#!/usr/bin/env python

# Copyright (c) 2004-2006 ActiveState Software Inc.

#

# Contributors:

# Trent Mick ([email protected])

"""

pythoncile - a Code Intelligence Language Engine for the Python language

Module Usage:

from pythoncile import scan

mtime = os.stat("foo.py")[stat.ST_MTIME]

content = open("foo.py", "r").read()

scan(content, "foo.py", mtime=mtime)

Command-line Usage:

pythoncile.py [<options>...] [<Python files>...]

Options:

-h, --help dump this help and exit

-V, --version dump this script's version and exit

-v, --verbose verbose output, use twice for more verbose output

-f, --filename <path> specify the filename of the file content

passed in on stdin, this is used for the "path"

attribute of the emitted <file> tag.

--md5=<string> md5 hash for the input

--mtime=<secs> modification time for output info, in #secs since

1/1/70.

-L, --language <name>

the language of the file being scanned

-c, --clock print timing info for scans (CIX is not printed)

One or more Python files can be specified as arguments or content can be

passed in on stdin. A directory can also be specified, in which case

all .py files in that directory are scanned.

This is a Language Engine for the Code Intelligence (codeintel) system.

Code Intelligence XML format. See:

http://specs.activestate.com/Komodo_3.0/func/code_intelligence.html

The command-line interface will return non-zero iff the scan failed.

"""

# Dev Notes:

# <none>

#

# TODO:

# - type inferencing: asserts

# - type inferencing: return statements

# - type inferencing: calls to isinstance

# - special handling for None may be required

# - Comments and doc strings. What format?

# - JavaDoc - type hard to parse and not reliable

# (http://java.sun.com/j2se/javadoc/writingdoccomments/).

# - PHPDoc? Possibly, but not that rigorous.

# - Grouch (http://www.mems-exchange.org/software/grouch/) -- dunno yet.

# - Don't like requirement for "Instance attributes:" landmark in doc

# strings.

# - This can't be a full solution because the requirement to repeat

# the argument name doesn't "fit" with having a near-by comment when

# variable is declared.

# - Two space indent is quite rigid

# - Only allowing attribute description on the next line is limiting.

# - Seems focussed just on class attributes rather than function

# arguments.

# - Perhaps what PerlCOM POD markup uses?

# - Home grown? My own style? Dunno

# - make type inferencing optional (because it will probably take a long

# time to generate), this is tricky though b/c should the CodeIntel system

# re-scan a file after "I want type inferencing now" is turned on? Hmmm.

# - [lower priority] handle staticmethod(methname) and

# classmethod(methname). This means having to delay emitting XML until

# end of class scope and adding .visitCallFunc().

# - [lower priority] look for associated comments for variable

# declarations (as per VS.NET's spec, c.f. "Supplying Code Comments" in

# the VS.NET user docs)

import os

import sys

import getopt

from hashlib import md5

import re

import logging

import pprint

import glob

import time

import stat

import types

import io

from functools import partial

# this particular ET is different from xml.etree and is expected

# to be returned from scan_et() by the clients of this module

import ciElementTree as ET

import ast

import parser

from codeintel2.common import CILEError

from codeintel2 import util

from codeintel2 import tdparser

#---- exceptions

class PythonCILEError(CILEError):

pass

#---- global data

_version_ = (0, 3, 0)

log = logging.getLogger("codeintel.pythoncile")

# log.setLevel(logging.DEBUG)

util.makePerformantLogger(log)

_gClockIt = 0 # if true then we are gathering timing data

_gClock = None # if gathering timing data this is set to time retrieval fn

_gStartTime = None # start time of current file being scanned

#---- internal routines and classes

def _isclass(namespace):

return (len(namespace["types"]) == 1

and "class" in namespace["types"])

def _isfunction(namespace):

return (len(namespace["types"]) == 1

and "function" in namespace["types"])

def getAttrStr(attrs):

"""Construct an XML-safe attribute string from the given attributes

"attrs" is a dictionary of attributes

The returned attribute string includes a leading space, if necessary,

so it is safe to use the string right after a tag name. Any Unicode

attributes will be encoded into UTF8 encoding as part of this process.

"""

from xml.sax.saxutils import quoteattr

s = ''

for attr, value in list(attrs.items()):

if not isinstance(value, str):

value = str(value)

elif isinstance(value, str):

value = value.encode("utf-8")

s += ' %s=%s' % (attr, quoteattr(value))

return s

# match 0x00-0x1f except TAB(0x09), LF(0x0A), and CR(0x0D)

_encre = re.compile('([\x00-\x08\x0b\x0c\x0e-\x1f])')

# XXX: this is not used anywhere, is it needed at all?

if sys.version_info >= (2, 3):

charrefreplace = 'xmlcharrefreplace'

else:

# Python 2.2 doesn't have 'xmlcharrefreplace'. Fallback to a

# literal '?' -- this is better than failing outright.

charrefreplace = 'replace'

def xmlencode(s):

"""Encode the given string for inclusion in a UTF-8 XML document.

Note: s must *not* be Unicode, it must be encoded before being passed in.

Specifically, illegal or unpresentable characters are encoded as

XML character entities.

"""

# As defined in the XML spec some of the character from 0x00 to 0x19

# are not allowed in well-formed XML. We replace those with entity

# references here.

# http://www.w3.org/TR/2000/REC-xml-20001006#charsets

#

# Dev Notes:

# - It would be nice if Python has a codec for this. Perhaps we

# should write one.

# - Eric, at one point, had this change to '_xmlencode' for rubycile:

# p4 diff2 -du \

# //depot/main/Apps/Komodo-devel/src/codeintel/ruby/rubycile.py#7 \

# //depot/main/Apps/Komodo-devel/src/codeintel/ruby/rubycile.py#8

# but:

# My guess is that there was a bug here, and explicitly

# utf-8-encoding non-ascii characters fixed it. This was a year

# ago, and I don't recall what I mean by "avoid shuffling the data

# around", but it must be related to something I observed without

# that code.

# replace with XML decimal char entity, e.g. '&#7;'

return _encre.sub(lambda m: '&#%d;' % ord(m.group(1)), s)

def cdataescape(s):

"""Return the string escaped for inclusion in an XML CDATA section.

Note: Any Unicode will be encoded to UTF8 encoding as part of this process.

A CDATA section is terminated with ']]>', therefore this token in the

content must be escaped. To my knowledge the XML spec does not define

how to do that. My chosen escape is (courteousy of EricP) is to split

that token into multiple CDATA sections, so that, for example:

blah...]]>...blah

becomes:

blah...]]]]><![CDATA[>...blah

and the resulting content should be copacetic:

<b><![CDATA[blah...]]]]><![CDATA[>...blah]]></b>

"""

if isinstance(s, str):

s = s.encode("utf-8")

parts = s.split("]]>")

return "]]]]><![CDATA[>".join(parts)

def _unistr(x):

if isinstance(x, str):

return x

elif isinstance(x, str):

return x.decode('utf8')

else:

return str(x)

def _et_attrs(attrs):

return dict((_unistr(k), xmlencode(_unistr(v))) for k, v in list(attrs.items())

if v is not None)

def _et_data(data):

return xmlencode(_unistr(data))

def _node_attrs(node, **kw):

return dict(name=node["name"],

line=node.get("line"),

doc=node.get("doc"),

attributes=node.get("attributes") or None,

**kw)

def _node_citdl(node):

max_type = None

max_score = -1

#'guesses' is a types dict: {<type guess>: <score>, ...}

guesses = node.get("types", {})

for type, score in list(guesses.items()):

if ' ' in type:

# XXX Drop the <start-scope> part of CITDL for now.

type = type.split(None, 1)[0]

# Don't emit None types, it does not help us. Fix for bug:

# http://bugs.activestate.com/show_bug.cgi?id=71989

if type != "None":

if score > max_score:

max_type = type

max_score = score

return max_type

class AST2CIXVisitor(ast.NodeVisitor):

"""Generate Code Intelligence XML (CIX) from walking a Python AST tree.

This just generates the CIX content _inside_ of the <file/> tag. The

prefix and suffix have to be added separately.

Note: All node text elements are encoded in UTF-8 format by the Python AST

tree processing, no matter what encoding is used for the file's

original content. The generated CIX XML will also be UTF-8 encoded.

"""

DEBUG = 0

def __init__(self, moduleName=None, content=None, lang="Python"):

self.lang = lang

if self.DEBUG is None:

self.DEBUG = log.isEnabledFor(logging.DEBUG)

self.moduleName = moduleName

self.content = content

if content and self.DEBUG:

self.lines = content.splitlines(0)

else:

self.lines = None

# Symbol Tables (dicts) are built up for each scope. The namespace

# stack to the global-level is maintain in self.nsstack.

self.st = { # the main module symbol table

# <scope name>: <namespace dict>

}

self.nsstack = []

self.cix = ET.TreeBuilder()

self.tree = None

def parse(self):

"""Parse text into a tree and walk the result"""

self.tree = ast.parse(self.content)

def walk(self):

return self.visit(self.tree)

def emit_start(self, s, attrs={}):

self.cix.start(s, _et_attrs(attrs))

def emit_data(self, data):

self.cix.data(_et_data(data))

def emit_end(self, s):

self.cix.end(s)

def emit_tag(self, s, attrs={}, data=None):

self.emit_start(s, _et_attrs(attrs))

if data is not None:

self.emit_data(data)

self.emit_end(s)

def cix_module(self, node):

"""Emit CIX for the given module namespace."""

# log.debug("cix_module(%s, level=%r)", '.'.join(node["nspath"]),

# level)

assert len(node["types"]) == 1 and "module" in node["types"]

attrs = _node_attrs(node, lang=self.lang, ilk="blob")

module = self.emit_start('scope', attrs)

for import_ in node.get("imports", []):

self.cix_import(import_)

self.cix_symbols(node["symbols"])

self.emit_end('scope')

def cix_import(self, node):

# log.debug("cix_import(%s, level=%r)", node["module"], level)

attrs = node

self.emit_tag('import', attrs)

def cix_symbols(self, node, parentIsClass=0):

# Sort variables by line order. This provide the most naturally

# readable comparison of document with its associate CIX content.

vars = sorted(list(node.values()), key=lambda v: v.get("line"))

for var in vars:

self.cix_symbol(var, parentIsClass)

def cix_symbol(self, node, parentIsClass=0):

if _isclass(node):

self.cix_class(node)

elif _isfunction(node):

self.cix_function(node)

else:

self.cix_variable(node, parentIsClass)

def cix_variable(self, node, parentIsClass=0):

# log.debug("cix_variable(%s, level=%r, parentIsClass=%r)",

# '.'.join(node["nspath"]), level, parentIsClass)

attrs = _node_attrs(node, citdl=_node_citdl(node))

if parentIsClass and "is-class-var" not in node:

# Special CodeIntel <variable> attribute to distinguish from the

# usual class variables.

if attrs["attributes"]:

attrs["attributes"] += " __instancevar__"

else:

attrs["attributes"] = "__instancevar__"

self.emit_tag('variable', attrs)

def cix_class(self, node):

# log.debug("cix_class(%s, level=%r)", '.'.join(node["nspath"]), level)

if node["classrefs"]:

citdls = (t for t in (_node_citdl(n) for n in node["classrefs"])

if t is not None)

classrefs = " ".join(citdls)

else:

classrefs = None

attrs = _node_attrs(node,

lineend=node.get("lineend"),

signature=node.get("signature"),

ilk="class",

classrefs=classrefs)

self.emit_start('scope', attrs)

for import_ in node.get("imports", []):

self.cix_import(import_)

self.cix_symbols(node["symbols"], parentIsClass=1)

self.emit_end('scope')

def cix_argument(self, node):

# log.debug("cix_argument(%s, level=%r)", '.'.join(node["nspath"]),

# level)

attrs = _node_attrs(node, citdl=_node_citdl(node), ilk="argument")

self.emit_tag('variable', attrs)

def cix_function(self, node):

# log.debug("cix_function(%s, level=%r)", '.'.join(node["nspath"]), level)

# Determine the best return type.

best_citdl = None

max_count = 0

for citdl, count in list(node["returns"].items()):

if count > max_count:

best_citdl = citdl

attrs = _node_attrs(node,

lineend=node.get("lineend"),

returns=best_citdl,

signature=node.get("signature"),

ilk="function")

self.emit_start("scope", attrs)

for import_ in node.get("imports", []):

self.cix_import(import_)

argNames = []

for arg in node["arguments"]:

argNames.append(arg["name"])

self.cix_argument(arg)

symbols = {} # don't re-emit the function arguments

for symbolName, symbol in list(node["symbols"].items()):

if symbolName not in argNames:

symbols[symbolName] = symbol

self.cix_symbols(symbols)

# XXX <returns/> if one is defined

self.emit_end('scope')

def getCIX(self, path):

"""Return CIX content for parsed data."""

log.debug("getCIX")

self.emit_start('file', dict(lang=self.lang, path=path))

if self.st:

moduleNS = self.st[()]

self.cix_module(moduleNS)

self.emit_end('file')

file = self.cix.close()

return file

# def generic_visit(self, node):

# method = 'visit_' + node.__class__.__name__

# if not hasattr(self, method):

# log.info("generic visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

# return super(AST2CIXVisitor, self).generic_visit(node)

def visit_Module(self, node):

log.info("visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

nspath = ()

namespace = {"name": self.moduleName,

"nspath": nspath,

"types": {"module": 1},

"symbols": {}}

doc = ast.get_docstring(node)

if doc:

summarylines = util.parseDocSummary(doc.splitlines(0))

namespace["doc"] = "\n".join(summarylines)

self.st[nspath] = namespace

self.nsstack.append(namespace)

self.generic_visit(node)

self.nsstack.pop()

def visit_Return(self, node):

log.info("visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

citdl_types = self._guessTypes(node.value)

for citdl in citdl_types:

if citdl:

citdl = citdl.split(None, 1)[0]

if citdl and citdl not in ("None", "NoneType"):

if citdl in ("False", "True"):

citdl = "bool"

func_node = self.nsstack[-1]

t = func_node["returns"]

t[citdl] = t.get(citdl, 0) + 1

def visit_ClassDef(self, node):

log.info("visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

locals = self.nsstack[-1]

name = node.name

nspath = locals["nspath"] + (name,)

namespace = {

"nspath": nspath,

"name": name,

"types": {"class": 1},

# XXX Example of a base class that might surprise: the

# __metaclass__ class in

# c:\python22\lib\site-packages\ctypes\com\automation.py

# Should this be self._getCITDLExprRepr()???

"classrefs": [],

"symbols": {},

}

namespace["declaration"] = namespace

if node.lineno:

namespace["line"] = node.lineno

namespace["lineend"] = node.lineno

lastNode = node

while True:

try:

lastNode = list(ast.iter_child_nodes(lastNode))[-1]

if hasattr(lastNode, 'lineno'):

namespace["lineend"] = lastNode.lineno

except IndexError:

break

attributes = []

if name.startswith("__") and name.endswith("__"):

pass

elif name.startswith("__"):

attributes.append("private")

elif name.startswith("_"):

attributes.append("protected")

namespace["attributes"] = ' '.join(attributes)

if node.bases:

for baseNode in node.bases:

baseName = self._getExprRepr(baseNode)

classref = {"name": baseName, "types": {}}

for t in self._guessTypes(baseNode):

if t not in classref["types"]:

classref["types"][t] = 0

classref["types"][t] += 1

namespace["classrefs"].append(classref)

doc = ast.get_docstring(node)

if doc:

siglines, desclines = util.parsePyFuncDoc(doc)

if siglines:

namespace["signature"] = "\n".join(siglines)

if desclines:

namespace["doc"] = "\n".join(desclines)

self.st[nspath] = locals["symbols"][name] = namespace

self.nsstack.append(namespace)

self.generic_visit(node)

self.nsstack.pop()

def visit_FunctionDef(self, node):

log.info("visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

parent = self.nsstack[-1]

parentIsClass = _isclass(parent)

namespace = {

"types": {"function": 1},

"returns": {},

"arguments": [],

"symbols": {},

}

namespace["declaration"] = namespace

if node.lineno:

namespace["line"] = node.lineno

namespace["lineend"] = node.lineno

lastNode = node

while True:

try:

lastNode = list(ast.iter_child_nodes(lastNode))[-1]

if hasattr(lastNode, 'lineno'):

namespace["lineend"] = lastNode.lineno

except IndexError:

break

name = node.name

# Determine attributes

attributes = []

if name.startswith("__") and name.endswith("__"):

pass

elif name.startswith("__"):

attributes.append("private")

elif name.startswith("_"):

attributes.append("protected")

if name == "__init__" and parentIsClass:

attributes.append("__ctor__")

# process decorators

prop_var = None

if node.decorator_list:

for deco in node.decorator_list:

deco_name = getattr(deco, 'id', None)

prop_mode = None

if deco_name == 'staticmethod':

attributes.append("__staticmethod__")

continue

if deco_name == 'classmethod':

attributes.append("__classmethod__")

continue

if deco_name == 'property':

prop_mode = 'getter'

elif hasattr(deco, 'attrname') and deco.attrname in ('getter',

'setter',

'deleter'):

prop_mode = deco.attrname

if prop_mode:

if prop_mode == 'getter':

# it's a getter, create a pseudo-var

prop_var = parent["symbols"].get(name, None)

if prop_var is None:

prop_var = dict(name=name,

nspath=parent["nspath"] + (name,),

doc=None,

types={},

symbols={})

var_attrs = ['property']

if name.startswith("__") and name.endswith("__"):

pass

elif name.startswith("__"):

var_attrs.append("private")

elif name.startswith("_"):

var_attrs.append("protected")

prop_var["attributes"] = ' '.join(var_attrs)

prop_var["declaration"] = prop_var

parent["symbols"][name] = prop_var

if not "is-class-var" in prop_var:

prop_var["is-class-var"] = 1

# hide the function

attributes += ['__hidden__']

name += " (property %s)" % prop_mode

# only one property decorator makes sense

break

namespace["attributes"] = ' '.join(attributes)

if parentIsClass and name == "__init__":

fallbackSig = parent["name"]

else:

fallbackSig = name

namespace["name"] = name

nspath = parent["nspath"] + (name,)

namespace["nspath"] = nspath

# Handle arguments. The format of the relevant Function attributes

# makes this a little bit of pain.

node_args = node.args

defaultArgsBaseIndex = len(node_args.args) - len(node_args.defaults)

if node_args.kwarg:

defaultArgsBaseIndex -= 1

if node_args.vararg:

defaultArgsBaseIndex -= 1

varargsIndex = len(node_args.args) - 2

else:

varargsIndex = None

kwargsIndex = len(node_args.args) - 1

elif node_args.vararg:

defaultArgsBaseIndex -= 1

varargsIndex = len(node_args.args) - 1

kwargsIndex = None

else:

varargsIndex = kwargsIndex = None

sigArgs = []

for i in range(len(node_args.args)):

argName = node_args.args[i].arg

argument = {"name": argName,

"nspath": nspath + (argName,),

"doc": None,

"types": {},

"line": node.lineno,

"symbols": {}}

if i == kwargsIndex:

argument["attributes"] = "kwargs"

sigArgs.append("**" + argName)

elif i == varargsIndex:

argument["attributes"] = "varargs"

sigArgs.append("*" + argName)

elif i >= defaultArgsBaseIndex:

defaultNode = node_args.defaults[i - defaultArgsBaseIndex]

try:

argument["default"] = self._getExprRepr(defaultNode)

except PythonCILEError as ex:

raise PythonCILEError("unexpected default argument node "

"type for Function '%s': %s"

% (node.name, ex))

sigArgs.append(argName + '=' + argument["default"])

for t in self._guessTypes(defaultNode):

log.info("guessed type: %s ::= %s", argName, t)

if t not in argument["types"]:

argument["types"][t] = 0

argument["types"][t] += 1

else:

sigArgs.append(argName)

if i == 0 and parentIsClass:

# If this is a class method, then the first arg is the class

# instance.

className = self.nsstack[-1]["nspath"][-1]

argument["types"][className] = 1

argument["declaration"] = self.nsstack[-1]

arguments = [argument]

for argument in arguments:

if "declaration" not in argument:

argument[

"declaration"] = argument # namespace dict of the declaration

namespace["arguments"].append(argument)

namespace["symbols"][argument["name"]] = argument

# Drop first "self" argument from class method signatures.

# - This is a little bit of a compromise as the "self" argument

# should *sometimes* be included in a method's call signature.

if _isclass(parent) and sigArgs and "__staticmethod__" not in attributes:

# Delete the first "self" argument.

del sigArgs[0]

fallbackSig += "(%s)" % (", ".join(sigArgs))

if "__staticmethod__" in attributes:

fallbackSig += " - staticmethod"

elif "__classmethod__" in attributes:

fallbackSig += " - classmethod"

doc = ast.get_docstring(node)

if doc:

siglines, desclines = util.parsePyFuncDoc(doc, [fallbackSig])

namespace["signature"] = "\n".join(siglines)

if desclines:

namespace["doc"] = "\n".join(desclines)

else:

namespace["signature"] = fallbackSig

self.st[nspath] = parent["symbols"][name] = namespace

self.nsstack.append(namespace)

self.generic_visit(node)

self.nsstack.pop()

if prop_var:

# this is a property getter function,

# copy its return types to the corresponding property variable...

var_types = prop_var["types"]

for t in namespace["returns"]:

if t not in var_types:

var_types[t] = 0

else:

var_types[t] += 1

# ... as well as its line number

if "line" in namespace:

prop_var["line"] = namespace["line"]

def visit_Import(self, node):

log.info("visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

imports = self.nsstack[-1].setdefault("imports", [])

for alias in node.names:

import_ = {"module": alias.name}

if node.lineno:

import_["line"] = node.lineno

if alias.asname:

import_["alias"] = alias.asname

imports.append(import_)

def visit_ImportFrom(self, node):

log.info("visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

imports = self.nsstack[-1].setdefault("imports", [])

module = node.module or ''

if node.level > 0:

module = ("." * node.level) + module

for alias in node.names:

import_ = {"module": module, "symbol": alias.name}

if node.lineno:

import_["line"] = node.lineno

if alias.asname:

import_["alias"] = alias.asname

imports.append(import_)

# XXX

# def visit_Return(self, node):

# # set __rettypes__ on Functions

# pass

# def visit_Global(self, node):

# # note for future visitAssign to control namespace

# pass

# def visit_Yield(self, node):

# # modify the Function into a generator??? what are the implications?

# pass

# def visit_Assert(self, node):

# # support the assert hints that Wing does

# pass

def _assignVariable(self, varName, namespace, rhsNode, line,

isClassVar=0):

"""Handle a simple variable name assignment.

"varName" is the variable name being assign to.

"namespace" is the namespace dict to which to assign the variable.

"rhsNode" is the ast.Node of the right-hand side of the

assignment.

"line" is the line number on which the variable is being assigned.

"isClassVar" (optional) is a boolean indicating if this var is

a class variable, as opposed to an instance variable

"""

log.debug("_assignVariable(varName=%r, namespace %s, rhsNode=%r, "

"line, isClassVar=%r)", varName,

'.'.join(namespace["nspath"]), rhsNode, isClassVar)

variable = namespace["symbols"].get(varName, None)

new_var = False

if variable is None:

new_var = True

variable = {"name": varName,

"nspath": namespace["nspath"]+(varName,),

# Could try to parse documentation from a near-by

# string.

"doc": None,

# 'types' is a dict mapping a type name to the number

# of times this was guessed as the variable type.

"types": {},

"symbols": {}}

# Determine attributes

attributes = []

if varName.startswith("__") and varName.endswith("__"):

pass

elif varName.startswith("__"):

attributes.append("private")

elif varName.startswith("_"):

attributes.append("protected")

variable["attributes"] = ' '.join(attributes)

variable["declaration"] = variable

if line:

variable["line"] = line

namespace["symbols"][varName] = variable

if isClassVar and not "is-class-var" in variable:

variable["is-class-var"] = 1

# line number of first class-level assignment wins

if line:

variable["line"] = line

if (not new_var and

_isfunction(variable) and

isinstance(rhsNode, ast.Call) and

rhsNode.args and

isinstance(rhsNode.args[0], ast.Name) and

variable["name"] == rhsNode.args[0].id

):

# a speial case for 2.4-styled decorators

return

varTypes = variable["types"]

for t in self._guessTypes(rhsNode, namespace):

log.info("guessed type: %s ::= %s", varName, t)

if t not in varTypes:

varTypes[t] = 0

varTypes[t] += 1

def _visitSimpleAssign(self, lhsNode, rhsNode, line):

"""Handle a simple assignment: assignment to a symbol name or to

an attribute of a symbol name. If the given left-hand side (lhsNode)

is not an node type that can be handled, it is dropped.

"""

log.debug("_visitSimpleAssign(lhsNode=%r, rhsNode=%r)", lhsNode,

rhsNode)

if isinstance(lhsNode, ast.Name):

# E.g.: foo = ...

# Assign this to the local namespace, unless there was a

# 'global' statement. (XXX Not handling 'global' yet.)

ns = self.nsstack[-1]

self._assignVariable(lhsNode.id, ns, rhsNode, line,

isClassVar=_isclass(ns))

elif isinstance(lhsNode, ast.Attribute):

# E.g.: foo.bar = ...

# If we can resolve "foo", then we update that namespace.

variable, citdl = self._resolveObjectRef(lhsNode.value)

if variable:

self._assignVariable(lhsNode.attr,

variable["declaration"], rhsNode, line)

else:

log.debug("could not handle simple assign (module '%s'): "

"lhsNode=%r, rhsNode=%r", self.moduleName, lhsNode,

rhsNode)

def visit_Assign(self, node):

log.info("visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

lhsNode = node.targets[0]

rhsNode = node.value

if isinstance(lhsNode, (ast.Name, ast.Attribute)):

# E.g.:

# foo = ... (Name)

# foo.bar = ... (Attribute)

self._visitSimpleAssign(lhsNode, rhsNode, node.lineno)

elif isinstance(lhsNode, (ast.Tuple, ast.List)):

# E.g.:

# foo, bar = ...

# [foo, bar] = ...

# If the RHS is a sequence with the same number of elements,

# then we update each assigned-to variable. Otherwise, bail.

if isinstance(rhsNode, (ast.Tuple, ast.List)):

if len(lhsNode.elts) == len(rhsNode.elts):

for i in range(len(lhsNode.elts)):

self._visitSimpleAssign(lhsNode.elts[i],

rhsNode.elts[i],

node.lineno)

elif isinstance(rhsNode, ast.Dict):

if len(lhsNode.elts) == len(rhsNode.keys):

for i in range(len(lhsNode.elts)):

self._visitSimpleAssign(lhsNode.elts[i],

rhsNode.keys[i],

node.lineno)

elif isinstance(rhsNode, ast.Call):

for i in range(len(lhsNode.elts)):

self._visitSimpleAssign(lhsNode.elts[i],

None, # we don't have a good type.

node.lineno)

else:

log.info(

"visitAssign:: skipping unknown rhsNode type: %r - %r",

type(rhsNode), rhsNode)

elif isinstance(lhsNode, ast.Slice):

# E.g.: bar[1:2] = "foo"

# We don't bother with these: too hard.

log.info("visitAssign:: skipping slice - too hard")

pass

elif isinstance(lhsNode, ast.Subscript):

# E.g.: bar[1] = "foo"

# We don't bother with these: too hard.

log.info("visitAssign:: skipping subscript - too hard")

pass

else:

raise PythonCILEError("unexpected type of LHS of assignment: %r"

% lhsNode)

def _handleUnknownAssignment(self, assignNode, lineno):

if isinstance(assignNode, ast.Name):

self._visitSimpleAssign(assignNode, None, lineno)

elif isinstance(assignNode, ast.Tuple):

for anode in assignNode.elts:

self._visitSimpleAssign(anode, None, lineno)

def visit_For(self, node):

log.info("visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

# E.g.:

# for foo in ...

# None: don't bother trying to resolve the type of the RHS

self._handleUnknownAssignment(node.target, node.lineno)

self.generic_visit(node)

def visit_With(self, node):

log.info("visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

for item in node.items:

self._handleUnknownAssignment(item.context_expr, node.lineno)

self.generic_visit(node)

def visit_Try(self, node):

log.info("visit_%s:%s: %r %r", node.__class__.__name__, getattr(node, 'lineno', '?'), self.lines and hasattr(node, 'lineno') and self.lines[node.lineno - 1], node._fields)

for body in node.body:

self.visit(body)

for handler in node.handlers:

try:

if handler.name:

lineno = handler.lineno

self._handleUnknownAssignment(handler.name, lineno)

for body in handler.body:

self.visit(body)

except IndexError:

pass

for orelse in node.orelse:

self.visit(orelse)

for finalbody in node.finalbody:

self.visit(finalbody)

def _resolveObjectRef(self, expr):

"""Try to resolve the given expression to a variable namespace.

"expr" is some kind of ast.Node instance.

Returns the following 2-tuple for the object:

(<variable dict>, <CITDL string>)

where,

<variable dict> is the defining dict for the variable, e.g.

{'name': 'classvar', 'types': {'int': 1}}.

This is None if the variable could not be resolved.

<CITDL string> is a string of CITDL code (see the spec) describing

how to resolve the variable later. This is None if the

variable could be resolved or if the expression is not

expressible in CITDL (CITDL does not attempt to be a panacea).

"""

log.debug("_resolveObjectRef(expr=%r)", expr)

if isinstance(expr, ast.Name):

name = expr.id

nspath = self.nsstack[-1]["nspath"]

for i in range(len(nspath), -1, -1):

ns = self.st[nspath[:i]]

if name in ns["symbols"]:

return (ns["symbols"][name], None)

else:

log.debug(

"_resolveObjectRef: %r not in namespace %r", name,

'.'.join(ns["nspath"]))

elif isinstance(expr, ast.Attribute):

obj, citdl = self._resolveObjectRef(expr.value)

decl = obj and obj["declaration"] or None # want the declaration

if (decl # and "symbols" in decl #XXX this "and"-part necessary?

and expr.attr in decl["symbols"]):

return (decl["symbols"][expr.attr], None)

elif isinstance(expr.value, ast.Num):

# Special case: specifically refer to type object for

# attribute access on constants, e.g.:

# ' '.join

citdl = "__builtins__.%s.%s"\

% ((type(expr.value.n).__name__), expr.attr)

return (None, citdl)

# XXX Could optimize here for common built-in attributes. E.g.,

# we *know* that str.join() returns a string.