python/helpers/epydoc/docintrospecter.py

# epydoc -- Introspection
#
# Copyright (C) 2005 Edward Loper
# Author: Edward Loper <edloper@loper.org>
# URL: <http://epydoc.sf.net>
#
# $Id: docintrospecter.py 1678 2008-01-29 17:21:29Z edloper $

"""
Extract API documentation about python objects by directly introspecting
their values.

The function L{introspect_docs()}, which provides the main interface
of this module, examines a Python objects via introspection, and uses
the information it finds to create an L{APIDoc} objects containing the
API documentation for that objects.

The L{register_introspecter()} method can be used to extend the
functionality of C{docintrospector}, by providing methods that handle
special value types.
"""
__docformat__ = 'epytext en'

######################################################################
## Imports
######################################################################

import inspect, re, sys, os.path, imp
# API documentation encoding:
from epydoc.apidoc import *
# Type comparisons:
from types import *
# Error reporting:
from epydoc import log
# Helper functions:
from epydoc.util import *
# For extracting encoding for docstrings:
import epydoc.docparser
# Builtin values
import __builtin__
# Backwards compatibility
from epydoc.compat import * 

######################################################################
## Caches
######################################################################

_valuedoc_cache = {}
"""A cache containing the API documentation for values that we've
already seen.  This cache is implemented as a dictionary that maps a
value's pyid to its L{ValueDoc}.

Note that if we encounter a value but decide not to introspect it
(because it's imported from another module), then C{_valuedoc_cache}
will contain an entry for the value, but the value will not be listed
in L{_introspected_values}."""

_introspected_values = {}
"""A record which values we've introspected, encoded as a dictionary from
pyid to C{bool}."""

def clear_cache():
    """
    Discard any cached C{APIDoc} values that have been computed for
    introspected values.
    """
    _valuedoc_cache.clear()
    _introspected_values.clear()

######################################################################
## Introspection
######################################################################

def introspect_docs(value=None, name=None, filename=None, context=None,
                    is_script=False, module_name=None):
    """
    Generate the API documentation for a specified object by
    introspecting Python values, and return it as a L{ValueDoc}.  The
    object to generate documentation for may be specified using
    the C{value} parameter, the C{filename} parameter, I{or} the
    C{name} parameter.  (It is an error to specify more than one
    of these three parameters, or to not specify any of them.)

    @param value: The python object that should be documented.
    @param filename: The name of the file that contains the python
        source code for a package, module, or script.  If
        C{filename} is specified, then C{introspect} will return a
        C{ModuleDoc} describing its contents.
    @param name: The fully-qualified python dotted name of any
        value (including packages, modules, classes, and
        functions).  C{DocParser} will automatically figure out
        which module(s) it needs to import in order to find the
        documentation for the specified object.
    @param context: The API documentation for the class of module
        that contains C{value} (if available).
    @param module_name: The name of the module where the value is defined.
        Useful to retrieve the docstring encoding if there is no way to
        detect the module by introspection (such as in properties)
    """
    if value is None and name is not None and filename is None:
        value = get_value_from_name(DottedName(name))
    elif value is None and name is None and filename is not None:
        if is_script:
            value = get_value_from_scriptname(filename)
        else:
            value = get_value_from_filename(filename, context)
    elif name is None and filename is None:
        # it's ok if value is None -- that's a value, after all.
        pass 
    else:
        raise ValueError("Expected exactly one of the following "
                         "arguments: value, name, filename")
    
    pyid = id(value)

    # If we've already introspected this value, then simply return
    # its ValueDoc from our cache.
    if pyid in _introspected_values:
        # If the file is a script, then adjust its name.
        if is_script and filename is not None:
            _valuedoc_cache[pyid].canonical_name = DottedName(
                munge_script_name(str(filename)))
        return _valuedoc_cache[pyid]

    # Create an initial value doc for this value & add it to the cache.
    val_doc = _get_valuedoc(value)

    # Introspect the value.
    _introspected_values[pyid] = True
    introspect_func = _get_introspecter(value)
    introspect_func(value, val_doc, module_name=module_name)

    # Set canonical name, if it was given
    if val_doc.canonical_name is UNKNOWN and name is not None:
        val_doc.canonical_name = DottedName(name)

    # If the file is a script, then adjust its name.
    if is_script and filename is not None:
        val_doc.canonical_name = DottedName(munge_script_name(str(filename)))
        
    if val_doc.canonical_name is UNKNOWN and filename is not None:
        shadowed_name = DottedName(value.__name__)
        log.warning("Module %s is shadowed by a variable with "
                    "the same name." % shadowed_name)
        val_doc.canonical_name = DottedName(str(shadowed_name)+"'")

    return val_doc

def _get_valuedoc(value):
    """
    If a C{ValueDoc} for the given value exists in the valuedoc
    cache, then return it; otherwise, create a new C{ValueDoc},
    add it to the cache, and return it.  When possible, the new
    C{ValueDoc}'s C{pyval}, C{repr}, and C{canonical_name}
    attributes will be set appropriately.
    """
    pyid = id(value)
    val_doc = _valuedoc_cache.get(pyid)
    if val_doc is None:
        try: canonical_name = get_canonical_name(value, strict=True)
        except DottedName.InvalidDottedName: canonical_name = UNKNOWN
        val_doc = ValueDoc(pyval=value, canonical_name = canonical_name,
                           docs_extracted_by='introspecter')
        _valuedoc_cache[pyid] = val_doc
        
        # If it's a module, then do some preliminary introspection.
        # Otherwise, check what the containing module is (used e.g.
        # to decide what markup language should be used for docstrings)
        if inspect.ismodule(value):
            introspect_module(value, val_doc, preliminary=True)
            val_doc.defining_module = val_doc
        else:
            module_name = str(get_containing_module(value))
            module = sys.modules.get(module_name)
            if module is not None and inspect.ismodule(module):
                val_doc.defining_module = _get_valuedoc(module)
            
    return val_doc

#////////////////////////////////////////////////////////////
# Module Introspection
#////////////////////////////////////////////////////////////

#: A list of module variables that should not be included in a
#: module's API documentation.
UNDOCUMENTED_MODULE_VARS = (
    '__builtins__', '__doc__', '__all__', '__file__', '__path__',
    '__name__', '__extra_epydoc_fields__', '__docformat__')

def introspect_module(module, module_doc, module_name=None, preliminary=False):
    """
    Add API documentation information about the module C{module}
    to C{module_doc}.
    """
    module_doc.specialize_to(ModuleDoc)

    # Record the module's docformat
    if hasattr(module, '__docformat__'):
        module_doc.docformat = unicode(module.__docformat__)
                                  
    # Record the module's filename
    if hasattr(module, '__file__'):
        try: module_doc.filename = unicode(module.__file__)
        except KeyboardInterrupt: raise
        except: pass
        if module_doc.filename is not UNKNOWN:
            try: module_doc.filename = py_src_filename(module_doc.filename)
            except ValueError: pass

    # If this is just a preliminary introspection, then don't do
    # anything else.  (Typically this is true if this module was
    # imported, but is not included in the set of modules we're
    # documenting.)
    module_doc.variables = {}
    if preliminary: return

    # Record the module's docstring
    if hasattr(module, '__doc__'):
        module_doc.docstring = get_docstring(module)

    # If the module has a __path__, then it's (probably) a
    # package; so set is_package=True and record its __path__.
    if hasattr(module, '__path__'):
        module_doc.is_package = True
        try: module_doc.path = [unicode(p) for p in module.__path__]
        except KeyboardInterrupt: raise
        except: pass
    else:
        module_doc.is_package = False

    # Make sure we have a name for the package.
    dotted_name = module_doc.canonical_name
    if dotted_name is UNKNOWN:
        dotted_name = DottedName(module.__name__)
    name_without_primes = DottedName(str(dotted_name).replace("'", ""))
        
    # Record the module's parent package, if it has one.
    if len(dotted_name) > 1:
        package_name = str(dotted_name.container())
        package = sys.modules.get(package_name)
        if package is not None:
            module_doc.package = introspect_docs(package)
    else:
        module_doc.package = None

    # Initialize the submodules property
    module_doc.submodules = []

    # Add the module to its parent package's submodules list.
    if module_doc.package not in (None, UNKNOWN):
        module_doc.package.submodules.append(module_doc)

    # Look up the module's __all__ attribute (public names).
    public_names = None
    if hasattr(module, '__all__'):
        try:
            public_names = set([str(name) for name in module.__all__])
        except KeyboardInterrupt: raise
        except: pass

    # Record the module's variables.
    module_doc.variables = {}
    for child_name in dir(module):
        if child_name in UNDOCUMENTED_MODULE_VARS: continue
        child = getattr(module, child_name)

        # Create a VariableDoc for the child, and introspect its
        # value if it's defined in this module.
        container = get_containing_module(child)
        if ((container is not None and
             container == name_without_primes) or
            (public_names is not None and
             child_name in public_names)):
            # Local variable.
            child_val_doc = introspect_docs(child, context=module_doc,
                                            module_name=dotted_name)
            child_var_doc = VariableDoc(name=child_name,
                                        value=child_val_doc,
                                        is_imported=False,
                                        container=module_doc,
                                        docs_extracted_by='introspecter')
        elif container is None or module_doc.canonical_name is UNKNOWN:

            # Don't introspect stuff "from __future__"
            if is_future_feature(child): continue

            # Possibly imported variable.
            child_val_doc = introspect_docs(child, context=module_doc)
            child_var_doc = VariableDoc(name=child_name,
                                        value=child_val_doc,
                                        container=module_doc,
                                        docs_extracted_by='introspecter')
        else:
            # Imported variable.
            child_val_doc = _get_valuedoc(child)
            child_var_doc = VariableDoc(name=child_name,
                                        value=child_val_doc,
                                        is_imported=True,
                                        container=module_doc,
                                        docs_extracted_by='introspecter')

        # If the module's __all__ attribute is set, use it to set the
        # variables public/private status and imported status.
        if public_names is not None:
            if child_name in public_names:
                child_var_doc.is_public = True
                if not isinstance(child_var_doc, ModuleDoc):
                    child_var_doc.is_imported = False
            else:
                child_var_doc.is_public = False

        module_doc.variables[child_name] = child_var_doc

    return module_doc

#////////////////////////////////////////////////////////////
# Class Introspection
#////////////////////////////////////////////////////////////

#: A list of class variables that should not be included in a
#: class's API documentation.
UNDOCUMENTED_CLASS_VARS = (
    '__doc__', '__module__', '__dict__', '__weakref__', '__slots__',
    '__pyx_vtable__')

def introspect_class(cls, class_doc, module_name=None):
    """
    Add API documentation information about the class C{cls}
    to C{class_doc}.
    """
    class_doc.specialize_to(ClassDoc)

    # Record the class's docstring.
    class_doc.docstring = get_docstring(cls)

    # Record the class's __all__ attribute (public names).
    public_names = None
    if hasattr(cls, '__all__'):
        try:
            public_names = set([str(name) for name in cls.__all__])
        except KeyboardInterrupt: raise
        except: pass

    # Start a list of subclasses.
    class_doc.subclasses = []

    # Sometimes users will define a __metaclass__ that copies all
    # class attributes from bases directly into the derived class's
    # __dict__ when the class is created.  (This saves the lookup time
    # needed to search the base tree for an attribute.)  But for the
    # docs, we only want to list these copied attributes in the
    # parent.  So only add an attribute if it is not identical to an
    # attribute of a base class.  (Unfortunately, this can sometimes
    # cause an attribute to look like it was inherited, even though it
    # wasn't, if it happens to have the exact same value as the
    # corresponding base's attribute.)  An example of a case where
    # this helps is PyQt -- subclasses of QWidget get about 300
    # methods injected into them.
    base_children = {}
    
    # Record the class's base classes; and add the class to its
    # base class's subclass lists.
    if hasattr(cls, '__bases__'):
        try: bases = list(cls.__bases__)
        except:
            bases = None
            log.warning("Class '%s' defines __bases__, but it does not "
                        "contain an iterable; ignoring base list."
                        % getattr(cls, '__name__', '??'))
        if bases is not None:
            class_doc.bases = []
            for base in bases:
                basedoc = introspect_docs(base)
                class_doc.bases.append(basedoc)
                basedoc.subclasses.append(class_doc)
            
            bases.reverse()
            for base in bases:
                if hasattr(base, '__dict__'):
                    base_children.update(base.__dict__)

    # The module name is not defined if the class is being introspected
    # as another class base.
    if module_name is None and class_doc.defining_module not in (None, UNKNOWN):
        module_name = class_doc.defining_module.canonical_name
        
    # Record the class's local variables.
    class_doc.variables = {}
    if hasattr(cls, '__dict__'):
        private_prefix = '_%s__' % getattr(cls, '__name__', '<none>')
        for child_name, child in cls.__dict__.items():
            if (child_name in base_children
                and base_children[child_name] == child):
                continue

            if child_name.startswith(private_prefix):
                child_name = child_name[len(private_prefix)-2:]
            if child_name in UNDOCUMENTED_CLASS_VARS: continue
            val_doc = introspect_docs(child, context=class_doc,
                                      module_name=module_name)
            var_doc = VariableDoc(name=child_name, value=val_doc,
                                  container=class_doc,
                                  docs_extracted_by='introspecter')
            if public_names is not None:
                var_doc.is_public = (child_name in public_names)
            class_doc.variables[child_name] = var_doc

    return class_doc

#////////////////////////////////////////////////////////////
# Routine Introspection
#////////////////////////////////////////////////////////////

def introspect_routine(routine, routine_doc, module_name=None):
    """Add API documentation information about the function
    C{routine} to C{routine_doc} (specializing it to C{Routine_doc})."""
    routine_doc.specialize_to(RoutineDoc)
    
    # Extract the underying function
    if isinstance(routine, MethodType):
        func = routine.im_func
    elif isinstance(routine, staticmethod):
        func = routine.__get__(0)
    elif isinstance(routine, classmethod):
        func = routine.__get__(0).im_func
    else:
        func = routine

    # Record the function's docstring.
    routine_doc.docstring = get_docstring(func)

    # Record the function's signature.
    if isinstance(func, FunctionType):
        (args, vararg, kwarg, defaults) = inspect.getargspec(func)

        # Add the arguments.
        routine_doc.posargs = args
        routine_doc.vararg = vararg
        routine_doc.kwarg = kwarg

        # Set default values for positional arguments.
        routine_doc.posarg_defaults = [None]*len(args)
        if defaults is not None:
            offset = len(args)-len(defaults)
            for i in range(len(defaults)):
                default_val = introspect_docs(defaults[i])
                routine_doc.posarg_defaults[i+offset] = default_val

        # If it's a bound method, then strip off the first argument.
        if isinstance(routine, MethodType) and routine.im_self is not None:
            routine_doc.posargs = routine_doc.posargs[1:]
            routine_doc.posarg_defaults = routine_doc.posarg_defaults[1:]

        # Set the routine's line number.
        if hasattr(func, 'func_code'):
            routine_doc.lineno = func.func_code.co_firstlineno

    else:
        # [XX] I should probably use UNKNOWN here??
        # dvarrazzo: if '...' is to be changed, also check that
        # `docstringparser.process_arg_field()` works correctly.
        # See SF bug #1556024.
        routine_doc.posargs = ['...']
        routine_doc.posarg_defaults = [None]
        routine_doc.kwarg = None
        routine_doc.vararg = None

    # Change type, if appropriate.
    if isinstance(routine, staticmethod):
        routine_doc.specialize_to(StaticMethodDoc)
    if isinstance(routine, classmethod):
        routine_doc.specialize_to(ClassMethodDoc)
        
    return routine_doc

#////////////////////////////////////////////////////////////
# Property Introspection
#////////////////////////////////////////////////////////////

def introspect_property(prop, prop_doc, module_name=None):
    """Add API documentation information about the property
    C{prop} to C{prop_doc} (specializing it to C{PropertyDoc})."""
    prop_doc.specialize_to(PropertyDoc)

    # Record the property's docstring.
    prop_doc.docstring = get_docstring(prop, module_name=module_name)

    # Record the property's access functions.
    if hasattr(prop, 'fget'):
        prop_doc.fget = introspect_docs(prop.fget)
        prop_doc.fset = introspect_docs(prop.fset)
        prop_doc.fdel = introspect_docs(prop.fdel)
    
    return prop_doc

#////////////////////////////////////////////////////////////
# Generic Value Introspection
#////////////////////////////////////////////////////////////

def introspect_other(val, val_doc, module_name=None):
    """Specialize val_doc to a C{GenericValueDoc} and return it."""
    val_doc.specialize_to(GenericValueDoc)
    return val_doc

#////////////////////////////////////////////////////////////
# Helper functions
#////////////////////////////////////////////////////////////

def isclass(object):
    """
    Return true if the given object is a class.  In particular, return
    true if object is an instance of C{types.TypeType} or of
    C{types.ClassType}.  This is used instead of C{inspect.isclass()},
    because the latter returns true for objects that are not classes
    (in particular, it returns true for any object that has a
    C{__bases__} attribute, including objects that define
    C{__getattr__} to always return a value).
    """
    return isinstance(object, tuple(_CLASS_TYPES))

_CLASS_TYPES = set([TypeType, ClassType])
"""A list of types that should be treated as classes."""

def register_class_type(typ):
    """Add a type to the lists of types that should be treated as
    classes.  By default, this list contains C{TypeType} and
    C{ClassType}."""
    _CLASS_TYPES.add(typ)

__future_check_works = None

def is_future_feature(object):
    """
    Return True if C{object} results from a C{from __future__ import feature}
    statement.
    """
    # Guard from unexpected implementation changes of the __future__ module.
    global __future_check_works
    if __future_check_works is not None:
        if __future_check_works:
            import __future__
            return isinstance(object, __future__._Feature)
        else:
            return False
    else:
        __future_check_works = True
        try:
            return is_future_feature(object)
        except:
            __future_check_works = False
            log.warning("Troubles inspecting __future__. Python implementation"
                        " may have been changed.")
            return False

def get_docstring(value, module_name=None):
    """
    Return the docstring for the given value; or C{None} if it
    does not have a docstring.
    @rtype: C{unicode}
    """
    docstring = getattr(value, '__doc__', None)
    if docstring is None:
        return None
    elif isinstance(docstring, unicode):
        return docstring
    elif isinstance(docstring, str):
        try: return unicode(docstring, 'ascii')
        except UnicodeDecodeError:
            if module_name is None:
                module_name = get_containing_module(value)
            if module_name is not None:
                try:
                    module = get_value_from_name(module_name)
                    filename = py_src_filename(module.__file__)
                    encoding = epydoc.docparser.get_module_encoding(filename)
                    return unicode(docstring, encoding)
                except KeyboardInterrupt: raise
                except Exception: pass
            if hasattr(value, '__name__'): name = value.__name__
            else: name = repr(value)
            log.warning("%s's docstring is not a unicode string, but it "
                        "contains non-ascii data -- treating it as "
                        "latin-1." % name)
            return unicode(docstring, 'latin-1')
        return None
    elif value is BuiltinMethodType:
        # Don't issue a warning for this special case.
        return None
    else:
        if hasattr(value, '__name__'): name = value.__name__
        else: name = repr(value)
        log.warning("%s's docstring is not a string -- ignoring it." %
                    name)
        return None

def get_canonical_name(value, strict=False):
    """
    @return: the canonical name for C{value}, or C{UNKNOWN} if no
    canonical name can be found.  Currently, C{get_canonical_name}
    can find canonical names for: modules; functions; non-nested
    classes; methods of non-nested classes; and some class methods
    of non-nested classes.
    
    @rtype: L{DottedName} or C{UNKNOWN}
    """
    if not hasattr(value, '__name__'): return UNKNOWN

    # Get the name via introspection.
    if isinstance(value, ModuleType):
        try:
            dotted_name = DottedName(value.__name__, strict=strict)
            # If the module is shadowed by a variable in its parent
            # package(s), then add a prime mark to the end, to
            # differentiate it from the variable that shadows it.
            if verify_name(value, dotted_name) is UNKNOWN:
                log.warning("Module %s is shadowed by a variable with "
                            "the same name." % dotted_name)
                # Note -- this return bypasses verify_name check:
                return DottedName(value.__name__+"'")
        except DottedName.InvalidDottedName:
            # Name is not a valid Python identifier -- treat as script.
            if hasattr(value, '__file__'):
                filename = '%s' % value.__str__
                dotted_name = DottedName(munge_script_name(filename))
        
    elif isclass(value):
        if value.__module__ == '__builtin__':
            dotted_name = DottedName(value.__name__, strict=strict)
        else:
            dotted_name = DottedName(value.__module__, value.__name__,
                                     strict=strict)
            
    elif (inspect.ismethod(value) and value.im_self is not None and
          value.im_class is ClassType and
          not value.__name__.startswith('<')): # class method.
        class_name = get_canonical_name(value.im_self)
        if class_name is UNKNOWN: return UNKNOWN
        dotted_name = DottedName(class_name, value.__name__, strict=strict)
    elif (inspect.ismethod(value) and
          not value.__name__.startswith('<')):
        class_name = get_canonical_name(value.im_class)
        if class_name is UNKNOWN: return UNKNOWN
        dotted_name = DottedName(class_name, value.__name__, strict=strict)
    elif (isinstance(value, FunctionType) and
          not value.__name__.startswith('<')):
        module_name = _find_function_module(value)
        if module_name is None: return UNKNOWN
        dotted_name = DottedName(module_name, value.__name__, strict=strict)
    else:
        return UNKNOWN

    return verify_name(value, dotted_name)

def verify_name(value, dotted_name):
    """
    Verify the name.  E.g., if it's a nested class, then we won't be
    able to find it with the name we constructed.
    """
    if dotted_name is UNKNOWN: return UNKNOWN
    if len(dotted_name) == 1 and hasattr(__builtin__, dotted_name[0]):
        return dotted_name
    named_value = sys.modules.get(dotted_name[0])
    if named_value is None: return UNKNOWN
    for identifier in dotted_name[1:]:
        try: named_value = getattr(named_value, identifier)
        except: return UNKNOWN
    if value is named_value:
        return dotted_name
    else:
        return UNKNOWN

# [xx] not used:
def value_repr(value):
    try:
        s = '%r' % value
        if isinstance(s, str):
            s = decode_with_backslashreplace(s)
        return s
    except:
        return UNKNOWN

def get_containing_module(value):
    """
    Return the name of the module containing the given value, or
    C{None} if the module name can't be determined.
    @rtype: L{DottedName}
    """
    if inspect.ismodule(value):
        return DottedName(value.__name__)
    elif isclass(value):
        return DottedName(value.__module__)
    elif (inspect.ismethod(value) and value.im_self is not None and
          value.im_class is ClassType): # class method.
        return DottedName(value.im_self.__module__)
    elif inspect.ismethod(value):
        return DottedName(value.im_class.__module__)
    elif inspect.isroutine(value):
        module = _find_function_module(value)
        if module is None: return None
        return DottedName(module)
    else:
        return None

def _find_function_module(func):
    """
    @return: The module that defines the given function.
    @rtype: C{module}
    @param func: The function whose module should be found.
    @type func: C{function}
    """
    if hasattr(func, '__module__'):
        return func.__module__
    try:
        module = inspect.getmodule(func)
        if module: return module.__name__
    except KeyboardInterrupt: raise
    except: pass

    # This fallback shouldn't usually be needed.  But it is needed in
    # a couple special cases (including using epydoc to document
    # itself).  In particular, if a module gets loaded twice, using
    # two different names for the same file, then this helps.
    for module in sys.modules.values():
        if (hasattr(module, '__dict__') and
            hasattr(func, 'func_globals') and
            func.func_globals is module.__dict__):
            return module.__name__
    return None

#////////////////////////////////////////////////////////////
# Introspection Dispatch Table
#////////////////////////////////////////////////////////////

_introspecter_registry = []
def register_introspecter(applicability_test, introspecter, priority=10):
    """
    Register an introspecter function.  Introspecter functions take
    two arguments, a python value and a C{ValueDoc} object, and should
    add information about the given value to the the C{ValueDoc}.
    Usually, the first line of an inspecter function will specialize
    it to a sublass of C{ValueDoc}, using L{ValueDoc.specialize_to()}:

        >>> def typical_introspecter(value, value_doc):
        ...     value_doc.specialize_to(SomeSubclassOfValueDoc)
        ...     <add info to value_doc>

    @param priority: The priority of this introspecter, which determines
    the order in which introspecters are tried -- introspecters with lower
    numbers are tried first.  The standard introspecters have priorities
    ranging from 20 to 30.  The default priority (10) will place new
    introspecters before standard introspecters.
    """
    _introspecter_registry.append( (priority, applicability_test,
                                    introspecter) )
    _introspecter_registry.sort()
    
def _get_introspecter(value):
    for (priority, applicability_test, introspecter) in _introspecter_registry:
        if applicability_test(value):
            return introspecter
    else:
        return introspect_other

# Register the standard introspecter functions.
def is_classmethod(v): return isinstance(v, classmethod)
def is_staticmethod(v): return isinstance(v, staticmethod)
def is_property(v): return isinstance(v, property)
register_introspecter(inspect.ismodule, introspect_module, priority=20)
register_introspecter(isclass, introspect_class, priority=24)
register_introspecter(inspect.isroutine, introspect_routine, priority=28)
register_introspecter(is_property, introspect_property, priority=30)

# Register getset_descriptor as a property
try:
    import array
    getset_type = type(array.array.typecode)
    del array
    def is_getset(v): return isinstance(v, getset_type)
    register_introspecter(is_getset, introspect_property, priority=32)
except:
    pass

# Register member_descriptor as a property
try:
    import datetime
    member_type = type(datetime.timedelta.days)
    del datetime
    def is_member(v): return isinstance(v, member_type)
    register_introspecter(is_member, introspect_property, priority=34)
except:
    pass

#////////////////////////////////////////////////////////////
# Import support
#////////////////////////////////////////////////////////////

def get_value_from_filename(filename, context=None):
    # Normalize the filename.
    filename = os.path.normpath(os.path.abspath(filename))

    # Divide the filename into a base directory and a name.  (For
    # packages, use the package's parent directory as the base, and
    # the directory name as its name).
    basedir = os.path.split(filename)[0]
    name = os.path.splitext(os.path.split(filename)[1])[0]
    if name == '__init__':
        basedir, name = os.path.split(basedir)
    name = DottedName(name)

    # If the context wasn't provided, then check if the file is in a
    # package directory.  If so, then update basedir & name to contain
    # the topmost package's directory and the fully qualified name for
    # this file.  (This update assume the default value of __path__
    # for the parent packages; if the parent packages override their
    # __path__s, then this can cause us not to find the value.)
    if context is None:
        while is_package_dir(basedir):
            basedir, pkg_name = os.path.split(basedir)
            name = DottedName(pkg_name, name)
            
    # If a parent package was specified, then find the directory of
    # the topmost package, and the fully qualified name for this file.
    if context is not None:
        # Combine the name.
        name = DottedName(context.canonical_name, name)
        # Find the directory of the base package.
        while context not in (None, UNKNOWN):
            pkg_dir = os.path.split(context.filename)[0]
            basedir = os.path.split(pkg_dir)[0]
            context = context.package

    # Import the module.  (basedir is the directory of the module's
    # topmost package, or its own directory if it's not in a package;
    # and name is the fully qualified dotted name for the module.)
    old_sys_path = sys.path[:]
    try:
        sys.path.insert(0, basedir)
        # This will make sure that we get the module itself, even
        # if it is shadowed by a variable.  (E.g., curses.wrapper):
        _import(str(name))
        if str(name) in sys.modules:
            return sys.modules[str(name)]
        else:
            # Use this as a fallback -- it *shouldn't* ever be needed.
            return get_value_from_name(name)
    finally:
        sys.path = old_sys_path

def get_value_from_scriptname(filename):
    name = munge_script_name(filename)
    return _import(name, filename)

def get_value_from_name(name, globs=None):
    """
    Given a name, return the corresponding value.
    
    @param globs: A namespace to check for the value, if there is no
        module containing the named value.  Defaults to __builtin__.
    """
    name = DottedName(name)

    # Import the topmost module/package.  If we fail, then check if
    # the requested name refers to a builtin.
    try:
        module = _import(name[0])
    except ImportError, e:
        if globs is None: globs = __builtin__.__dict__
        if name[0] in globs:
            try: return _lookup(globs[name[0]], name[1:])
            except: raise e
        else:
            raise

    # Find the requested value in the module/package or its submodules.
    for i in range(1, len(name)):
        try: return _lookup(module, name[i:])
        except ImportError: pass
        module = _import('.'.join(name[:i+1]))
        module = _lookup(module, name[1:i+1])
    return module

def _lookup(module, name):
    val = module
    for i, identifier in enumerate(name):
        try: val = getattr(val, identifier)
        except AttributeError:
            exc_msg = ('no variable named %s in %s' %
                       (identifier, '.'.join(name[:1+i])))
            raise ImportError(exc_msg)
    return val
            
def _import(name, filename=None):
    """
    Run the given callable in a 'sandboxed' environment.
    Currently, this includes saving and restoring the contents of
    sys and __builtins__; and suppressing stdin, stdout, and stderr.
    """
    # Note that we just do a shallow copy of sys.  In particular,
    # any changes made to sys.modules will be kept.  But we do
    # explicitly store sys.path.
    old_sys = sys.__dict__.copy()
    old_sys_path = sys.path[:]
    old_builtins = __builtin__.__dict__.copy()

    # Add the current directory to sys.path, in case they're trying to
    # import a module by name that resides in the current directory.
    # But add it to the end -- otherwise, the explicit directory added
    # in get_value_from_filename might get overwritten
    sys.path.append('')

    # Suppress input and output.  (These get restored when we restore
    # sys to old_sys).  
    sys.stdin = sys.stdout = sys.stderr = _dev_null
    sys.__stdin__ = sys.__stdout__ = sys.__stderr__ = _dev_null

    # Remove any command-line arguments
    sys.argv = ['(imported)']

    try:
        try:
            if filename is None:
                return __import__(name)
            else:
                # For importing scripts:
                return imp.load_source(name, filename)
        except KeyboardInterrupt: raise
        except:
            exc_typ, exc_val, exc_tb = sys.exc_info()
            if exc_val is None:
                estr = '%s' % (exc_typ,)
            else:
                estr = '%s: %s' % (exc_typ.__name__, exc_val)
            if exc_tb.tb_next is not None:
                estr += ' (line %d)' % (exc_tb.tb_next.tb_lineno,)
            raise ImportError(estr)
    finally:
        # Restore the important values that we saved.
        __builtin__.__dict__.clear()
        __builtin__.__dict__.update(old_builtins)
        sys.__dict__.clear()
        sys.__dict__.update(old_sys)
        sys.path = old_sys_path
        
def introspect_docstring_lineno(api_doc):
    """
    Try to determine the line number on which the given item's
    docstring begins.  Return the line number, or C{None} if the line
    number can't be determined.  The line number of the first line in
    the file is 1.
    """
    if api_doc.docstring_lineno is not UNKNOWN:
        return api_doc.docstring_lineno
    if isinstance(api_doc, ValueDoc) and api_doc.pyval is not UNKNOWN:
        try:
            lines, lineno = inspect.findsource(api_doc.pyval)
            if not isinstance(api_doc, ModuleDoc): lineno += 1
            for lineno in range(lineno, len(lines)):
                if lines[lineno].split('#', 1)[0].strip():
                    api_doc.docstring_lineno = lineno + 1
                    return lineno + 1
        except IOError: pass
        except TypeError: pass
        except IndexError:
            log.warning('inspect.findsource(%s) raised IndexError'
                        % api_doc.canonical_name)
    return None

class _DevNull:
    """
    A "file-like" object that discards anything that is written and
    always reports end-of-file when read.  C{_DevNull} is used by
    L{_import()} to discard output when importing modules; and to
    ensure that stdin appears closed.
    """
    def __init__(self):
        self.closed = 1
        self.mode = 'r+'
        self.softspace = 0
        self.name='</dev/null>'
    def close(self): pass
    def flush(self): pass
    def read(self, size=0): return ''
    def readline(self, size=0): return ''
    def readlines(self, sizehint=0): return []
    def seek(self, offset, whence=0): pass
    def tell(self): return 0L
    def truncate(self, size=0): pass
    def write(self, str): pass
    def writelines(self, sequence): pass
    xreadlines = readlines
_dev_null = _DevNull()
    
######################################################################
## Zope InterfaceClass
######################################################################

try:
    from zope.interface.interface import InterfaceClass as _ZopeInterfaceClass
    register_class_type(_ZopeInterfaceClass)
except:
    pass

######################################################################
## Zope Extension classes
######################################################################

try:
    # Register type(ExtensionClass.ExtensionClass)
    from ExtensionClass import ExtensionClass as _ExtensionClass
    _ZopeType = type(_ExtensionClass)
    def _is_zope_type(val):
        return isinstance(val, _ZopeType)
    register_introspecter(_is_zope_type, introspect_class)

    # Register ExtensionClass.*MethodType
    from ExtensionClass import PythonMethodType as _ZopeMethodType
    from ExtensionClass import ExtensionMethodType as _ZopeCMethodType
    def _is_zope_method(val):
        return isinstance(val, (_ZopeMethodType, _ZopeCMethodType))
    register_introspecter(_is_zope_method, introspect_routine)
except:
    pass
                         


    
# [xx]
0 # hm..  otherwise the following gets treated as a docstring!  ouch!
"""
######################################################################
## Zope Extension...
######################################################################
class ZopeIntrospecter(Introspecter):
    VALUEDOC_CLASSES = Introspecter.VALUEDOC_CLASSES.copy()
    VALUEDOC_CLASSES.update({
        'module': ZopeModuleDoc,
        'class': ZopeClassDoc,
        'interface': ZopeInterfaceDoc,
        'attribute': ZopeAttributeDoc,
        })
    
    def add_module_child(self, child, child_name, module_doc):
        if isinstance(child, zope.interfaces.Interface):
            module_doc.add_zope_interface(child_name)
        else:
            Introspecter.add_module_child(self, child, child_name, module_doc)

    def add_class_child(self, child, child_name, class_doc):
        if isinstance(child, zope.interfaces.Interface):
            class_doc.add_zope_interface(child_name)
        else:
            Introspecter.add_class_child(self, child, child_name, class_doc)

    def introspect_zope_interface(self, interface, interfacename):
        pass # etc...
"""