(_to_system_newlines(new_text))
- except os.error, err:
+ except os.error as err:
self.log_error("Can't write %s: %s", filename, err)
finally:
f.close()
"""
try:
tree = self.parse_block(block, lineno, indent)
- except Exception, err:
+ except Exception as err:
if self.logger.isEnabledFor(logging.DEBUG):
for line in block:
self.log_debug("Source: %s", line.rstrip(u"\n"))
# coding: utf-8
print "BOM BOOM!"
-
raise Exception(5).with_traceback(6) # foo"""
self.check(b, a)
+ def test_None_value(self):
+ b = """raise Exception(5), None, tb"""
+ a = """raise Exception(5).with_traceback(tb)"""
+ self.check(b, a)
+
def test_tuple_value(self):
b = """raise Exception, (5, 6, 7)"""
a = """raise Exception(5, 6, 7)"""
b = "from %s import %s as foo_bar" % (old, member)
a = "from %s import %s as foo_bar" % (new, member)
self.check(b, a)
+ b = "from %s import %s as blah, %s" % (old, member, member)
+ a = "from %s import %s as blah, %s" % (new, member, member)
+ self.check(b, a)
def test_star(self):
for old in self.modules:
self.assertEqual(str(n1), "foo**bar")
self.assertTrue(isinstance(n1.children, list))
+ def test_leaves(self):
+ l1 = pytree.Leaf(100, "foo")
+ l2 = pytree.Leaf(100, "bar")
+ l3 = pytree.Leaf(100, "fooey")
+ n2 = pytree.Node(1000, [l1, l2])
+ n3 = pytree.Node(1000, [l3])
+ n1 = pytree.Node(1000, [n2, n3])
+
+ self.assertEqual(list(n1.leaves()), [l1, l2, l3])
+
+ def test_depth(self):
+ l1 = pytree.Leaf(100, "foo")
+ l2 = pytree.Leaf(100, "bar")
+ n2 = pytree.Node(1000, [l1, l2])
+ n3 = pytree.Node(1000, [])
+ n1 = pytree.Node(1000, [n2, n3])
+
+ self.assertEqual(l1.depth(), 2)
+ self.assertEqual(n3.depth(), 1)
+ self.assertEqual(n1.depth(), 0)
+
def test_post_order(self):
l1 = pytree.Leaf(100, "foo")
l2 = pytree.Leaf(100, "bar")
projects / python / commitdiff