homes/common/.local/lib/python2.7/site-packages/jedi/parsing.py
Christophe Buffenoir 8385ed00af Add jedi
Fix error in powerline (ascii can't decode)
Get vim plugins directly from git
2013-05-03 10:58:55 +02:00

1855 lines
66 KiB
Python

"""
Maintainer: David Halter <davidhalter88@gmail.com>
Version: 0.1
py_fuzzyparser parses python code, with the goal of a good representation of
the code within a tree structure. Variables, Classes and Functions are defined
within this tree structure, containing their exact locations in the code.
It is also a primary goal to work with code which contains syntax errors.
This behaviour may be used to refactor, modify, search and complete code.
To understand this code it is extremely important to understand the behaviour
of the python module 'tokenize'.
This original codebase of this parser, which has been refactored and heavily
changed, was programmed by Aaron Griffin <aaronmgriffin@gmail.com>.
**The structure of the following script:**
A Scope has
- imports (Import)
- subscopes (Scope, Class, Function, Flow)
- statements (Statement)
All these objects have `Name`s. `Call` and `Array` are used as detail objects
of a statement.
All those classes are being generated by PyFuzzyParser, which takes python text
as input and ignores just all the non-python stuff. Basically you could feed it
a perl script, and it should still work (which means throw no error.
"""
from _compatibility import (next, literal_eval, StringIO,
property, cleandoc, Python3Method)
import tokenize
import re
import keyword
import os
import debug
class ParserError(Exception):
pass
def indent_block(text, indention=' '):
""" This function indents a text block with a default of four spaces """
temp = ''
while text and text[-1] == '\n':
temp += text[-1]
text = text[:-1]
lines = text.split('\n')
return '\n'.join(map(lambda s: indention + s, lines)) + temp
class PushBackIterator(object):
def __init__(self, iterator):
self.pushes = []
self.iterator = iterator
def push_back(self, value):
self.pushes.append(value)
def __iter__(self):
return self
def next(self):
""" Python 2 Compatibility """
return self.__next__()
def __next__(self):
if self.pushes:
return self.pushes.pop()
else:
return next(self.iterator)
class Base(object):
"""
This is just here to have an isinstance check, which is also used on
evaluate classes. But since they have sometimes a special type of
delegation, it is important for those classes to override this method.
I know that there is a chance to do such things with __instancecheck__, but
since Python 2.5 doesn't support it, I decided to do it this way.
"""
def isinstance(self, *cls):
return isinstance(self, cls)
class Simple(Base):
"""
The super class for Scope, Import, Name and Statement. Every object in
the parser tree inherits from this class.
"""
def __init__(self, start_pos, end_pos=(None, None)):
self.start_pos = start_pos
self.end_pos = end_pos
self.parent = None
@Python3Method
def get_parent_until(self, classes=(), reverse=False,
include_current=False):
""" Takes always the parent, until one class (not a Class) """
if type(classes) not in (tuple, list):
classes = (classes,)
scope = self
while scope.parent is not None:
if classes and reverse != scope.isinstance(*classes):
if include_current:
return scope
break
scope = scope.parent
return scope
def __repr__(self):
code = self.get_code().replace('\n', ' ')
return "<%s: %s@%s>" % \
(type(self).__name__, code, self.start_pos[0])
class Scope(Simple):
"""
Super class for the parser tree, which represents the state of a python
text file.
A Scope manages and owns its subscopes, which are classes and functions, as
well as variables and imports. It is used to access the structure of python
files.
:param start_pos: The position (line and column) of the scope.
:type start_pos: tuple(int, int)
:param docstr: The docstring for the current Scope.
:type docstr: str
"""
def __init__(self, start_pos, docstr=''):
super(Scope, self).__init__(start_pos)
self.subscopes = []
self.imports = []
self.statements = []
self.docstr = docstr
self.asserts = []
def add_scope(self, sub, decorators):
sub.parent = self
sub.decorators = decorators
for d in decorators:
# the parent is the same, because the decorator has not the scope
# of the function
d.parent = sub.parent
self.subscopes.append(sub)
return sub
def add_statement(self, stmt):
"""
Used to add a Statement or a Scope.
A statement would be a normal command (Statement) or a Scope (Flow).
"""
stmt.parent = self
self.statements.append(stmt)
return stmt
def add_docstr(self, string):
""" Clean up a docstring """
self.docstr = cleandoc(literal_eval(string))
def add_import(self, imp):
self.imports.append(imp)
imp.parent = self
def get_imports(self):
""" Gets also the imports within flow statements """
i = [] + self.imports
for s in self.statements:
if isinstance(s, Scope):
i += s.get_imports()
return i
def get_code(self, first_indent=False, indention=' '):
"""
:return: Returns the code of the current scope.
:rtype: str
"""
string = ""
if len(self.docstr) > 0:
string += '"""' + self.docstr + '"""\n'
for i in self.imports:
string += i.get_code()
for sub in self.subscopes:
string += sub.get_code(first_indent=True, indention=indention)
for stmt in self.statements:
string += stmt.get_code()
if first_indent:
string = indent_block(string, indention=indention)
return string
@Python3Method
def get_set_vars(self):
"""
Get all the names, that are active and accessible in the current
scope.
:return: list of Name
:rtype: list
"""
n = []
for stmt in self.statements:
try:
n += stmt.get_set_vars(True)
except TypeError:
n += stmt.get_set_vars()
# function and class names
n += [s.name for s in self.subscopes]
for i in self.imports:
if not i.star:
n += i.get_defined_names()
return n
def get_defined_names(self):
return [n for n in self.get_set_vars()
if isinstance(n, Import) or len(n) == 1]
def is_empty(self):
"""
:return: True if there are no subscopes, imports and statements.
:rtype: bool
"""
return not (self.imports or self.subscopes or self.statements)
@Python3Method
def get_statement_for_position(self, pos):
checks = self.statements + self.asserts
if self.isinstance(Function):
checks += self.params + self.decorators + self.returns
for s in checks:
if isinstance(s, Flow):
p = s.get_statement_for_position(pos)
while s.next and not p:
s = s.next
p = s.get_statement_for_position(pos)
if p:
return p
elif s.start_pos <= pos < s.end_pos:
return s
for s in self.subscopes:
if s.start_pos <= pos <= s.end_pos:
p = s.get_statement_for_position(pos)
if p:
return p
def __repr__(self):
try:
name = self.path
except AttributeError:
try:
name = self.name
except AttributeError:
name = self.command
return "<%s: %s@%s-%s>" % (type(self).__name__, name,
self.start_pos[0], self.end_pos[0])
class Module(Scope):
"""
The top scope, which is always a module.
"""
def __init__(self, path, docstr=''):
super(Module, self).__init__((1, 0), docstr)
self.path = path
self.global_vars = []
self._name = None
self.used_names = {}
self.temp_used_names = []
def add_global(self, name):
"""
Global means in these context a function (subscope) which has a global
statement.
This is only relevant for the top scope.
:param name: The name of the global.
:type name: Name
"""
self.global_vars.append(name)
# set no parent here, because globals are not defined in this scope.
def get_set_vars(self):
n = super(Module, self).get_set_vars()
n += self.global_vars
return n
@property
def name(self):
""" This is used for the goto function. """
if self._name is not None:
return self._name
if self.path is None:
string = '' # no path -> empty name
else:
sep = (re.escape(os.path.sep),) * 2
r = re.search(r'([^%s]*?)(%s__init__)?(\.py|\.so)?$' % sep,
self.path)
string = r.group(1)
names = [(string, (0, 0))]
self._name = Name(names, self.start_pos, self.end_pos, self)
return self._name
def is_builtin(self):
return not (self.path is None or self.path.endswith('.py'))
class Class(Scope):
"""
Used to store the parsed contents of a python class.
:param name: The Class name.
:type name: string
:param supers: The super classes of a Class.
:type supers: list
:param start_pos: The start position (line, column) of the class.
:type start_pos: tuple(int, int)
:param docstr: The docstring for the current Scope.
:type docstr: str
"""
def __init__(self, name, supers, start_pos, docstr=''):
super(Class, self).__init__(start_pos, docstr)
self.name = name
name.parent = self
self.supers = supers
for s in self.supers:
s.parent = self
self.decorators = []
def get_code(self, first_indent=False, indention=' '):
string = "\n".join('@' + stmt.get_code() for stmt in self.decorators)
string += 'class %s' % (self.name)
if len(self.supers) > 0:
sup = ','.join(stmt.code for stmt in self.supers)
string += '(%s)' % sup
string += ':\n'
string += super(Class, self).get_code(True, indention)
if self.is_empty():
string += "pass\n"
return string
class Function(Scope):
"""
Used to store the parsed contents of a python function.
:param name: The Function name.
:type name: string
:param params: The parameters (Statement) of a Function.
:type params: list
:param start_pos: The start position (line, column) the Function.
:type start_pos: tuple(int, int)
:param docstr: The docstring for the current Scope.
:type docstr: str
"""
def __init__(self, name, params, start_pos, annotation):
Scope.__init__(self, start_pos)
self.name = name
name.parent = self
self.params = params
for p in params:
p.parent = self
p.parent_function = self
self.decorators = []
self.returns = []
self.is_generator = False
self.listeners = set() # not used here, but in evaluation.
if annotation is not None:
annotation.parent = self
self.annotation = annotation
def get_code(self, first_indent=False, indention=' '):
string = "\n".join('@' + stmt.get_code() for stmt in self.decorators)
params = ','.join([stmt.code for stmt in self.params])
string += "def %s(%s):\n" % (self.name, params)
string += super(Function, self).get_code(True, indention)
if self.is_empty():
string += "pass\n"
return string
def get_set_vars(self):
n = super(Function, self).get_set_vars()
for p in self.params:
try:
n.append(p.get_name())
except IndexError:
debug.warning("multiple names in param %s" % n)
return n
def get_call_signature(self, width=72):
"""
Generate call signature of this function.
:param width: Fold lines if a line is longer than this value.
:type width: int
:rtype: str
"""
l = self.name.names[-1] + '('
lines = []
for (i, p) in enumerate(self.params):
code = p.get_code(False)
if i != len(self.params) - 1:
code += ', '
if len(l + code) > width:
lines.append(l[:-1] if l[-1] == ' ' else l)
l = code
else:
l += code
if l:
lines.append(l)
lines[-1] += ')'
return '\n'.join(lines)
@property
def doc(self):
""" Return a document string including call signature. """
return '%s\n\n%s' % (self.get_call_signature(), self.docstr)
class Flow(Scope):
"""
Used to describe programming structure - flow statements,
which indent code, but are not classes or functions:
- for
- while
- if
- try
- with
Therefore statements like else, except and finally are also here,
they are now saved in the root flow elements, but in the next variable.
:param command: The flow command, if, while, else, etc.
:type command: str
:param inits: The initializations of a flow -> while 'statement'.
:type inits: list(Statement)
:param start_pos: Position (line, column) of the Flow statement.
:type start_pos: tuple(int, int)
:param set_vars: Local variables used in the for loop (only there).
:type set_vars: list
"""
def __init__(self, command, inits, start_pos, set_vars=None):
self.next = None
self.command = command
super(Flow, self).__init__(start_pos, '')
self._parent = None
# These have to be statements, because of with, which takes multiple.
self.inits = inits
for s in inits:
s.parent = self
if set_vars is None:
self.set_vars = []
else:
self.set_vars = set_vars
for s in self.set_vars:
s.parent.parent = self
s.parent = self
@property
def parent(self):
return self._parent
@parent.setter
def parent(self, value):
self._parent = value
if self.next:
self.next.parent = value
def get_code(self, first_indent=False, indention=' '):
stmts = []
for s in self.inits:
stmts.append(s.get_code(new_line=False))
stmt = ', '.join(stmts)
string = "%s %s:\n" % (self.command, vars, stmt)
string += super(Flow, self).get_code(True, indention)
if self.next:
string += self.next.get_code()
return string
def get_set_vars(self, is_internal_call=False):
"""
Get the names for the flow. This includes also a call to the super
class.
:param is_internal_call: defines an option for internal files to crawl\
through this class. Normally it will just call its superiors, to\
generate the output.
"""
if is_internal_call:
n = list(self.set_vars)
for s in self.inits:
n += s.set_vars
if self.next:
n += self.next.get_set_vars(is_internal_call)
n += super(Flow, self).get_set_vars()
return n
else:
return self.get_parent_until((Class, Function)).get_set_vars()
def get_imports(self):
i = super(Flow, self).get_imports()
if self.next:
i += self.next.get_imports()
return i
def set_next(self, next):
""" Set the next element in the flow, those are else, except, etc. """
if self.next:
return self.next.set_next(next)
else:
self.next = next
self.next.parent = self.parent
return next
class ForFlow(Flow):
"""
Used for the for loop, because there are two statement parts.
"""
def __init__(self, inits, start_pos, set_stmt, is_list_comp=False):
super(ForFlow, self).__init__('for', inits, start_pos,
set_stmt.used_vars)
self.set_stmt = set_stmt
self.is_list_comp = is_list_comp
def get_code(self, first_indent=False, indention=" " * 4):
vars = ",".join(x.get_code() for x in self.set_vars)
stmts = []
for s in self.inits:
stmts.append(s.get_code(new_line=False))
stmt = ', '.join(stmts)
s = "for %s in %s:\n" % (vars, stmt)
return s + super(Flow, self).get_code(True, indention)
class Import(Simple):
"""
Stores the imports of any Scopes.
>>> 1+1
2
:param start_pos: Position (line, column) of the Import.
:type start_pos: tuple(int, int)
:param namespace: The import, can be empty if a star is given
:type namespace: Name
:param alias: The alias of a namespace(valid in the current namespace).
:type alias: Name
:param from_ns: Like the namespace, can be equally used.
:type from_ns: Name
:param star: If a star is used -> from time import *.
:type star: bool
:param defunct: An Import is valid or not.
:type defunct: bool
"""
def __init__(self, start_pos, end_pos, namespace, alias=None,
from_ns=None, star=False, relative_count=0, defunct=False):
super(Import, self).__init__(start_pos, end_pos)
self.namespace = namespace
self.alias = alias
self.from_ns = from_ns
for n in [namespace, alias, from_ns]:
if n:
n.parent = self
self.star = star
self.relative_count = relative_count
self.defunct = defunct
def get_code(self, new_line=True):
# in case one of the names is None
alias = self.alias or ''
namespace = self.namespace or ''
from_ns = self.from_ns or ''
if self.alias:
ns_str = "%s as %s" % (namespace, alias)
else:
ns_str = str(namespace)
nl = '\n' if new_line else ''
if self.from_ns or self.relative_count:
if self.star:
ns_str = '*'
dots = '.' * self.relative_count
return "from %s%s import %s%s" % (dots, from_ns, ns_str, nl)
else:
return "import %s%s" % (ns_str, nl)
def get_defined_names(self):
if self.defunct:
return []
if self.star:
return [self]
if self.alias:
return [self.alias]
if len(self.namespace) > 1:
o = self.namespace
n = Name([(o.names[0], o.start_pos)], o.start_pos, o.end_pos,
parent=o.parent)
return [n]
else:
return [self.namespace]
def get_set_vars(self):
return self.get_defined_names()
def get_all_import_names(self):
n = []
if self.from_ns:
n.append(self.from_ns)
if self.namespace:
n.append(self.namespace)
if self.alias:
n.append(self.alias)
return n
class Statement(Simple):
"""
This is the class for all the possible statements. Which means, this class
stores pretty much all the Python code, except functions, classes, imports,
and flow functions like if, for, etc.
:param code: The full code of a statement. This is import, if one wants \
to execute the code at some level.
:param code: str
:param set_vars: The variables which are defined by the statement.
:param set_vars: str
:param used_funcs: The functions which are used by the statement.
:param used_funcs: str
:param used_vars: The variables which are used by the statement.
:param used_vars: str
:param token_list: Token list which is also peppered with Name.
:param token_list: list
:param start_pos: Position (line, column) of the Statement.
:type start_pos: tuple(int, int)
"""
def __init__(self, code, set_vars, used_funcs, used_vars, token_list,
start_pos, end_pos):
super(Statement, self).__init__(start_pos, end_pos)
self.code = code
self.used_funcs = used_funcs
self.used_vars = used_vars
self.token_list = token_list
for s in set_vars + used_funcs + used_vars:
s.parent = self
self.set_vars = self._remove_executions_from_set_vars(set_vars)
# cache
self._assignment_calls = None
self._assignment_details = None
# this is important for other scripts
self._assignment_calls_calculated = False
def _remove_executions_from_set_vars(self, set_vars):
"""
Important mainly for assosiative arrays:
>>> a = 3
>>> b = {}
>>> b[a] = 3
`a` is in this case not a set_var, it is used to index the dict.
"""
if not set_vars:
return set_vars
result = set(set_vars)
last = None
in_execution = 0
for tok in self.token_list:
if isinstance(tok, Name):
if tok not in result:
break
if in_execution:
result.remove(tok)
elif isinstance(tok, tuple):
tok = tok[1]
if tok in ['(', '['] and isinstance(last, Name):
in_execution += 1
elif tok in [')', ']'] and in_execution > 0:
in_execution -= 1
last = tok
return list(result)
def get_code(self, new_line=True):
if new_line:
return self.code + '\n'
else:
return self.code
def get_set_vars(self):
""" Get the names for the statement. """
return list(self.set_vars)
@property
def assignment_details(self):
if self._assignment_details is None:
# normally, this calls sets this variable
self.get_assignment_calls()
# it may not have been set by get_assignment_calls -> just use an empty
# array
return self._assignment_details or []
def is_global(self):
# first keyword of the first token is global -> must be a global
return str(self.token_list[0]) == "global"
def get_assignment_calls(self):
"""
This is not done in the main parser, because it might be slow and
most of the statements won't need this data anyway. This is something
'like' a lazy execution.
This is not really nice written, sorry for that. If you plan to replace
it and make it nicer, that would be cool :-)
"""
if self._assignment_calls_calculated:
return self._assignment_calls
self._assignment_details = []
top = result = Array(self.start_pos, Array.NOARRAY, self)
level = 0
is_chain = False
close_brackets = False
tok_iter = enumerate(self.token_list)
for i, tok_temp in tok_iter:
#print 'tok', tok_temp, result
if isinstance(tok_temp, ListComprehension):
result.add_to_current_field(tok_temp)
continue
try:
token_type, tok, start_pos = tok_temp
except TypeError:
# the token is a Name, which has already been parsed
tok = tok_temp
token_type = None
start_pos = tok.start_pos
except ValueError:
debug.warning("unkown value, shouldn't happen",
tok_temp, type(tok_temp))
raise
else:
if tok in ['return', 'yield'] or level == 0 and \
tok.endswith('=') and not tok in ['>=', '<=', '==', '!=']:
# This means, there is an assignment here.
# Add assignments, which can be more than one
self._assignment_details.append((tok, top))
# All these calls wouldn't be important if nonlocal would
# exist. -> Initialize the first item again.
result = Array(start_pos, Array.NOARRAY, self)
top = result
level = 0
close_brackets = False
is_chain = False
continue
elif tok == 'as':
next(tok_iter, None)
continue
brackets = {'(': Array.TUPLE, '[': Array.LIST, '{': Array.SET}
is_call = lambda: type(result) == Call
is_call_or_close = lambda: is_call() or close_brackets
is_literal = token_type in [tokenize.STRING, tokenize.NUMBER]
if isinstance(tok, Name) or is_literal:
c_type = Call.NAME
if is_literal:
tok = literal_eval(tok)
if token_type == tokenize.STRING:
c_type = Call.STRING
elif token_type == tokenize.NUMBER:
c_type = Call.NUMBER
if is_chain:
call = Call(tok, c_type, start_pos, parent=result)
result = result.set_next_chain_call(call)
is_chain = False
close_brackets = False
else:
if close_brackets:
result = result.parent
close_brackets = False
if type(result) == Call:
result = result.parent
call = Call(tok, c_type, start_pos, parent=result)
result.add_to_current_field(call)
result = call
elif tok in brackets.keys(): # brackets
level += 1
if is_call_or_close():
result = Array(start_pos, brackets[tok], parent=result)
result = result.parent.add_execution(result)
close_brackets = False
else:
result = Array(start_pos, brackets[tok], parent=result)
result.parent.add_to_current_field(result)
elif tok == ':':
while is_call_or_close():
result = result.parent
close_brackets = False
if result.type == Array.LIST: # [:] lookups
result.add_to_current_field(tok)
else:
result.add_dictionary_key()
elif tok == '.':
if close_brackets and result.parent != top:
# only get out of the array, if it is a array execution
result = result.parent
close_brackets = False
is_chain = True
elif tok == ',':
while is_call_or_close():
result = result.parent
close_brackets = False
result.add_field((start_pos[0], start_pos[1] + 1))
# important - it cannot be empty anymore
if result.type == Array.NOARRAY:
result.type = Array.TUPLE
elif tok in [')', '}', ']']:
while is_call_or_close():
result = result.parent
close_brackets = False
if tok == '}' and not len(result):
# this is a really special case - empty brackets {} are
# always dictionaries and not sets.
result.type = Array.DICT
level -= 1
result.end_pos = start_pos[0], start_pos[1] + 1
close_brackets = True
else:
while is_call_or_close():
result = result.parent
close_brackets = False
if tok != '\n':
result.add_to_current_field(tok)
if level != 0:
debug.warning("Brackets don't match: %s."
"This is not normal behaviour." % level)
self._assignment_calls_calculated = True
self._assignment_calls = top
return top
class Param(Statement):
"""
The class which shows definitions of params of classes and functions.
But this is not to define function calls.
"""
def __init__(self, code, set_vars, used_funcs, used_vars,
token_list, start_pos, end_pos):
super(Param, self).__init__(code, set_vars, used_funcs, used_vars,
token_list, start_pos, end_pos)
# this is defined by the parser later on, not at the initialization
# it is the position in the call (first argument, second...)
self.position_nr = None
self.is_generated = False
self.annotation_stmt = None
self.parent_function = None
def add_annotation(self, annotation_stmt):
annotation_stmt.parent = self
self.annotation_stmt = annotation_stmt
def get_name(self):
""" get the name of the param """
n = self.set_vars or self.used_vars
if len(n) > 1:
debug.warning("Multiple param names (%s)." % n)
return n[0]
class Call(Base):
"""
`Call` contains a call, e.g. `foo.bar` and owns the executions of those
calls, which are `Array`s.
"""
NAME = 1
NUMBER = 2
STRING = 3
def __init__(self, name, type, start_pos, parent_stmt=None, parent=None):
self.name = name
# parent is not the oposite of next. The parent of c: a = [b.c] would
# be an array.
self.parent = parent
self.type = type
self.start_pos = start_pos
self.next = None
self.execution = None
self._parent_stmt = parent_stmt
@property
def parent_stmt(self):
if self._parent_stmt is not None:
return self._parent_stmt
elif self.parent:
return self.parent.parent_stmt
else:
return None
@parent_stmt.setter
def parent_stmt(self, value):
self._parent_stmt = value
def set_next_chain_call(self, call):
""" Adds another part of the statement"""
self.next = call
call.parent = self.parent
return call
def add_execution(self, call):
"""
An execution is nothing else than brackets, with params in them, which
shows access on the internals of this name.
"""
self.execution = call
# there might be multiple executions, like a()[0], in that case, they
# have the same parent. Otherwise it's not possible to parse proper.
if self.parent.execution == self:
call.parent = self.parent
else:
call.parent = self
return call
def generate_call_path(self):
""" Helps to get the order in which statements are executed. """
# TODO include previous nodes? As an option?
try:
for name_part in self.name.names:
yield name_part
except AttributeError:
yield self
if self.execution is not None:
for y in self.execution.generate_call_path():
yield y
if self.next is not None:
for y in self.next.generate_call_path():
yield y
def get_code(self):
if self.type == Call.NAME:
s = self.name.get_code()
else:
s = repr(self.name)
if self.execution is not None:
s += '(%s)' % self.execution.get_code()
if self.next is not None:
s += self.next.get_code()
return s
def __repr__(self):
return "<%s: %s>" % \
(type(self).__name__, self.name)
class Array(Call):
"""
Describes the different python types for an array, but also empty
statements. In the Python syntax definitions this type is named 'atom'.
http://docs.python.org/py3k/reference/grammar.html
Array saves sub-arrays as well as normal operators and calls to methods.
:param array_type: The type of an array, which can be one of the constants\
below.
:type array_type: int
"""
NOARRAY = None
TUPLE = 'tuple'
LIST = 'list'
DICT = 'dict'
SET = 'set'
def __init__(self, start_pos, arr_type=NOARRAY, parent_stmt=None,
parent=None, values=None):
super(Array, self).__init__(None, arr_type, start_pos, parent_stmt,
parent)
self.values = values if values else []
self.arr_el_pos = []
self.keys = []
self.end_pos = None
def add_field(self, start_pos):
"""
Just add a new field to the values.
Each value has a sub-array, because there may be different tokens in
one array.
"""
self.arr_el_pos.append(start_pos)
self.values.append([])
def add_to_current_field(self, tok):
""" Adds a token to the latest field (in content). """
if not self.values:
# An empty round brace is just a tuple, filled it is unnecessary.
if self.type == Array.TUPLE:
self.type = Array.NOARRAY
# Add the first field, this is done here, because if nothing
# gets added, the list is empty, which is also needed sometimes.
self.values.append([])
self.values[-1].append(tok)
def add_dictionary_key(self):
"""
Only used for dictionaries, automatically adds the tokens added by now
from the values to keys, because the parser works this way.
"""
if self.type in (Array.LIST, Array.TUPLE):
return # these are basically code errors, just ignore
self.keys.append(self.values.pop())
if self.type == Array.SET:
self.type = Array.DICT
self.values.append([])
def get_only_subelement(self):
"""
Returns the only element that an array contains. If it contains
more than one element, raise an exception.
"""
if len(self.values) != 1 or len(self.values[0]) != 1:
raise AttributeError("More than one value found")
return self.values[0][0]
@staticmethod
def is_type(instance, *types):
"""
This is not only used for calls on the actual object, but for
ducktyping, to invoke this function with anything as `self`.
"""
if isinstance(instance, Array):
if instance.type in types:
return True
return False
def __len__(self):
return len(self.values)
def __getitem__(self, key):
return self.values[key]
def __iter__(self):
if self.type == self.DICT:
return iter(zip(self.keys, self.values))
else:
return iter(self.values)
def get_code(self):
def to_str(el):
try:
return el.get_code()
except AttributeError:
return str(el)
map = {Array.NOARRAY: '%s',
Array.TUPLE: '(%s)',
Array.LIST: '[%s]',
Array.DICT: '{%s}',
Array.SET: '{%s}'
}
inner = []
for i, value in enumerate(self.values):
s = ''
try:
key = self.keys[i]
except IndexError:
pass
else:
for el in key[i]:
s += to_str(el)
for el in value:
s += to_str(el)
inner.append(s)
return map[self.type] % ', '.join(inner)
def __repr__(self):
if self.type == self.NOARRAY:
typ = 'noarray'
else:
typ = self.type
return "<%s: %s%s>" % (type(self).__name__, typ, self.values)
class NamePart(str):
"""
A string. Sometimes it is important to know if the string belongs to a name
or not.
"""
def __new__(cls, s, start_pos):
self = super(NamePart, cls).__new__(cls, s)
self.start_pos = start_pos
return self
@property
def end_pos(self):
return self.start_pos[0], self.start_pos[1] + len(self)
class Name(Simple):
"""
Used to define names in python.
Which means the whole namespace/class/function stuff.
So a name like "module.class.function"
would result in an array of [module, class, function]
"""
def __init__(self, names, start_pos, end_pos, parent=None):
super(Name, self).__init__(start_pos, end_pos)
self.names = tuple(n if isinstance(n, NamePart) else NamePart(*n)
for n in names)
if parent is not None:
self.parent = parent
def get_code(self):
""" Returns the names in a full string format """
return ".".join(self.names)
def __str__(self):
return self.get_code()
def __len__(self):
return len(self.names)
class ListComprehension(object):
""" Helper class for list comprehensions """
def __init__(self, stmt, middle, input):
self.stmt = stmt
self.middle = middle
self.input = input
def __repr__(self):
return "<%s: %s>" % \
(type(self).__name__, self.get_code())
def get_code(self):
statements = self.stmt, self.middle, self.input
code = [s.get_code().replace('\n', '') for s in statements]
return "%s for %s in %s" % tuple(code)
class PyFuzzyParser(object):
"""
This class is used to parse a Python file, it then divides them into a
class structure of different scopes.
:param code: The codebase for the parser.
:type code: str
:param user_position: The line/column, the user is currently on.
:type user_position: tuple(int, int)
"""
def __init__(self, code, module_path=None, user_position=None,
no_docstr=False, line_offset=0):
self.user_position = user_position
self.user_scope = None
self.user_stmt = None
self.code = code + '\n' # end with \n, because the parser needs it
self.no_docstr = no_docstr
# initialize global Scope
self.module = Module(module_path)
self.scope = self.module
self.current = (None, None)
# Stuff to fix tokenize errors. The parser is pretty good in tolerating
# any errors of tokenize and just parse ahead.
self._line_of_tokenize_restart = line_offset
self.parse()
# delete code again, only the parser needs it
del self.code
def __repr__(self):
return "<%s: %s>" % (type(self).__name__, self.module)
@property
def start_pos(self):
return (self._line_of_tokenize_restart + self._tokenize_start_pos[0],
self._tokenize_start_pos[1])
@property
def end_pos(self):
return (self._line_of_tokenize_restart + self._tokenize_end_pos[0],
self._tokenize_end_pos[1])
def _check_user_stmt(self, simple):
if not isinstance(simple, Param):
for tok_name in self.module.temp_used_names:
try:
self.module.used_names[tok_name].add(simple)
except KeyError:
self.module.used_names[tok_name] = set([simple])
self.module.temp_used_names = []
if not self.user_position:
return
# the position is right
if simple.start_pos <= self.user_position <= simple.end_pos:
if self.user_stmt is not None:
# if there is already a user position (another import, because
# imports are splitted) the names are checked.
for n in simple.get_set_vars():
if n.start_pos < self.user_position <= n.end_pos:
self.user_stmt = simple
else:
self.user_stmt = simple
def _parsedotname(self, pre_used_token=None):
"""
The dot name parser parses a name, variable or function and returns
their names.
:return: Tuple of Name, token_type, nexttoken.
:rtype: tuple(Name, int, str)
"""
def append(el):
names.append(el)
self.module.temp_used_names.append(el[0])
names = []
if pre_used_token is None:
token_type, tok = self.next()
if token_type != tokenize.NAME and tok != '*':
return [], token_type, tok
else:
token_type, tok = pre_used_token
if token_type != tokenize.NAME and tok != '*':
# token maybe a name or star
return None, token_type, tok
append((tok, self.start_pos))
first_pos = self.start_pos
while True:
token_type, tok = self.next()
if tok != '.':
break
token_type, tok = self.next()
if token_type != tokenize.NAME:
break
append((tok, self.start_pos))
n = Name(names, first_pos, self.end_pos) if names else None
return n, token_type, tok
def _parseimportlist(self):
"""
The parser for the imports. Unlike the class and function parse
function, this returns no Import class, but rather an import list,
which is then added later on.
The reason, why this is not done in the same class lies in the nature
of imports. There are two ways to write them:
- from ... import ...
- import ...
To distinguish, this has to be processed after the parser.
:return: List of imports.
:rtype: list
"""
imports = []
brackets = False
continue_kw = [",", ";", "\n", ')'] \
+ list(set(keyword.kwlist) - set(['as']))
while True:
defunct = False
token_type, tok = self.next()
if token_type == tokenize.ENDMARKER:
break
if brackets and tok == '\n':
self.next()
if tok == '(': # python allows only one `(` in the statement.
brackets = True
self.next()
i, token_type, tok = self._parsedotname(self.current)
if not i:
defunct = True
name2 = None
if tok == 'as':
name2, token_type, tok = self._parsedotname()
imports.append((i, name2, defunct))
while tok not in continue_kw:
token_type, tok = self.next()
if not (tok == "," or brackets and tok == '\n'):
break
return imports
def _parseparen(self):
"""
Functions and Classes have params (which means for classes
super-classes). They are parsed here and returned as Statements.
:return: List of Statements
:rtype: list
"""
names = []
tok = None
pos = 0
breaks = [',', ':']
while tok not in [')', ':']:
param, tok = self._parse_statement(added_breaks=breaks,
stmt_class=Param)
if param and tok == ':':
# parse annotations
annotation, tok = self._parse_statement(added_breaks=breaks)
if annotation:
param.add_annotation(annotation)
# params without vars are usually syntax errors.
if param and (param.set_vars or param.used_vars):
param.position_nr = pos
names.append(param)
pos += 1
return names
def _parsefunction(self):
"""
The parser for a text functions. Process the tokens, which follow a
function definition.
:return: Return a Scope representation of the tokens.
:rtype: Function
"""
first_pos = self.start_pos
token_type, fname = self.next()
if token_type != tokenize.NAME:
return None
fname = Name([(fname, self.start_pos)], self.start_pos, self.end_pos)
token_type, open = self.next()
if open != '(':
return None
params = self._parseparen()
token_type, colon = self.next()
annotation = None
if colon in ['-', '->']:
# parse annotations
if colon == '-':
# The Python 2 tokenizer doesn't understand this
token_type, colon = self.next()
if colon != '>':
return None
annotation, colon = self._parse_statement(added_breaks=[':'])
if colon != ':':
return None
# because of 2 line func param definitions
scope = Function(fname, params, first_pos, annotation)
if self.user_scope and scope != self.user_scope \
and self.user_position > first_pos:
self.user_scope = scope
return scope
def _parseclass(self):
"""
The parser for a text class. Process the tokens, which follow a
class definition.
:return: Return a Scope representation of the tokens.
:rtype: Class
"""
first_pos = self.start_pos
token_type, cname = self.next()
if token_type != tokenize.NAME:
debug.warning("class: syntax err, token is not a name@%s (%s: %s)"
% (self.start_pos[0], tokenize.tok_name[token_type], cname))
return None
cname = Name([(cname, self.start_pos)], self.start_pos, self.end_pos)
super = []
token_type, next = self.next()
if next == '(':
super = self._parseparen()
token_type, next = self.next()
if next != ':':
debug.warning("class syntax: %s@%s" % (cname, self.start_pos[0]))
return None
# because of 2 line class initializations
scope = Class(cname, super, first_pos)
if self.user_scope and scope != self.user_scope \
and self.user_position > first_pos:
self.user_scope = scope
return scope
def _parse_statement(self, pre_used_token=None, added_breaks=None,
stmt_class=Statement, list_comp=False):
"""
Parses statements like:
>>> a = test(b)
>>> a += 3 - 2 or b
and so on. One line at a time.
:param pre_used_token: The pre parsed token.
:type pre_used_token: set
:return: Statement + last parsed token.
:rtype: (Statement, str)
"""
string = ''
set_vars = []
used_funcs = []
used_vars = []
level = 0 # The level of parentheses
is_return = None
if pre_used_token:
token_type, tok = pre_used_token
else:
token_type, tok = self.next()
while token_type == tokenize.COMMENT:
# remove newline and comment
self.next()
token_type, tok = self.next()
first_pos = self.start_pos
opening_brackets = ['{', '(', '[']
closing_brackets = ['}', ')', ']']
# the difference between "break" and "always break" is that the latter
# will even break in parentheses. This is true for typical flow
# commands like def and class and the imports, which will never be used
# in a statement.
breaks = ['\n', ':', ')']
always_break = [';', 'import', 'from', 'class', 'def', 'try', 'except',
'finally', 'while']
if added_breaks:
breaks += added_breaks
tok_list = []
while not (tok in always_break or tok in breaks and level <= 0):
try:
set_string = None
#print 'parse_stmt', tok, tokenize.tok_name[token_type]
tok_list.append(self.current + (self.start_pos,))
if tok == 'as':
string += " %s " % tok
token_type, tok = self.next()
if token_type == tokenize.NAME:
n, token_type, tok = self._parsedotname(self.current)
if n:
set_vars.append(n)
tok_list.append(n)
string += ".".join(n.names)
continue
elif token_type == tokenize.NAME:
if tok in ['return', 'yield', 'del', 'raise']:
if len(tok_list) > 1:
# this happens, when a statement has opening
# brackets, which are not closed again, here I just
# start a new statement. This is a hack, but I
# could not come up with a better solution.
# This is basically a reset of the statement.
debug.warning('keyword in statement %s@%s',
tok_list, self.start_pos[0])
tok_list = [self.current + (self.start_pos,)]
set_vars = []
used_funcs = []
used_vars = []
level = 0
set_string = tok + ' '
if tok in ['return', 'yield']:
is_return = tok
elif tok == 'for':
# list comprehensions!
middle, tok = self._parse_statement(
added_breaks=['in'])
if tok != 'in' or middle is None:
if middle is None:
level -= 1
else:
middle.parent = self.scope
debug.warning('list comprehension formatting @%s' %
self.start_pos[0])
continue
b = [')', ']']
in_clause, tok = self._parse_statement(added_breaks=b,
list_comp=True)
if tok not in b or in_clause is None:
middle.parent = self.scope
if in_clause is None:
self.gen.push_back(self._current_full)
else:
in_clause.parent = self.scope
in_clause.parent = self.scope
debug.warning('list comprehension in_clause %s@%s'
% (tok, self.start_pos[0]))
continue
other_level = 0
for i, tok in enumerate(reversed(tok_list)):
if not isinstance(tok, (Name, ListComprehension)):
tok = tok[1]
if tok in closing_brackets:
other_level -= 1
elif tok in opening_brackets:
other_level += 1
if other_level > 0:
break
else:
# could not detect brackets -> nested list comp
i = 0
tok_list, toks = tok_list[:-i], tok_list[-i:-1]
src = ''
for t in toks:
src += t[1] if isinstance(t, tuple) \
else t.get_code()
st = Statement(src, [], [], [],
toks, first_pos, self.end_pos)
for s in [st, middle, in_clause]:
s.parent = self.scope
tok = ListComprehension(st, middle, in_clause)
tok_list.append(tok)
if list_comp:
string = ''
string += tok.get_code()
continue
else:
n, token_type, tok = self._parsedotname(self.current)
# removed last entry, because we add Name
tok_list.pop()
if n:
tok_list.append(n)
if tok == '(':
# it must be a function
used_funcs.append(n)
else:
used_vars.append(n)
if string and re.match(r'[\w\d\'"]', string[-1]):
string += ' '
string += ".".join(n.names)
continue
elif tok.endswith('=') and tok not in ['>=', '<=', '==', '!=']:
# there has been an assignement -> change vars
if level == 0:
set_vars += used_vars
used_vars = []
elif tok in opening_brackets:
level += 1
elif tok in closing_brackets:
level -= 1
string = set_string if set_string is not None else string + tok
token_type, tok = self.next()
except StopIteration:
# comes from tokenizer
break
if not string:
return None, tok
#print 'new_stat', string, set_vars, used_funcs, used_vars
if self.freshscope and not self.no_docstr and len(tok_list) == 1 \
and self.last_token[0] == tokenize.STRING:
self.scope.add_docstr(self.last_token[1])
return None, tok
else:
stmt = stmt_class(string, set_vars, used_funcs, used_vars,
tok_list, first_pos, self.end_pos)
self._check_user_stmt(stmt)
if is_return:
# add returns to the scope
func = self.scope.get_parent_until(Function)
if is_return == 'yield':
func.is_generator = True
try:
func.returns.append(stmt)
except AttributeError:
debug.warning('return in non-function')
if tok in always_break:
self.gen.push_back(self._current_full)
return stmt, tok
def next(self):
return self.__next__()
def __iter__(self):
return self
def __next__(self):
""" Generate the next tokenize pattern. """
try:
self._current_full = next(self.gen)
except tokenize.TokenError:
# We just ignore this error, I try to handle it earlier - as
# good as possible
debug.warning('parentheses not closed error')
except IndentationError:
# This is an error, that tokenize may produce, because the code
# is not indented as it should. Here it just ignores this line
# and restarts the parser.
# (This is a rather unlikely error message, for normal code,
# tokenize seems to be pretty tolerant)
debug.warning('indentation error on line %s, ignoring it' %
(self.start_pos[0]))
self._line_of_tokenize_restart = self.start_pos[0] + 1
self.gen = PushBackIterator(tokenize.generate_tokens(
self.buf.readline))
return self.next()
except StopIteration:
# set end_pos correctly, if we finish
s = self.scope
while s is not None:
s.end_pos = self.end_pos
s = s.parent
raise
type, tok, self._tokenize_start_pos, self._tokenize_end_pos, \
self.parserline = self._current_full
if self.user_position and (self.start_pos[0] == self.user_position[0]
or self.user_scope is None
and self.start_pos[0] >= self.user_position[0]):
debug.dbg('user scope found [%s] = %s' % \
(self.parserline.replace('\n', ''), repr(self.scope)))
self.user_scope = self.scope
self.last_token = self.current
self.current = (type, tok)
return self.current
def parse(self):
"""
The main part of the program. It analyzes the given code-text and
returns a tree-like scope. For a more detailed description, see the
class description.
:param text: The code which should be parsed.
:param type: str
:raises: IndentationError
"""
self.buf = StringIO(self.code)
self.gen = PushBackIterator(tokenize.generate_tokens(
self.buf.readline))
extended_flow = ['else', 'elif', 'except', 'finally']
statement_toks = ['{', '[', '(', '`']
decorators = []
self.freshscope = True
self.iterator = iter(self)
# This iterator stuff is not intentional. It grew historically.
for token_type, tok in self.iterator:
self.module.temp_used_names = []
#debug.dbg('main: tok=[%s] type=[%s] indent=[%s]'\
# % (tok, tokenize.tok_name[token_type], start_position[0]))
while token_type == tokenize.DEDENT and self.scope != self.module:
token_type, tok = self.next()
if self.start_pos[1] <= self.scope.start_pos[1]:
self.scope.end_pos = self.start_pos
self.scope = self.scope.parent
# check again for unindented stuff. this is true for syntax
# errors. only check for names, because thats relevant here. If
# some docstrings are not indented, I don't care.
while self.start_pos[1] <= self.scope.start_pos[1] \
and (token_type == tokenize.NAME or tok in ['(', '['])\
and self.scope != self.module:
self.scope.end_pos = self.start_pos
self.scope = self.scope.parent
first_pos = self.start_pos
if tok == 'def':
func = self._parsefunction()
if func is None:
debug.warning("function: syntax error@%s" %
self.start_pos[0])
continue
self.freshscope = True
self.scope = self.scope.add_scope(func, decorators)
decorators = []
elif tok == 'class':
cls = self._parseclass()
if cls is None:
debug.warning("class: syntax error@%s" % self.start_pos[0])
continue
self.freshscope = True
self.scope = self.scope.add_scope(cls, decorators)
decorators = []
# import stuff
elif tok == 'import':
imports = self._parseimportlist()
for m, alias, defunct in imports:
i = Import(first_pos, self.end_pos, m, alias,
defunct=defunct)
self._check_user_stmt(i)
self.scope.add_import(i)
if not imports:
i = Import(first_pos, self.end_pos, None, defunct=True)
self._check_user_stmt(i)
self.freshscope = False
elif tok == 'from':
defunct = False
# take care for relative imports
relative_count = 0
while 1:
token_type, tok = self.next()
if tok != '.':
break
relative_count += 1
# the from import
mod, token_type, tok = self._parsedotname(self.current)
if str(mod) == 'import' and relative_count:
self.gen.push_back(self._current_full)
tok = 'import'
mod = None
if not mod and not relative_count or tok != "import":
debug.warning("from: syntax error@%s" % self.start_pos[0])
defunct = True
if tok != 'import':
self.gen.push_back(self._current_full)
names = self._parseimportlist()
for name, alias, defunct2 in names:
star = name is not None and name.names[0] == '*'
if star:
name = None
i = Import(first_pos, self.end_pos, name, alias, mod,
star, relative_count, defunct=defunct or defunct2)
self._check_user_stmt(i)
self.scope.add_import(i)
self.freshscope = False
#loops
elif tok == 'for':
set_stmt, tok = self._parse_statement(added_breaks=['in'])
if tok == 'in':
statement, tok = self._parse_statement()
if tok == ':':
s = [] if statement is None else [statement]
f = ForFlow(s, first_pos, set_stmt)
self.scope = self.scope.add_statement(f)
else:
debug.warning('syntax err, for flow started @%s',
self.start_pos[0])
if statement is not None:
statement.parent = self.scope
if set_stmt is not None:
set_stmt.parent = self.scope
else:
debug.warning('syntax err, for flow incomplete @%s',
self.start_pos[0])
if set_stmt is not None:
set_stmt.parent = self.scope
elif tok in ['if', 'while', 'try', 'with'] + extended_flow:
added_breaks = []
command = tok
if command in ['except', 'with']:
added_breaks.append(',')
# multiple statements because of with
inits = []
first = True
while first or command == 'with' \
and tok not in [':', '\n']:
statement, tok = \
self._parse_statement(added_breaks=added_breaks)
if command == 'except' and tok in added_breaks:
# the except statement defines a var
# this is only true for python 2
n, token_type, tok = self._parsedotname()
if n:
statement.set_vars.append(n)
statement.code += ',' + n.get_code()
if statement:
inits.append(statement)
first = False
if tok == ':':
f = Flow(command, inits, first_pos)
if command in extended_flow:
# the last statement has to be another part of
# the flow statement, because a dedent releases the
# main scope, so just take the last statement.
try:
s = self.scope.statements[-1].set_next(f)
except (AttributeError, IndexError):
# If set_next doesn't exist, just add it.
s = self.scope.add_statement(f)
else:
s = self.scope.add_statement(f)
self.scope = s
else:
for i in inits:
i.parent = self.scope
debug.warning('syntax err, flow started @%s',
self.start_pos[0])
# globals
elif tok == 'global':
stmt, tok = self._parse_statement(self.current)
if stmt:
self.scope.add_statement(stmt)
for name in stmt.used_vars:
# add the global to the top, because there it is
# important.
self.module.add_global(name)
# decorator
elif tok == '@':
stmt, tok = self._parse_statement()
decorators.append(stmt)
elif tok == 'pass':
continue
elif tok == 'assert':
stmt, tok = self._parse_statement()
stmt.parent = self.scope
self.scope.asserts.append(stmt)
# default
elif token_type in [tokenize.NAME, tokenize.STRING,
tokenize.NUMBER] \
or tok in statement_toks:
# this is the main part - a name can be a function or a
# normal var, which can follow anything. but this is done
# by the statement parser.
stmt, tok = self._parse_statement(self.current)
if stmt:
self.scope.add_statement(stmt)
self.freshscope = False
else:
if token_type not in [tokenize.COMMENT, tokenize.INDENT,
tokenize.NEWLINE, tokenize.NL,
tokenize.ENDMARKER]:
debug.warning('token not classified', tok, token_type,
self.start_pos[0])
del self.buf
return self.module