IPython/utils/traitlets.py

# encoding: utf-8
"""
A lightweight Traits like module.

This is designed to provide a lightweight, simple, pure Python version of
many of the capabilities of enthought.traits.  This includes:

* Validation
* Type specification with defaults
* Static and dynamic notification
* Basic predefined types
* An API that is similar to enthought.traits

We don't support:

* Delegation
* Automatic GUI generation
* A full set of trait types.  Most importantly, we don't provide container
  traits (list, dict, tuple) that can trigger notifications if their
  contents change.
* API compatibility with enthought.traits

There are also some important difference in our design:

* enthought.traits does not validate default values.  We do.

We choose to create this module because we need these capabilities, but
we need them to be pure Python so they work in all Python implementations,
including Jython and IronPython.

Inheritance diagram:

.. inheritance-diagram:: IPython.utils.traitlets
   :parts: 3
"""

# Copyright (c) IPython Development Team.
# Distributed under the terms of the Modified BSD License.
#
# Adapted from enthought.traits, Copyright (c) Enthought, Inc.,
# also under the terms of the Modified BSD License.

import contextlib
import inspect
import re
import sys
import types
from types import FunctionType
try:
    from types import ClassType, InstanceType
    ClassTypes = (ClassType, type)
except:
    ClassTypes = (type,)
from warnings import warn

from .getargspec import getargspec
from .importstring import import_item
from IPython.utils import py3compat
from IPython.utils import eventful
from IPython.utils.py3compat import iteritems, string_types
from IPython.testing.skipdoctest import skip_doctest

SequenceTypes = (list, tuple, set, frozenset)

#-----------------------------------------------------------------------------
# Basic classes
#-----------------------------------------------------------------------------


class NoDefaultSpecified ( object ): pass
NoDefaultSpecified = NoDefaultSpecified()


class Undefined ( object ): pass
Undefined = Undefined()

class TraitError(Exception):
    pass

#-----------------------------------------------------------------------------
# Utilities
#-----------------------------------------------------------------------------


def class_of ( object ):
    """ Returns a string containing the class name of an object with the
    correct indefinite article ('a' or 'an') preceding it (e.g., 'an Image',
    'a PlotValue').
    """
    if isinstance( object, py3compat.string_types ):
        return add_article( object )

    return add_article( object.__class__.__name__ )


def add_article ( name ):
    """ Returns a string containing the correct indefinite article ('a' or 'an')
    prefixed to the specified string.
    """
    if name[:1].lower() in 'aeiou':
       return 'an ' + name

    return 'a ' + name


def repr_type(obj):
    """ Return a string representation of a value and its type for readable
    error messages.
    """
    the_type = type(obj)
    if (not py3compat.PY3) and the_type is InstanceType:
        # Old-style class.
        the_type = obj.__class__
    msg = '%r %r' % (obj, the_type)
    return msg


def is_trait(t):
    """ Returns whether the given value is an instance or subclass of TraitType.
    """
    return (isinstance(t, TraitType) or
            (isinstance(t, type) and issubclass(t, TraitType)))


def parse_notifier_name(name):
    """Convert the name argument to a list of names.

    Examples
    --------

    >>> parse_notifier_name('a')
    ['a']
    >>> parse_notifier_name(['a','b'])
    ['a', 'b']
    >>> parse_notifier_name(None)
    ['anytrait']
    """
    if isinstance(name, string_types):
        return [name]
    elif name is None:
        return ['anytrait']
    elif isinstance(name, (list, tuple)):
        for n in name:
            assert isinstance(n, string_types), "names must be strings"
        return name


class _SimpleTest:
    def __init__ ( self, value ): self.value = value
    def __call__ ( self, test  ):
        return test == self.value
    def __repr__(self):
        return "<SimpleTest(%r)" % self.value
    def __str__(self):
        return self.__repr__()


def getmembers(object, predicate=None):
    """A safe version of inspect.getmembers that handles missing attributes.

    This is useful when there are descriptor based attributes that for
    some reason raise AttributeError even though they exist.  This happens
    in zope.inteface with the __provides__ attribute.
    """
    results = []
    for key in dir(object):
        try:
            value = getattr(object, key)
        except AttributeError:
            pass
        else:
            if not predicate or predicate(value):
                results.append((key, value))
    results.sort()
    return results

def _validate_link(*tuples):
    """Validate arguments for traitlet link functions"""
    for t in tuples:
        if not len(t) == 2:
            raise TypeError("Each linked traitlet must be specified as (HasTraits, 'trait_name'), not %r" % t)
        obj, trait_name = t
        if not isinstance(obj, HasTraits):
            raise TypeError("Each object must be HasTraits, not %r" % type(obj))
        if not trait_name in obj.traits():
            raise TypeError("%r has no trait %r" % (obj, trait_name))

@skip_doctest
class link(object):
    """Link traits from different objects together so they remain in sync.

    Parameters
    ----------
    *args : pairs of objects/attributes

    Examples
    --------

    >>> c = link((obj1, 'value'), (obj2, 'value'), (obj3, 'value'))
    >>> obj1.value = 5 # updates other objects as well
    """
    updating = False
    def __init__(self, *args):
        if len(args) < 2:
            raise TypeError('At least two traitlets must be provided.')
        _validate_link(*args)

        self.objects = {}

        initial = getattr(args[0][0], args[0][1])
        for obj, attr in args:
            setattr(obj, attr, initial)

            callback = self._make_closure(obj, attr)
            obj.on_trait_change(callback, attr)
            self.objects[(obj, attr)] = callback

    @contextlib.contextmanager
    def _busy_updating(self):
        self.updating = True
        try:
            yield
        finally:
            self.updating = False

    def _make_closure(self, sending_obj, sending_attr):
        def update(name, old, new):
            self._update(sending_obj, sending_attr, new)
        return update

    def _update(self, sending_obj, sending_attr, new):
        if self.updating:
            return
        with self._busy_updating():
            for obj, attr in self.objects.keys():
                setattr(obj, attr, new)

    def unlink(self):
        for key, callback in self.objects.items():
            (obj, attr) = key
            obj.on_trait_change(callback, attr, remove=True)

@skip_doctest
class directional_link(object):
    """Link the trait of a source object with traits of target objects.

    Parameters
    ----------
    source : pair of object, name
    targets : pairs of objects/attributes

    Examples
    --------

    >>> c = directional_link((src, 'value'), (tgt1, 'value'), (tgt2, 'value'))
    >>> src.value = 5  # updates target objects
    >>> tgt1.value = 6 # does not update other objects
    """
    updating = False

    def __init__(self, source, *targets):
        if len(targets) < 1:
            raise TypeError('At least two traitlets must be provided.')
        _validate_link(source, *targets)
        self.source = source
        self.targets = targets

        # Update current value
        src_attr_value = getattr(source[0], source[1])
        for obj, attr in targets:
            setattr(obj, attr, src_attr_value)

        # Wire
        self.source[0].on_trait_change(self._update, self.source[1])

    @contextlib.contextmanager
    def _busy_updating(self):
        self.updating = True
        try:
            yield
        finally:
            self.updating = False

    def _update(self, name, old, new):
        if self.updating:
            return
        with self._busy_updating():
            for obj, attr in self.targets:
                setattr(obj, attr, new)

    def unlink(self):
        self.source[0].on_trait_change(self._update, self.source[1], remove=True)
        self.source = None
        self.targets = []

dlink = directional_link


#-----------------------------------------------------------------------------
# Base TraitType for all traits
#-----------------------------------------------------------------------------


class TraitType(object):
    """A base class for all trait descriptors.

    Notes
    -----
    Our implementation of traits is based on Python's descriptor
    prototol.  This class is the base class for all such descriptors.  The
    only magic we use is a custom metaclass for the main :class:`HasTraits`
    class that does the following:

    1. Sets the :attr:`name` attribute of every :class:`TraitType`
       instance in the class dict to the name of the attribute.
    2. Sets the :attr:`this_class` attribute of every :class:`TraitType`
       instance in the class dict to the *class* that declared the trait.
       This is used by the :class:`This` trait to allow subclasses to
       accept superclasses for :class:`This` values.
    """


    metadata = {}
    default_value = Undefined
    allow_none = False
    info_text = 'any value'

    def __init__(self, default_value=NoDefaultSpecified, allow_none=None, **metadata):
        """Create a TraitType.
        """
        if default_value is not NoDefaultSpecified:
            self.default_value = default_value
        if allow_none is not None:
            self.allow_none = allow_none

        if 'default' in metadata:
            # Warn the user that they probably meant default_value.
            warn(
                "Parameter 'default' passed to TraitType. "
                "Did you mean 'default_value'?"
            )

        if len(metadata) > 0:
            if len(self.metadata) > 0:
                self._metadata = self.metadata.copy()
                self._metadata.update(metadata)
            else:
                self._metadata = metadata
        else:
            self._metadata = self.metadata

        self.init()

    def init(self):
        pass

    def get_default_value(self):
        """Create a new instance of the default value."""
        return self.default_value

    def instance_init(self):
        """Part of the initialization which may depends on the underlying
        HasTraits instance.

        It is typically overloaded for specific trait types.

        This method is called by :meth:`HasTraits.__new__` and in the
        :meth:`TraitType.instance_init` method of trait types holding
        other trait types.
        """
        pass

    def init_default_value(self, obj):
        """Instantiate the default value for the trait type.

        This method is called by :meth:`TraitType.set_default_value` in the
        case a default value is provided at construction time or later when
        accessing the trait value for the first time in
        :meth:`HasTraits.__get__`.
        """
        value = self.get_default_value()
        value = self._validate(obj, value)
        obj._trait_values[self.name] = value
        return value

    def set_default_value(self, obj):
        """Set the default value on a per instance basis.

        This method is called by :meth:`HasTraits.__new__` to instantiate and
        validate the default value. The creation and validation of
        default values must be delayed until the parent :class:`HasTraits`
        class has been instantiated.
        Parameters
        ----------
        obj : :class:`HasTraits` instance
            The parent :class:`HasTraits` instance that has just been
            created.
        """
        # Check for a deferred initializer defined in the same class as the
        # trait declaration or above.
        mro = type(obj).mro()
        meth_name = '_%s_default' % self.name
        for cls in mro[:mro.index(self.this_class)+1]:
            if meth_name in cls.__dict__:
                break
        else:
            # We didn't find one. Do static initialization.
            self.init_default_value(obj)
            return
        # Complete the dynamic initialization.
        obj._trait_dyn_inits[self.name] = meth_name

    def __get__(self, obj, cls=None):
        """Get the value of the trait by self.name for the instance.

        Default values are instantiated when :meth:`HasTraits.__new__`
        is called.  Thus by the time this method gets called either the
        default value or a user defined value (they called :meth:`__set__`)
        is in the :class:`HasTraits` instance.
        """
        if obj is None:
            return self
        else:
            try:
                value = obj._trait_values[self.name]
            except KeyError:
                # Check for a dynamic initializer.
                if self.name in obj._trait_dyn_inits:
                    method = getattr(obj, obj._trait_dyn_inits[self.name])
                    value = method()
                    # FIXME: Do we really validate here?
                    value = self._validate(obj, value)
                    obj._trait_values[self.name] = value
                    return value
                else:
                    return self.init_default_value(obj)
            except Exception:
                # HasTraits should call set_default_value to populate
                # this.  So this should never be reached.
                raise TraitError('Unexpected error in TraitType: '
                                 'default value not set properly')
            else:
                return value

    def __set__(self, obj, value):
        new_value = self._validate(obj, value)
        try:
            old_value = obj._trait_values[self.name]
        except KeyError:
            old_value = None

        obj._trait_values[self.name] = new_value
        try:
            silent = bool(old_value == new_value)
        except:
            # if there is an error in comparing, default to notify
            silent = False
        if silent is not True:
            # we explicitly compare silent to True just in case the equality
            # comparison above returns something other than True/False
            obj._notify_trait(self.name, old_value, new_value)

    def _validate(self, obj, value):
        if value is None and self.allow_none:
            return value
        if hasattr(self, 'validate'):
            value = self.validate(obj, value)
        try:
            obj_validate = getattr(obj, '_%s_validate' % self.name)
        except (AttributeError, RuntimeError):
            # Qt mixins raise RuntimeError on missing attrs accessed before __init__
            pass
        else:
            value = obj_validate(value, self)
        return value

    def __or__(self, other):
        if isinstance(other, Union):
            return Union([self] + other.trait_types)
        else:
            return Union([self, other])

    def info(self):
        return self.info_text

    def error(self, obj, value):
        if obj is not None:
            e = "The '%s' trait of %s instance must be %s, but a value of %s was specified." \
                % (self.name, class_of(obj),
                   self.info(), repr_type(value))
        else:
            e = "The '%s' trait must be %s, but a value of %r was specified." \
                % (self.name, self.info(), repr_type(value))
        raise TraitError(e)

    def get_metadata(self, key, default=None):
        return getattr(self, '_metadata', {}).get(key, default)

    def set_metadata(self, key, value):
        getattr(self, '_metadata', {})[key] = value


#-----------------------------------------------------------------------------
# The HasTraits implementation
#-----------------------------------------------------------------------------


class MetaHasTraits(type):
    """A metaclass for HasTraits.

    This metaclass makes sure that any TraitType class attributes are
    instantiated and sets their name attribute.
    """

    def __new__(mcls, name, bases, classdict):
        """Create the HasTraits class.

        This instantiates all TraitTypes in the class dict and sets their
        :attr:`name` attribute.
        """
        # print "MetaHasTraitlets (mcls, name): ", mcls, name
        # print "MetaHasTraitlets (bases): ", bases
        # print "MetaHasTraitlets (classdict): ", classdict
        for k,v in iteritems(classdict):
            if isinstance(v, TraitType):
                v.name = k
            elif inspect.isclass(v):
                if issubclass(v, TraitType):
                    vinst = v()
                    vinst.name = k
                    classdict[k] = vinst
        return super(MetaHasTraits, mcls).__new__(mcls, name, bases, classdict)

    def __init__(cls, name, bases, classdict):
        """Finish initializing the HasTraits class.

        This sets the :attr:`this_class` attribute of each TraitType in the
        class dict to the newly created class ``cls``.
        """
        for k, v in iteritems(classdict):
            if isinstance(v, TraitType):
                v.this_class = cls
        super(MetaHasTraits, cls).__init__(name, bases, classdict)

class HasTraits(py3compat.with_metaclass(MetaHasTraits, object)):

    def __new__(cls, *args, **kw):
        # This is needed because object.__new__ only accepts
        # the cls argument.
        new_meth = super(HasTraits, cls).__new__
        if new_meth is object.__new__:
            inst = new_meth(cls)
        else:
            inst = new_meth(cls, **kw)
        inst._trait_values = {}
        inst._trait_notifiers = {}
        inst._trait_dyn_inits = {}
        # Here we tell all the TraitType instances to set their default
        # values on the instance.
        for key in dir(cls):
            # Some descriptors raise AttributeError like zope.interface's
            # __provides__ attributes even though they exist.  This causes
            # AttributeErrors even though they are listed in dir(cls).
            try:
                value = getattr(cls, key)
            except AttributeError:
                pass
            else:
                if isinstance(value, TraitType):
                    value.instance_init()
                    if key not in kw:
                        value.set_default_value(inst)

        return inst

    def __init__(self, *args, **kw):
        # Allow trait values to be set using keyword arguments.
        # We need to use setattr for this to trigger validation and
        # notifications.
        
        with self.hold_trait_notifications():
            for key, value in iteritems(kw):
                setattr(self, key, value)
    
    @contextlib.contextmanager
    def hold_trait_notifications(self):
        """Context manager for bundling trait change notifications
        
        Use this when doing multiple trait assignments (init, config),
        to avoid race conditions in trait notifiers requesting other trait values.
        All trait notifications will fire after all values have been assigned.
        """
        _notify_trait = self._notify_trait
        notifications = []
        self._notify_trait = lambda *a: notifications.append(a)
        
        try:
            yield
        finally:
            self._notify_trait = _notify_trait
            if isinstance(_notify_trait, types.MethodType):
                # FIXME: remove when support is bumped to 3.4.
                # when original method is restored,
                # remove the redundant value from __dict__
                # (only used to preserve pickleability on Python < 3.4)
                self.__dict__.pop('_notify_trait', None)
        # trigger delayed notifications
        for args in notifications:
            self._notify_trait(*args)

    def _notify_trait(self, name, old_value, new_value):

        # First dynamic ones
        callables = []
        callables.extend(self._trait_notifiers.get(name,[]))
        callables.extend(self._trait_notifiers.get('anytrait',[]))

        # Now static ones
        try:
            cb = getattr(self, '_%s_changed' % name)
        except:
            pass
        else:
            callables.append(cb)

        # Call them all now
        for c in callables:
            # Traits catches and logs errors here.  I allow them to raise
            if callable(c):
                argspec = getargspec(c)

                nargs = len(argspec[0])
                # Bound methods have an additional 'self' argument
                # I don't know how to treat unbound methods, but they
                # can't really be used for callbacks.
                if isinstance(c, types.MethodType):
                    offset = -1
                else:
                    offset = 0
                if nargs + offset == 0:
                    c()
                elif nargs + offset == 1:
                    c(name)
                elif nargs + offset == 2:
                    c(name, new_value)
                elif nargs + offset == 3:
                    c(name, old_value, new_value)
                else:
                    raise TraitError('a trait changed callback '
                                        'must have 0-3 arguments.')
            else:
                raise TraitError('a trait changed callback '
                                    'must be callable.')


    def _add_notifiers(self, handler, name):
        if name not in self._trait_notifiers:
            nlist = []
            self._trait_notifiers[name] = nlist
        else:
            nlist = self._trait_notifiers[name]
        if handler not in nlist:
            nlist.append(handler)

    def _remove_notifiers(self, handler, name):
        if name in self._trait_notifiers:
            nlist = self._trait_notifiers[name]
            try:
                index = nlist.index(handler)
            except ValueError:
                pass
            else:
                del nlist[index]

    def on_trait_change(self, handler, name=None, remove=False):
        """Setup a handler to be called when a trait changes.

        This is used to setup dynamic notifications of trait changes.

        Static handlers can be created by creating methods on a HasTraits
        subclass with the naming convention '_[traitname]_changed'.  Thus,
        to create static handler for the trait 'a', create the method
        _a_changed(self, name, old, new) (fewer arguments can be used, see
        below).

        Parameters
        ----------
        handler : callable
            A callable that is called when a trait changes.  Its
            signature can be handler(), handler(name), handler(name, new)
            or handler(name, old, new).
        name : list, str, None
            If None, the handler will apply to all traits.  If a list
            of str, handler will apply to all names in the list.  If a
            str, the handler will apply just to that name.
        remove : bool
            If False (the default), then install the handler.  If True
            then unintall it.
        """
        if remove:
            names = parse_notifier_name(name)
            for n in names:
                self._remove_notifiers(handler, n)
        else:
            names = parse_notifier_name(name)
            for n in names:
                self._add_notifiers(handler, n)

    @classmethod
    def class_trait_names(cls, **metadata):
        """Get a list of all the names of this class' traits.

        This method is just like the :meth:`trait_names` method,
        but is unbound.
        """
        return cls.class_traits(**metadata).keys()

    @classmethod
    def class_traits(cls, **metadata):
        """Get a `dict` of all the traits of this class.  The dictionary
        is keyed on the name and the values are the TraitType objects.

        This method is just like the :meth:`traits` method, but is unbound.

        The TraitTypes returned don't know anything about the values
        that the various HasTrait's instances are holding.

        The metadata kwargs allow functions to be passed in which
        filter traits based on metadata values.  The functions should
        take a single value as an argument and return a boolean.  If
        any function returns False, then the trait is not included in
        the output.  This does not allow for any simple way of
        testing that a metadata name exists and has any
        value because get_metadata returns None if a metadata key
        doesn't exist.
        """
        traits = dict([memb for memb in getmembers(cls) if
                     isinstance(memb[1], TraitType)])

        if len(metadata) == 0:
            return traits

        for meta_name, meta_eval in metadata.items():
            if type(meta_eval) is not FunctionType:
                metadata[meta_name] = _SimpleTest(meta_eval)

        result = {}
        for name, trait in traits.items():
            for meta_name, meta_eval in metadata.items():
                if not meta_eval(trait.get_metadata(meta_name)):
                    break
            else:
                result[name] = trait

        return result

    def trait_names(self, **metadata):
        """Get a list of all the names of this class' traits."""
        return self.traits(**metadata).keys()

    def traits(self, **metadata):
        """Get a `dict` of all the traits of this class.  The dictionary
        is keyed on the name and the values are the TraitType objects.

        The TraitTypes returned don't know anything about the values
        that the various HasTrait's instances are holding.

        The metadata kwargs allow functions to be passed in which
        filter traits based on metadata values.  The functions should
        take a single value as an argument and return a boolean.  If
        any function returns False, then the trait is not included in
        the output.  This does not allow for any simple way of
        testing that a metadata name exists and has any
        value because get_metadata returns None if a metadata key
        doesn't exist.
        """
        traits = dict([memb for memb in getmembers(self.__class__) if
                     isinstance(memb[1], TraitType)])

        if len(metadata) == 0:
            return traits

        for meta_name, meta_eval in metadata.items():
            if type(meta_eval) is not FunctionType:
                metadata[meta_name] = _SimpleTest(meta_eval)

        result = {}
        for name, trait in traits.items():
            for meta_name, meta_eval in metadata.items():
                if not meta_eval(trait.get_metadata(meta_name)):
                    break
            else:
                result[name] = trait

        return result

    def trait_metadata(self, traitname, key, default=None):
        """Get metadata values for trait by key."""
        try:
            trait = getattr(self.__class__, traitname)
        except AttributeError:
            raise TraitError("Class %s does not have a trait named %s" %
                                (self.__class__.__name__, traitname))
        else:
            return trait.get_metadata(key, default)

#-----------------------------------------------------------------------------
# Actual TraitTypes implementations/subclasses
#-----------------------------------------------------------------------------

#-----------------------------------------------------------------------------
# TraitTypes subclasses for handling classes and instances of classes
#-----------------------------------------------------------------------------


class ClassBasedTraitType(TraitType):
    """
    A trait with error reporting and string -> type resolution for Type,
    Instance and This.
    """

    def _resolve_string(self, string):
        """
        Resolve a string supplied for a type into an actual object.
        """
        return import_item(string)

    def error(self, obj, value):
        kind = type(value)
        if (not py3compat.PY3) and kind is InstanceType:
            msg = 'class %s' % value.__class__.__name__
        else:
            msg = '%s (i.e. %s)' % ( str( kind )[1:-1], repr( value ) )

        if obj is not None:
            e = "The '%s' trait of %s instance must be %s, but a value of %s was specified." \
                % (self.name, class_of(obj),
                   self.info(), msg)
        else:
            e = "The '%s' trait must be %s, but a value of %r was specified." \
                % (self.name, self.info(), msg)

        raise TraitError(e)


class Type(ClassBasedTraitType):
    """A trait whose value must be a subclass of a specified class."""

    def __init__ (self, default_value=None, klass=None, allow_none=True, **metadata ):
        """Construct a Type trait

        A Type trait specifies that its values must be subclasses of
        a particular class.

        If only ``default_value`` is given, it is used for the ``klass`` as
        well.

        Parameters
        ----------
        default_value : class, str or None
            The default value must be a subclass of klass.  If an str,
            the str must be a fully specified class name, like 'foo.bar.Bah'.
            The string is resolved into real class, when the parent
            :class:`HasTraits` class is instantiated.
        klass : class, str, None
            Values of this trait must be a subclass of klass.  The klass
            may be specified in a string like: 'foo.bar.MyClass'.
            The string is resolved into real class, when the parent
            :class:`HasTraits` class is instantiated.
        allow_none : bool [ default True ]
            Indicates whether None is allowed as an assignable value. Even if
            ``False``, the default value may be ``None``.
        """
        if default_value is None:
            if klass is None:
                klass = object
        elif klass is None:
            klass = default_value

        if not (inspect.isclass(klass) or isinstance(klass, py3compat.string_types)):
            raise TraitError("A Type trait must specify a class.")

        self.klass       = klass

        super(Type, self).__init__(default_value, allow_none=allow_none, **metadata)

    def validate(self, obj, value):
        """Validates that the value is a valid object instance."""
        if isinstance(value, py3compat.string_types):
            try:
                value = self._resolve_string(value)
            except ImportError:
                raise TraitError("The '%s' trait of %s instance must be a type, but "
                                 "%r could not be imported" % (self.name, obj, value))
        try:
            if issubclass(value, self.klass):
                return value
        except:
            pass

        self.error(obj, value)

    def info(self):
        """ Returns a description of the trait."""
        if isinstance(self.klass, py3compat.string_types):
            klass = self.klass
        else:
            klass = self.klass.__name__
        result = 'a subclass of ' + klass
        if self.allow_none:
            return result + ' or None'
        return result

    def instance_init(self):
        self._resolve_classes()
        super(Type, self).instance_init()

    def _resolve_classes(self):
        if isinstance(self.klass, py3compat.string_types):
            self.klass = self._resolve_string(self.klass)
        if isinstance(self.default_value, py3compat.string_types):
            self.default_value = self._resolve_string(self.default_value)

    def get_default_value(self):
        return self.default_value


class DefaultValueGenerator(object):
    """A class for generating new default value instances."""

    def __init__(self, *args, **kw):
        self.args = args
        self.kw = kw

    def generate(self, klass):
        return klass(*self.args, **self.kw)


class Instance(ClassBasedTraitType):
    """A trait whose value must be an instance of a specified class.

    The value can also be an instance of a subclass of the specified class.

    Subclasses can declare default classes by overriding the klass attribute
    """

    klass = None

    def __init__(self, klass=None, args=None, kw=None,
                 allow_none=True, **metadata ):
        """Construct an Instance trait.

        This trait allows values that are instances of a particular
        class or its subclasses.  Our implementation is quite different
        from that of enthough.traits as we don't allow instances to be used
        for klass and we handle the ``args`` and ``kw`` arguments differently.

        Parameters
        ----------
        klass : class, str
            The class that forms the basis for the trait.  Class names
            can also be specified as strings, like 'foo.bar.Bar'.
        args : tuple
            Positional arguments for generating the default value.
        kw : dict
            Keyword arguments for generating the default value.
        allow_none : bool [default True]
            Indicates whether None is allowed as a value.

        Notes
        -----
        If both ``args`` and ``kw`` are None, then the default value is None.
        If ``args`` is a tuple and ``kw`` is a dict, then the default is
        created as ``klass(*args, **kw)``.  If exactly one of ``args`` or ``kw`` is
        None, the None is replaced by ``()`` or ``{}``, respectively.
        """
        if klass is None:
            klass = self.klass
        
        if (klass is not None) and (inspect.isclass(klass) or isinstance(klass, py3compat.string_types)):
            self.klass = klass
        else:
            raise TraitError('The klass attribute must be a class'
                                ' not: %r' % klass)

        # self.klass is a class, so handle default_value
        if args is None and kw is None:
            default_value = None
        else:
            if args is None:
                # kw is not None
                args = ()
            elif kw is None:
                # args is not None
                kw = {}

            if not isinstance(kw, dict):
                raise TraitError("The 'kw' argument must be a dict or None.")
            if not isinstance(args, tuple):
                raise TraitError("The 'args' argument must be a tuple or None.")

            default_value = DefaultValueGenerator(*args, **kw)

        super(Instance, self).__init__(default_value, allow_none=allow_none, **metadata)

    def validate(self, obj, value):
        if isinstance(value, self.klass):
            return value
        else:
            self.error(obj, value)

    def info(self):
        if isinstance(self.klass, py3compat.string_types):
            klass = self.klass
        else:
            klass = self.klass.__name__
        result = class_of(klass)
        if self.allow_none:
            return result + ' or None'

        return result

    def instance_init(self):
        self._resolve_classes()
        super(Instance, self).instance_init()

    def _resolve_classes(self):
        if isinstance(self.klass, py3compat.string_types):
            self.klass = self._resolve_string(self.klass)

    def get_default_value(self):
        """Instantiate a default value instance.

        This is called when the containing HasTraits classes'
        :meth:`__new__` method is called to ensure that a unique instance
        is created for each HasTraits instance.
        """
        dv  = self.default_value
        if isinstance(dv, DefaultValueGenerator):
            return dv.generate(self.klass)
        else:
            return dv


class ForwardDeclaredMixin(object):
    """
    Mixin for forward-declared versions of Instance and Type.
    """
    def _resolve_string(self, string):
        """
        Find the specified class name by looking for it in the module in which
        our this_class attribute was defined.
        """
        modname = self.this_class.__module__
        return import_item('.'.join([modname, string]))


class ForwardDeclaredType(ForwardDeclaredMixin, Type):
    """
    Forward-declared version of Type.
    """
    pass


class ForwardDeclaredInstance(ForwardDeclaredMixin, Instance):
    """
    Forward-declared version of Instance.
    """
    pass


class This(ClassBasedTraitType):
    """A trait for instances of the class containing this trait.

    Because how how and when class bodies are executed, the ``This``
    trait can only have a default value of None.  This, and because we
    always validate default values, ``allow_none`` is *always* true.
    """

    info_text = 'an instance of the same type as the receiver or None'

    def __init__(self, **metadata):
        super(This, self).__init__(None, **metadata)

    def validate(self, obj, value):
        # What if value is a superclass of obj.__class__?  This is
        # complicated if it was the superclass that defined the This
        # trait.
        if isinstance(value, self.this_class) or (value is None):
            return value
        else:
            self.error(obj, value)


class Union(TraitType):
    """A trait type representing a Union type."""

    def __init__(self, trait_types, **metadata):
        """Construct a Union  trait.

        This trait allows values that are allowed by at least one of the
        specified trait types. A Union traitlet cannot have metadata on
        its own, besides the metadata of the listed types.

        Parameters
        ----------
        trait_types: sequence
            The list of trait types of length at least 1.

        Notes
        -----
        Union([Float(), Bool(), Int()]) attempts to validate the provided values
        with the validation function of Float, then Bool, and finally Int.
        """
        self.trait_types = trait_types
        self.info_text = " or ".join([tt.info_text for tt in self.trait_types])
        self.default_value = self.trait_types[0].get_default_value()
        super(Union, self).__init__(**metadata)

    def instance_init(self):
        for trait_type in self.trait_types:
            trait_type.name = self.name
            trait_type.this_class = self.this_class
            trait_type.instance_init()
        super(Union, self).instance_init()

    def validate(self, obj, value):
        for trait_type in self.trait_types:
            try:
                v = trait_type._validate(obj, value)
                self._metadata = trait_type._metadata
                return v
            except TraitError:
                continue
        self.error(obj, value)

    def __or__(self, other):
        if isinstance(other, Union):
            return Union(self.trait_types + other.trait_types)
        else:
            return Union(self.trait_types + [other])

#-----------------------------------------------------------------------------
# Basic TraitTypes implementations/subclasses
#-----------------------------------------------------------------------------


class Any(TraitType):
    default_value = None
    info_text = 'any value'


class Int(TraitType):
    """An int trait."""

    default_value = 0
    info_text = 'an int'

    def validate(self, obj, value):
        if isinstance(value, int):
            return value
        self.error(obj, value)

class CInt(Int):
    """A casting version of the int trait."""

    def validate(self, obj, value):
        try:
            return int(value)
        except:
            self.error(obj, value)

if py3compat.PY3:
    Long, CLong = Int, CInt
    Integer = Int
else:
    class Long(TraitType):
        """A long integer trait."""

        default_value = 0
        info_text = 'a long'

        def validate(self, obj, value):
            if isinstance(value, long):
                return value
            if isinstance(value, int):
                return long(value)
            self.error(obj, value)


    class CLong(Long):
        """A casting version of the long integer trait."""

        def validate(self, obj, value):
            try:
                return long(value)
            except:
                self.error(obj, value)

    class Integer(TraitType):
        """An integer trait.

        Longs that are unnecessary (<= sys.maxint) are cast to ints."""

        default_value = 0
        info_text = 'an integer'

        def validate(self, obj, value):
            if isinstance(value, int):
                return value
            if isinstance(value, long):
                # downcast longs that fit in int:
                # note that int(n > sys.maxint) returns a long, so
                # we don't need a condition on this cast
                return int(value)
            if sys.platform == "cli":
                from System import Int64
                if isinstance(value, Int64):
                    return int(value)
            self.error(obj, value)


class Float(TraitType):
    """A float trait."""

    default_value = 0.0
    info_text = 'a float'

    def validate(self, obj, value):
        if isinstance(value, float):
            return value
        if isinstance(value, int):
            return float(value)
        self.error(obj, value)


class CFloat(Float):
    """A casting version of the float trait."""

    def validate(self, obj, value):
        try:
            return float(value)
        except:
            self.error(obj, value)

class Complex(TraitType):
    """A trait for complex numbers."""

    default_value = 0.0 + 0.0j
    info_text = 'a complex number'

    def validate(self, obj, value):
        if isinstance(value, complex):
            return value
        if isinstance(value, (float, int)):
            return complex(value)
        self.error(obj, value)


class CComplex(Complex):
    """A casting version of the complex number trait."""

    def validate (self, obj, value):
        try:
            return complex(value)
        except:
            self.error(obj, value)

# We should always be explicit about whether we're using bytes or unicode, both
# for Python 3 conversion and for reliable unicode behaviour on Python 2. So
# we don't have a Str type.
class Bytes(TraitType):
    """A trait for byte strings."""

    default_value = b''
    info_text = 'a bytes object'

    def validate(self, obj, value):
        if isinstance(value, bytes):
            return value
        self.error(obj, value)


class CBytes(Bytes):
    """A casting version of the byte string trait."""

    def validate(self, obj, value):
        try:
            return bytes(value)
        except:
            self.error(obj, value)


class Unicode(TraitType):
    """A trait for unicode strings."""

    default_value = u''
    info_text = 'a unicode string'

    def validate(self, obj, value):
        if isinstance(value, py3compat.unicode_type):
            return value
        if isinstance(value, bytes):
            try:
                return value.decode('ascii', 'strict')
            except UnicodeDecodeError:
                msg = "Could not decode {!r} for unicode trait '{}' of {} instance."
                raise TraitError(msg.format(value, self.name, class_of(obj)))
        self.error(obj, value)


class CUnicode(Unicode):
    """A casting version of the unicode trait."""

    def validate(self, obj, value):
        try:
            return py3compat.unicode_type(value)
        except:
            self.error(obj, value)


class ObjectName(TraitType):
    """A string holding a valid object name in this version of Python.

    This does not check that the name exists in any scope."""
    info_text = "a valid object identifier in Python"

    if py3compat.PY3:
        # Python 3:
        coerce_str = staticmethod(lambda _,s: s)

    else:
        # Python 2:
        def coerce_str(self, obj, value):
            "In Python 2, coerce ascii-only unicode to str"
            if isinstance(value, unicode):
                try:
                    return str(value)
                except UnicodeEncodeError:
                    self.error(obj, value)
            return value

    def validate(self, obj, value):
        value = self.coerce_str(obj, value)

        if isinstance(value, string_types) and py3compat.isidentifier(value):
            return value
        self.error(obj, value)

class DottedObjectName(ObjectName):
    """A string holding a valid dotted object name in Python, such as A.b3._c"""
    def validate(self, obj, value):
        value = self.coerce_str(obj, value)

        if isinstance(value, string_types) and py3compat.isidentifier(value, dotted=True):
            return value
        self.error(obj, value)


class Bool(TraitType):
    """A boolean (True, False) trait."""

    default_value = False
    info_text = 'a boolean'

    def validate(self, obj, value):
        if isinstance(value, bool):
            return value
        self.error(obj, value)


class CBool(Bool):
    """A casting version of the boolean trait."""

    def validate(self, obj, value):
        try:
            return bool(value)
        except:
            self.error(obj, value)


class Enum(TraitType):
    """An enum that whose value must be in a given sequence."""

    def __init__(self, values, default_value=None, **metadata):
        self.values = values
        super(Enum, self).__init__(default_value, **metadata)

    def validate(self, obj, value):
        if value in self.values:
                return value
        self.error(obj, value)

    def info(self):
        """ Returns a description of the trait."""
        result = 'any of ' + repr(self.values)
        if self.allow_none:
            return result + ' or None'
        return result

class CaselessStrEnum(Enum):
    """An enum of strings that are caseless in validate."""

    def validate(self, obj, value):
        if not isinstance(value, py3compat.string_types):
            self.error(obj, value)

        for v in self.values:
            if v.lower() == value.lower():
                return v
        self.error(obj, value)

class Container(Instance):
    """An instance of a container (list, set, etc.)

    To be subclassed by overriding klass.
    """
    klass = None
    _cast_types = ()
    _valid_defaults = SequenceTypes
    _trait = None

    def __init__(self, trait=None, default_value=None, allow_none=False,
                **metadata):
        """Create a container trait type from a list, set, or tuple.

        The default value is created by doing ``List(default_value)``,
        which creates a copy of the ``default_value``.

        ``trait`` can be specified, which restricts the type of elements
        in the container to that TraitType.

        If only one arg is given and it is not a Trait, it is taken as
        ``default_value``:

        ``c = List([1,2,3])``

        Parameters
        ----------

        trait : TraitType [ optional ]
            the type for restricting the contents of the Container.  If unspecified,
            types are not checked.

        default_value : SequenceType [ optional ]
            The default value for the Trait.  Must be list/tuple/set, and
            will be cast to the container type.

        allow_none : bool [ default False ]
            Whether to allow the value to be None

        **metadata : any
            further keys for extensions to the Trait (e.g. config)

        """
        # allow List([values]):
        if default_value is None and not is_trait(trait):
            default_value = trait
            trait = None

        if default_value is None:
            args = ()
        elif isinstance(default_value, self._valid_defaults):
            args = (default_value,)
        else:
            raise TypeError('default value of %s was %s' %(self.__class__.__name__, default_value))

        if is_trait(trait):
            self._trait = trait() if isinstance(trait, type) else trait
            self._trait.name = 'element'
        elif trait is not None:
            raise TypeError("`trait` must be a Trait or None, got %s"%repr_type(trait))

        super(Container,self).__init__(klass=self.klass, args=args,
                                  allow_none=allow_none, **metadata)

    def element_error(self, obj, element, validator):
        e = "Element of the '%s' trait of %s instance must be %s, but a value of %s was specified." \
            % (self.name, class_of(obj), validator.info(), repr_type(element))
        raise TraitError(e)

    def validate(self, obj, value):
        if isinstance(value, self._cast_types):
            value = self.klass(value)
        value = super(Container, self).validate(obj, value)
        if value is None:
            return value

        value = self.validate_elements(obj, value)

        return value

    def validate_elements(self, obj, value):
        validated = []
        if self._trait is None or isinstance(self._trait, Any):
            return value
        for v in value:
            try:
                v = self._trait._validate(obj, v)
            except TraitError:
                self.element_error(obj, v, self._trait)
            else:
                validated.append(v)
        return self.klass(validated)

    def instance_init(self):
        if isinstance(self._trait, TraitType):
            self._trait.this_class = self.this_class
            self._trait.instance_init()
        super(Container, self).instance_init()


class List(Container):
    """An instance of a Python list."""
    klass = list
    _cast_types = (tuple,)

    def __init__(self, trait=None, default_value=None, minlen=0, maxlen=sys.maxsize, **metadata):
        """Create a List trait type from a list, set, or tuple.

        The default value is created by doing ``List(default_value)``,
        which creates a copy of the ``default_value``.

        ``trait`` can be specified, which restricts the type of elements
        in the container to that TraitType.

        If only one arg is given and it is not a Trait, it is taken as
        ``default_value``:

        ``c = List([1,2,3])``

        Parameters
        ----------

        trait : TraitType [ optional ]
            the type for restricting the contents of the Container.  If unspecified,
            types are not checked.

        default_value : SequenceType [ optional ]
            The default value for the Trait.  Must be list/tuple/set, and
            will be cast to the container type.

        minlen : Int [ default 0 ]
            The minimum length of the input list

        maxlen : Int [ default sys.maxsize ]
            The maximum length of the input list

        allow_none : bool [ default False ]
            Whether to allow the value to be None

        **metadata : any
            further keys for extensions to the Trait (e.g. config)

        """
        self._minlen = minlen
        self._maxlen = maxlen
        super(List, self).__init__(trait=trait, default_value=default_value,
                                **metadata)

    def length_error(self, obj, value):
        e = "The '%s' trait of %s instance must be of length %i <= L <= %i, but a value of %s was specified." \
            % (self.name, class_of(obj), self._minlen, self._maxlen, value)
        raise TraitError(e)

    def validate_elements(self, obj, value):
        length = len(value)
        if length < self._minlen or length > self._maxlen:
            self.length_error(obj, value)

        return super(List, self).validate_elements(obj, value)

    def validate(self, obj, value):
        value = super(List, self).validate(obj, value)
        value = self.validate_elements(obj, value)
        return value


class Set(List):
    """An instance of a Python set."""
    klass = set
    _cast_types = (tuple, list)


class Tuple(Container):
    """An instance of a Python tuple."""
    klass = tuple
    _cast_types = (list,)

    def __init__(self, *traits, **metadata):
        """Tuple(*traits, default_value=None, **medatata)

        Create a tuple from a list, set, or tuple.

        Create a fixed-type tuple with Traits:

        ``t = Tuple(Int, Str, CStr)``

        would be length 3, with Int,Str,CStr for each element.

        If only one arg is given and it is not a Trait, it is taken as
        default_value:

        ``t = Tuple((1,2,3))``

        Otherwise, ``default_value`` *must* be specified by keyword.

        Parameters
        ----------

        *traits : TraitTypes [ optional ]
            the types for restricting the contents of the Tuple.  If unspecified,
            types are not checked. If specified, then each positional argument
            corresponds to an element of the tuple.  Tuples defined with traits
            are of fixed length.

        default_value : SequenceType [ optional ]
            The default value for the Tuple.  Must be list/tuple/set, and
            will be cast to a tuple. If `traits` are specified, the
            `default_value` must conform to the shape and type they specify.

        allow_none : bool [ default False ]
            Whether to allow the value to be None

        **metadata : any
            further keys for extensions to the Trait (e.g. config)

        """
        default_value = metadata.pop('default_value', None)
        allow_none = metadata.pop('allow_none', True)

        # allow Tuple((values,)):
        if len(traits) == 1 and default_value is None and not is_trait(traits[0]):
            default_value = traits[0]
            traits = ()

        if default_value is None:
            args = ()
        elif isinstance(default_value, self._valid_defaults):
            args = (default_value,)
        else:
            raise TypeError('default value of %s was %s' %(self.__class__.__name__, default_value))

        self._traits = []
        for trait in traits:
            t = trait() if isinstance(trait, type) else trait
            t.name = 'element'
            self._traits.append(t)

        if self._traits and default_value is None:
            # don't allow default to be an empty container if length is specified
            args = None
        super(Container,self).__init__(klass=self.klass, args=args, **metadata)

    def validate_elements(self, obj, value):
        if not self._traits:
            # nothing to validate
            return value
        if len(value) != len(self._traits):
            e = "The '%s' trait of %s instance requires %i elements, but a value of %s was specified." \
                % (self.name, class_of(obj), len(self._traits), repr_type(value))
            raise TraitError(e)

        validated = []
        for t, v in zip(self._traits, value):
            try:
                v = t._validate(obj, v)
            except TraitError:
                self.element_error(obj, v, t)
            else:
                validated.append(v)
        return tuple(validated)

    def instance_init(self):
        for trait in self._traits:
            if isinstance(trait, TraitType):
                trait.this_class = self.this_class
                trait.instance_init()
        super(Container, self).instance_init()


class Dict(Instance):
    """An instance of a Python dict."""
    _trait = None

    def __init__(self, trait=None, default_value=NoDefaultSpecified, allow_none=False, **metadata):
        """Create a dict trait type from a dict.

        The default value is created by doing ``dict(default_value)``,
        which creates a copy of the ``default_value``.

        trait : TraitType [ optional ]
            the type for restricting the contents of the Container.  If unspecified,
            types are not checked.

        default_value : SequenceType [ optional ]
            The default value for the Dict.  Must be dict, tuple, or None, and
            will be cast to a dict if not None. If `trait` is specified, the
            `default_value` must conform to the constraints it specifies.

        allow_none : bool [ default False ]
            Whether to allow the value to be None

        """
        if default_value is NoDefaultSpecified and trait is not None:
            if not is_trait(trait):
                default_value = trait
                trait = None
        if default_value is NoDefaultSpecified:
            default_value = {}
        if default_value is None:
            args = None
        elif isinstance(default_value, dict):
            args = (default_value,)
        elif isinstance(default_value, SequenceTypes):
            args = (default_value,)
        else:
            raise TypeError('default value of Dict was %s' % default_value)

        if is_trait(trait):
            self._trait = trait() if isinstance(trait, type) else trait
            self._trait.name = 'element'
        elif trait is not None:
            raise TypeError("`trait` must be a Trait or None, got %s"%repr_type(trait))

        super(Dict,self).__init__(klass=dict, args=args,
                                  allow_none=allow_none, **metadata)

    def element_error(self, obj, element, validator):
        e = "Element of the '%s' trait of %s instance must be %s, but a value of %s was specified." \
            % (self.name, class_of(obj), validator.info(), repr_type(element))
        raise TraitError(e)

    def validate(self, obj, value):
        value = super(Dict, self).validate(obj, value)
        if value is None:
            return value
        value = self.validate_elements(obj, value)
        return value

    def validate_elements(self, obj, value):
        if self._trait is None or isinstance(self._trait, Any):
            return value
        validated = {}
        for key in value:
            v = value[key]
            try:
                v = self._trait._validate(obj, v)
            except TraitError:
                self.element_error(obj, v, self._trait)
            else:
                validated[key] = v
        return self.klass(validated)

    def instance_init(self):
        if isinstance(self._trait, TraitType):
            self._trait.this_class = self.this_class
            self._trait.instance_init()
        super(Dict, self).instance_init()


class EventfulDict(Instance):
    """An instance of an EventfulDict."""

    def __init__(self, default_value={}, allow_none=False, **metadata):
        """Create a EventfulDict trait type from a dict.

        The default value is created by doing
        ``eventful.EvenfulDict(default_value)``, which creates a copy of the
        ``default_value``.
        """
        if default_value is None:
            args = None
        elif isinstance(default_value, dict):
            args = (default_value,)
        elif isinstance(default_value, SequenceTypes):
            args = (default_value,)
        else:
            raise TypeError('default value of EventfulDict was %s' % default_value)

        super(EventfulDict, self).__init__(klass=eventful.EventfulDict, args=args,
                                  allow_none=allow_none, **metadata)


class EventfulList(Instance):
    """An instance of an EventfulList."""

    def __init__(self, default_value=None, allow_none=False, **metadata):
        """Create a EventfulList trait type from a dict.

        The default value is created by doing 
        ``eventful.EvenfulList(default_value)``, which creates a copy of the 
        ``default_value``.
        """
        if default_value is None:
            args = ((),)
        else:
            args = (default_value,)

        super(EventfulList, self).__init__(klass=eventful.EventfulList, args=args,
                                  allow_none=allow_none, **metadata)


class TCPAddress(TraitType):
    """A trait for an (ip, port) tuple.

    This allows for both IPv4 IP addresses as well as hostnames.
    """

    default_value = ('127.0.0.1', 0)
    info_text = 'an (ip, port) tuple'

    def validate(self, obj, value):
        if isinstance(value, tuple):
            if len(value) == 2:
                if isinstance(value[0], py3compat.string_types) and isinstance(value[1], int):
                    port = value[1]
                    if port >= 0 and port <= 65535:
                        return value
        self.error(obj, value)

class CRegExp(TraitType):
    """A casting compiled regular expression trait.

    Accepts both strings and compiled regular expressions. The resulting
    attribute will be a compiled regular expression."""

    info_text = 'a regular expression'

    def validate(self, obj, value):
        try:
            return re.compile(value)
        except:
            self.error(obj, value)