Merge pull request #6044 from minrk/core.log

add IPython.utils.log.get_logger

IPython/utils/log.py

@@ -0,0 +1,25 b''
+"""Grab the global logger instance."""
+
+# Copyright (c) IPython Development Team.
+# Distributed under the terms of the Modified BSD License.
+
+import logging
+
+_logger = None
+
+def get_logger():
+    """Grab the global logger instance.
+    
+    If a global IPython Application is instantiated, grab its logger.
+    Otherwise, grab the root logger.
+    """
+    global _logger
+    
+    if _logger is None:
+        from IPython.config import Application
+        if Application.initialized():
+            _logger = Application.instance().log
+        else:
+            logging.basicConfig()
+            _logger = logging.getLogger()
+    return _logger

IPython/config/configurable.py

@@ -1,390 +1,366 b''
 # encoding: utf-8
-"""
-A base class for objects that are configurable.
-
-Inheritance diagram:
-
-.. inheritance-diagram:: IPython.config.configurable
-   :parts: 3
+"""A base class for objects that are configurable."""
 
-Authors:
+# Copyright (c) IPython Development Team.
+# Distributed under the terms of the Modified BSD License.
 
-* Brian Granger
-* Fernando Perez
-* Min RK
-"""
 from __future__ import print_function
 
-#-----------------------------------------------------------------------------
-#  Copyright (C) 2008-2011  The IPython Development Team
-#
-#  Distributed under the terms of the BSD License.  The full license is in
-#  the file COPYING, distributed as part of this software.
-#-----------------------------------------------------------------------------
-
-#-----------------------------------------------------------------------------
-# Imports
-#-----------------------------------------------------------------------------
-
 import logging
 from copy import deepcopy
 
 from .loader import Config, LazyConfigValue
 from IPython.utils.traitlets import HasTraits, Instance
 from IPython.utils.text import indent, wrap_paragraphs
 from IPython.utils.py3compat import iteritems
 
 
 #-----------------------------------------------------------------------------
 # Helper classes for Configurables
 #-----------------------------------------------------------------------------
 
 
 class ConfigurableError(Exception):
     pass
 
 
 class MultipleInstanceError(ConfigurableError):
     pass
 
 #-----------------------------------------------------------------------------
 # Configurable implementation
 #-----------------------------------------------------------------------------
 
 class Configurable(HasTraits):
 
     config = Instance(Config, (), {})
     parent = Instance('IPython.config.configurable.Configurable')
 
     def __init__(self, **kwargs):
         """Create a configurable given a config config.
 
         Parameters
         ----------
         config : Config
             If this is empty, default values are used. If config is a
             :class:`Config` instance, it will be used to configure the
             instance.
         parent : Configurable instance, optional
             The parent Configurable instance of this object.
 
         Notes
         -----
         Subclasses of Configurable must call the :meth:`__init__` method of
         :class:`Configurable` *before* doing anything else and using
         :func:`super`::
 
             class MyConfigurable(Configurable):
                 def __init__(self, config=None):
                     super(MyConfigurable, self).__init__(config=config)
                     # Then any other code you need to finish initialization.
 
         This ensures that instances will be configured properly.
         """
         parent = kwargs.pop('parent', None)
         if parent is not None:
             # config is implied from parent
             if kwargs.get('config', None) is None:
                 kwargs['config'] = parent.config
             self.parent = parent
         
         config = kwargs.pop('config', None)
         if config is not None:
             # We used to deepcopy, but for now we are trying to just save
             # by reference.  This *could* have side effects as all components
             # will share config. In fact, I did find such a side effect in
             # _config_changed below. If a config attribute value was a mutable type
             # all instances of a component were getting the same copy, effectively
             # making that a class attribute.
             # self.config = deepcopy(config)
             self.config = config
         # This should go second so individual keyword arguments override
         # the values in config.
         super(Configurable, self).__init__(**kwargs)
 
     #-------------------------------------------------------------------------
     # Static trait notifiations
     #-------------------------------------------------------------------------
     
     @classmethod
     def section_names(cls):
         """return section names as a list"""
         return  [c.__name__ for c in reversed(cls.__mro__) if
             issubclass(c, Configurable) and issubclass(cls, c)
         ]
     
     def _find_my_config(self, cfg):
         """extract my config from a global Config object
         
         will construct a Config object of only the config values that apply to me
         based on my mro(), as well as those of my parent(s) if they exist.
         
         If I am Bar and my parent is Foo, and their parent is Tim,
         this will return merge following config sections, in this order::
         
             [Bar, Foo.bar, Tim.Foo.Bar]
         
         With the last item being the highest priority.
         """
         cfgs = [cfg]
         if self.parent:
             cfgs.append(self.parent._find_my_config(cfg))
         my_config = Config()
         for c in cfgs:
             for sname in self.section_names():
                 # Don't do a blind getattr as that would cause the config to
                 # dynamically create the section with name Class.__name__.
                 if c._has_section(sname):
                     my_config.merge(c[sname])
         return my_config
     
     def _load_config(self, cfg, section_names=None, traits=None):
         """load traits from a Config object"""
         
         if traits is None:
             traits = self.traits(config=True)
         if section_names is None:
             section_names = self.section_names()
         
         my_config = self._find_my_config(cfg)
         for name, config_value in iteritems(my_config):
             if name in traits:
                 if isinstance(config_value, LazyConfigValue):
                     # ConfigValue is a wrapper for using append / update on containers
                     # without having to copy the 
                     initial = getattr(self, name)
                     config_value = config_value.get_value(initial)
                 # We have to do a deepcopy here if we don't deepcopy the entire
                 # config object. If we don't, a mutable config_value will be
                 # shared by all instances, effectively making it a class attribute.
                 setattr(self, name, deepcopy(config_value))
 
     def _config_changed(self, name, old, new):
         """Update all the class traits having ``config=True`` as metadata.
 
         For any class trait with a ``config`` metadata attribute that is
         ``True``, we update the trait with the value of the corresponding
         config entry.
         """
         # Get all traits with a config metadata entry that is True
         traits = self.traits(config=True)
 
         # We auto-load config section for this class as well as any parent
         # classes that are Configurable subclasses.  This starts with Configurable
         # and works down the mro loading the config for each section.
         section_names = self.section_names()
         self._load_config(new, traits=traits, section_names=section_names)
 
     def update_config(self, config):
         """Fire the traits events when the config is updated."""
         # Save a copy of the current config.
         newconfig = deepcopy(self.config)
         # Merge the new config into the current one.
         newconfig.merge(config)
         # Save the combined config as self.config, which triggers the traits
         # events.
         self.config = newconfig
 
     @classmethod
     def class_get_help(cls, inst=None):
         """Get the help string for this class in ReST format.
         
         If `inst` is given, it's current trait values will be used in place of
         class defaults.
         """
         assert inst is None or isinstance(inst, cls)
         final_help = []
         final_help.append(u'%s options' % cls.__name__)
         final_help.append(len(final_help[0])*u'-')
         for k, v in sorted(cls.class_traits(config=True).items()):
             help = cls.class_get_trait_help(v, inst)
             final_help.append(help)
         return '\n'.join(final_help)
 
     @classmethod
     def class_get_trait_help(cls, trait, inst=None):
         """Get the help string for a single trait.
         
         If `inst` is given, it's current trait values will be used in place of
         the class default.
         """
         assert inst is None or isinstance(inst, cls)
         lines = []
         header = "--%s.%s=<%s>" % (cls.__name__, trait.name, trait.__class__.__name__)
         lines.append(header)
         if inst is not None:
             lines.append(indent('Current: %r' % getattr(inst, trait.name), 4))
         else:
             try:
                 dvr = repr(trait.get_default_value())
             except Exception:
                 dvr = None # ignore defaults we can't construct
             if dvr is not None:
                 if len(dvr) > 64:
                     dvr = dvr[:61]+'...'
                 lines.append(indent('Default: %s' % dvr, 4))
         if 'Enum' in trait.__class__.__name__:
             # include Enum choices
             lines.append(indent('Choices: %r' % (trait.values,)))
 
         help = trait.get_metadata('help')
         if help is not None:
             help = '\n'.join(wrap_paragraphs(help, 76))
             lines.append(indent(help, 4))
         return '\n'.join(lines)
 
     @classmethod
     def class_print_help(cls, inst=None):
         """Get the help string for a single trait and print it."""
         print(cls.class_get_help(inst))
 
     @classmethod
     def class_config_section(cls):
         """Get the config class config section"""
         def c(s):
             """return a commented, wrapped block."""
             s = '\n\n'.join(wrap_paragraphs(s, 78))
 
             return '# ' + s.replace('\n', '\n# ')
 
         # section header
         breaker = '#' + '-'*78
         s = "# %s configuration" % cls.__name__
         lines = [breaker, s, breaker, '']
         # get the description trait
         desc = cls.class_traits().get('description')
         if desc:
             desc = desc.default_value
         else:
             # no description trait, use __doc__
             desc = getattr(cls, '__doc__', '')
         if desc:
             lines.append(c(desc))
             lines.append('')
 
         parents = []
         for parent in cls.mro():
             # only include parents that are not base classes
             # and are not the class itself
             # and have some configurable traits to inherit
             if parent is not cls and issubclass(parent, Configurable) and \
                     parent.class_traits(config=True):
                 parents.append(parent)
 
         if parents:
             pstr = ', '.join([ p.__name__ for p in parents ])
             lines.append(c('%s will inherit config from: %s'%(cls.__name__, pstr)))
             lines.append('')
 
         for name, trait in iteritems(cls.class_traits(config=True)):
             help = trait.get_metadata('help') or ''
             lines.append(c(help))
             lines.append('# c.%s.%s = %r'%(cls.__name__, name, trait.get_default_value()))
             lines.append('')
         return '\n'.join(lines)
 
 
 
 class SingletonConfigurable(Configurable):
     """A configurable that only allows one instance.
 
     This class is for classes that should only have one instance of itself
     or *any* subclass. To create and retrieve such a class use the
     :meth:`SingletonConfigurable.instance` method.
     """
 
     _instance = None
 
     @classmethod
     def _walk_mro(cls):
         """Walk the cls.mro() for parent classes that are also singletons
 
         For use in instance()
         """
 
         for subclass in cls.mro():
             if issubclass(cls, subclass) and \
                     issubclass(subclass, SingletonConfigurable) and \
                     subclass != SingletonConfigurable:
                 yield subclass
 
     @classmethod
     def clear_instance(cls):
         """unset _instance for this class and singleton parents.
         """
         if not cls.initialized():
             return
         for subclass in cls._walk_mro():
             if isinstance(subclass._instance, cls):
                 # only clear instances that are instances
                 # of the calling class
                 subclass._instance = None
 
     @classmethod
     def instance(cls, *args, **kwargs):
         """Returns a global instance of this class.
 
         This method create a new instance if none have previously been created
         and returns a previously created instance is one already exists.
 
         The arguments and keyword arguments passed to this method are passed
         on to the :meth:`__init__` method of the class upon instantiation.
 
         Examples
         --------
 
         Create a singleton class using instance, and retrieve it::
 
             >>> from IPython.config.configurable import SingletonConfigurable
             >>> class Foo(SingletonConfigurable): pass
             >>> foo = Foo.instance()
             >>> foo == Foo.instance()
             True
 
         Create a subclass that is retrived using the base class instance::
 
             >>> class Bar(SingletonConfigurable): pass
             >>> class Bam(Bar): pass
             >>> bam = Bam.instance()
             >>> bam == Bar.instance()
             True
         """
         # Create and save the instance
         if cls._instance is None:
             inst = cls(*args, **kwargs)
             # Now make sure that the instance will also be returned by
             # parent classes' _instance attribute.
             for subclass in cls._walk_mro():
                 subclass._instance = inst
 
         if isinstance(cls._instance, cls):
             return cls._instance
         else:
             raise MultipleInstanceError(
                 'Multiple incompatible subclass instances of '
                 '%s are being created.' % cls.__name__
             )
 
     @classmethod
     def initialized(cls):
         """Has an instance been created?"""
         return hasattr(cls, "_instance") and cls._instance is not None
 
 
 class LoggingConfigurable(Configurable):
     """A parent class for Configurables that log.
 
     Subclasses have a log trait, and the default behavior
-    is to get the logger from the currently running Application
-    via Application.instance().log.
+    is to get the logger from the currently running Application.
     """
 
     log = Instance('logging.Logger')
     def _log_default(self):
-        from IPython.config.application import Application
-        if Application.initialized():
-            return Application.instance().log
-        else:
-            return logging.getLogger()
+        from IPython.utils import log
+        return log.get_logger()
 
 

IPython/config/loader.py

@@ -1,846 +1,824 b''
-"""A simple configuration system.
+# encoding: utf-8
+"""A simple configuration system."""
 
-Inheritance diagram:
-
-.. inheritance-diagram:: IPython.config.loader
-   :parts: 3
-
-Authors
--------
-* Brian Granger
-* Fernando Perez
-* Min RK
-"""
-
-#-----------------------------------------------------------------------------
-#  Copyright (C) 2008-2011  The IPython Development Team
-#
-#  Distributed under the terms of the BSD License.  The full license is in
-#  the file COPYING, distributed as part of this software.
-#-----------------------------------------------------------------------------
-
-#-----------------------------------------------------------------------------
-# Imports
-#-----------------------------------------------------------------------------
+# Copyright (c) IPython Development Team.
+# Distributed under the terms of the Modified BSD License.
 
 import argparse
 import copy
 import logging
 import os
 import re
 import sys
 import json
 
 from IPython.utils.path import filefind, get_ipython_dir
 from IPython.utils import py3compat
 from IPython.utils.encoding import DEFAULT_ENCODING
 from IPython.utils.py3compat import unicode_type, iteritems
 from IPython.utils.traitlets import HasTraits, List, Any
 
 #-----------------------------------------------------------------------------
 # Exceptions
 #-----------------------------------------------------------------------------
 
 
 class ConfigError(Exception):
     pass
 
 class ConfigLoaderError(ConfigError):
     pass
 
 class ConfigFileNotFound(ConfigError):
     pass
 
 class ArgumentError(ConfigLoaderError):
     pass
 
 #-----------------------------------------------------------------------------
 # Argparse fix
 #-----------------------------------------------------------------------------
 
 # Unfortunately argparse by default prints help messages to stderr instead of
 # stdout.  This makes it annoying to capture long help screens at the command
 # line, since one must know how to pipe stderr, which many users don't know how
 # to do.  So we override the print_help method with one that defaults to
 # stdout and use our class instead.
 
 class ArgumentParser(argparse.ArgumentParser):
     """Simple argparse subclass that prints help to stdout by default."""
 
     def print_help(self, file=None):
         if file is None:
             file = sys.stdout
         return super(ArgumentParser, self).print_help(file)
 
     print_help.__doc__ = argparse.ArgumentParser.print_help.__doc__
 
 #-----------------------------------------------------------------------------
 # Config class for holding config information
 #-----------------------------------------------------------------------------
 
 class LazyConfigValue(HasTraits):
     """Proxy object for exposing methods on configurable containers
     
     Exposes:
     
     - append, extend, insert on lists
     - update on dicts
     - update, add on sets
     """
     
     _value = None
     
     # list methods
     _extend = List()
     _prepend = List()
     
     def append(self, obj):
         self._extend.append(obj)
     
     def extend(self, other):
         self._extend.extend(other)
     
     def prepend(self, other):
         """like list.extend, but for the front"""
         self._prepend[:0] = other
     
     _inserts = List()
     def insert(self, index, other):
         if not isinstance(index, int):
             raise TypeError("An integer is required")
         self._inserts.append((index, other))
     
     # dict methods
     # update is used for both dict and set
     _update = Any()
     def update(self, other):
         if self._update is None:
             if isinstance(other, dict):
                 self._update = {}
             else:
                 self._update = set()
         self._update.update(other)
     
     # set methods
     def add(self, obj):
         self.update({obj})
     
     def get_value(self, initial):
         """construct the value from the initial one
         
         after applying any insert / extend / update changes
         """
         if self._value is not None:
             return self._value
         value = copy.deepcopy(initial)
         if isinstance(value, list):
             for idx, obj in self._inserts:
                 value.insert(idx, obj)
             value[:0] = self._prepend
             value.extend(self._extend)
         
         elif isinstance(value, dict):
             if self._update:
                 value.update(self._update)
         elif isinstance(value, set):
             if self._update:
                 value.update(self._update)
         self._value = value
         return value
     
     def to_dict(self):
         """return JSONable dict form of my data
         
         Currently update as dict or set, extend, prepend as lists, and inserts as list of tuples.
         """
         d = {}
         if self._update:
             d['update'] = self._update
         if self._extend:
             d['extend'] = self._extend
         if self._prepend:
             d['prepend'] = self._prepend
         elif self._inserts:
             d['inserts'] = self._inserts
         return d
 
 
 def _is_section_key(key):
     """Is a Config key a section name (does it start with a capital)?"""
     if key and key[0].upper()==key[0] and not key.startswith('_'):
         return True
     else:
         return False
 
 
 class Config(dict):
     """An attribute based dict that can do smart merges."""
 
     def __init__(self, *args, **kwds):
         dict.__init__(self, *args, **kwds)
         self._ensure_subconfig()
     
     def _ensure_subconfig(self):
         """ensure that sub-dicts that should be Config objects are
         
         casts dicts that are under section keys to Config objects,
         which is necessary for constructing Config objects from dict literals.
         """
         for key in self:
             obj = self[key]
             if _is_section_key(key) \
                     and isinstance(obj, dict) \
                     and not isinstance(obj, Config):
                 setattr(self, key, Config(obj))
     
     def _merge(self, other):
         """deprecated alias, use Config.merge()"""
         self.merge(other)
     
     def merge(self, other):
         """merge another config object into this one"""
         to_update = {}
         for k, v in iteritems(other):
             if k not in self:
                 to_update[k] = copy.deepcopy(v)
             else: # I have this key
                 if isinstance(v, Config) and isinstance(self[k], Config):
                     # Recursively merge common sub Configs
                     self[k].merge(v)
                 else:
                     # Plain updates for non-Configs
                     to_update[k] = copy.deepcopy(v)
 
         self.update(to_update)
 
     def __contains__(self, key):
         # allow nested contains of the form `"Section.key" in config`
         if '.' in key:
             first, remainder = key.split('.', 1)
             if first not in self:
                 return False
             return remainder in self[first]
         
         return super(Config, self).__contains__(key)
     
     # .has_key is deprecated for dictionaries.
     has_key = __contains__
     
     def _has_section(self, key):
         return _is_section_key(key) and key in self
     
     def copy(self):
         return type(self)(dict.copy(self))
 
     def __copy__(self):
         return self.copy()
 
     def __deepcopy__(self, memo):
         import copy
         return type(self)(copy.deepcopy(list(self.items())))
     
     def __getitem__(self, key):
         try:
             return dict.__getitem__(self, key)
         except KeyError:
             if _is_section_key(key):
                 c = Config()
                 dict.__setitem__(self, key, c)
                 return c
             elif not key.startswith('_'):
                 # undefined, create lazy value, used for container methods
                 v = LazyConfigValue()
                 dict.__setitem__(self, key, v)
                 return v
             else:
                 raise KeyError
 
     def __setitem__(self, key, value):
         if _is_section_key(key):
             if not isinstance(value, Config):
                 raise ValueError('values whose keys begin with an uppercase '
                                  'char must be Config instances: %r, %r' % (key, value))
         dict.__setitem__(self, key, value)
 
     def __getattr__(self, key):
         if key.startswith('__'):
             return dict.__getattr__(self, key)
         try:
             return self.__getitem__(key)
         except KeyError as e:
             raise AttributeError(e)
 
     def __setattr__(self, key, value):
         if key.startswith('__'):
             return dict.__setattr__(self, key, value)
         try:
             self.__setitem__(key, value)
         except KeyError as e:
             raise AttributeError(e)
 
     def __delattr__(self, key):
         if key.startswith('__'):
             return dict.__delattr__(self, key)
         try:
             dict.__delitem__(self, key)
         except KeyError as e:
             raise AttributeError(e)
 
 
 #-----------------------------------------------------------------------------
 # Config loading classes
 #-----------------------------------------------------------------------------
 
 
 class ConfigLoader(object):
     """A object for loading configurations from just about anywhere.
 
     The resulting configuration is packaged as a :class:`Config`.
 
     Notes
     -----
     A :class:`ConfigLoader` does one thing: load a config from a source
     (file, command line arguments) and returns the data as a :class:`Config` object.
     There are lots of things that :class:`ConfigLoader` does not do.  It does
     not implement complex logic for finding config files.  It does not handle
     default values or merge multiple configs.  These things need to be
     handled elsewhere.
     """
 
     def _log_default(self):
-        from IPython.config.application import Application
-        if Application.initialized():
-            return Application.instance().log
-        else:
-            return logging.getLogger()
+        from IPython.utils.log import get_logger
+        return get_logger()
 
     def __init__(self, log=None):
         """A base class for config loaders.
 
         log : instance of :class:`logging.Logger` to use.
               By default loger of :meth:`IPython.config.application.Application.instance()`
               will be used
 
         Examples
         --------
 
         >>> cl = ConfigLoader()
         >>> config = cl.load_config()
         >>> config
         {}
         """
         self.clear()
         if log is None:
             self.log = self._log_default()
             self.log.debug('Using default logger')
         else:
             self.log = log
 
     def clear(self):
         self.config = Config()
 
     def load_config(self):
         """Load a config from somewhere, return a :class:`Config` instance.
 
         Usually, this will cause self.config to be set and then returned.
         However, in most cases, :meth:`ConfigLoader.clear` should be called
         to erase any previous state.
         """
         self.clear()
         return self.config
 
 
 class FileConfigLoader(ConfigLoader):
     """A base class for file based configurations.
 
     As we add more file based config loaders, the common logic should go
     here.
     """
 
     def __init__(self, filename, path=None, **kw):
         """Build a config loader for a filename and path.
 
         Parameters
         ----------
         filename : str
             The file name of the config file.
         path : str, list, tuple
             The path to search for the config file on, or a sequence of
             paths to try in order.
         """
         super(FileConfigLoader, self).__init__(**kw)
         self.filename = filename
         self.path = path
         self.full_filename = ''
 
     def _find_file(self):
         """Try to find the file by searching the paths."""
         self.full_filename = filefind(self.filename, self.path)
 
 class JSONFileConfigLoader(FileConfigLoader):
     """A Json file loader for config"""
 
     def load_config(self):
         """Load the config from a file and return it as a Config object."""
         self.clear()
         try:
             self._find_file()
         except IOError as e:
             raise ConfigFileNotFound(str(e))
         dct = self._read_file_as_dict()
         self.config = self._convert_to_config(dct)
         return self.config
 
     def _read_file_as_dict(self):
         with open(self.full_filename) as f:
             return json.load(f)
 
     def _convert_to_config(self, dictionary):
         if 'version' in dictionary:
             version = dictionary.pop('version')
         else:
             version = 1
             self.log.warn("Unrecognized JSON config file version, assuming version {}".format(version))
 
         if version == 1:
             return Config(dictionary)
         else:
             raise ValueError('Unknown version of JSON config file: {version}'.format(version=version))
 
 
 class PyFileConfigLoader(FileConfigLoader):
     """A config loader for pure python files.
 
     This is responsible for locating a Python config file by filename and
     path, then executing it to construct a Config object.
     """
 
     def load_config(self):
         """Load the config from a file and return it as a Config object."""
         self.clear()
         try:
             self._find_file()
         except IOError as e:
             raise ConfigFileNotFound(str(e))
         self._read_file_as_dict()
         return self.config
 
 
     def _read_file_as_dict(self):
         """Load the config file into self.config, with recursive loading."""
         # This closure is made available in the namespace that is used
         # to exec the config file.  It allows users to call
         # load_subconfig('myconfig.py') to load config files recursively.
         # It needs to be a closure because it has references to self.path
         # and self.config.  The sub-config is loaded with the same path
         # as the parent, but it uses an empty config which is then merged
         # with the parents.
 
         # If a profile is specified, the config file will be loaded
         # from that profile
 
         def load_subconfig(fname, profile=None):
             # import here to prevent circular imports
             from IPython.core.profiledir import ProfileDir, ProfileDirError
             if profile is not None:
                 try:
                     profile_dir = ProfileDir.find_profile_dir_by_name(
                             get_ipython_dir(),
                             profile,
                     )
                 except ProfileDirError:
                     return
                 path = profile_dir.location
             else:
                 path = self.path
             loader = PyFileConfigLoader(fname, path)
             try:
                 sub_config = loader.load_config()
             except ConfigFileNotFound:
                 # Pass silently if the sub config is not there. This happens
                 # when a user s using a profile, but not the default config.
                 pass
             else:
                 self.config.merge(sub_config)
 
         # Again, this needs to be a closure and should be used in config
         # files to get the config being loaded.
         def get_config():
             return self.config
 
         namespace = dict(
             load_subconfig=load_subconfig,
             get_config=get_config,
             __file__=self.full_filename,
         )
         fs_encoding = sys.getfilesystemencoding() or 'ascii'
         conf_filename = self.full_filename.encode(fs_encoding)
         py3compat.execfile(conf_filename, namespace)
 
 
 class CommandLineConfigLoader(ConfigLoader):
     """A config loader for command line arguments.
 
     As we add more command line based loaders, the common logic should go
     here.
     """
 
     def _exec_config_str(self, lhs, rhs):
         """execute self.config.<lhs> = <rhs>
         
         * expands ~ with expanduser
         * tries to assign with raw eval, otherwise assigns with just the string,
           allowing `--C.a=foobar` and `--C.a="foobar"` to be equivalent.  *Not*
           equivalent are `--C.a=4` and `--C.a='4'`.
         """
         rhs = os.path.expanduser(rhs)
         try:
             # Try to see if regular Python syntax will work. This
             # won't handle strings as the quote marks are removed
             # by the system shell.
             value = eval(rhs)
         except (NameError, SyntaxError):
             # This case happens if the rhs is a string.
             value = rhs
 
         exec(u'self.config.%s = value' % lhs)
 
     def _load_flag(self, cfg):
         """update self.config from a flag, which can be a dict or Config"""
         if isinstance(cfg, (dict, Config)):
             # don't clobber whole config sections, update
             # each section from config:
             for sec,c in iteritems(cfg):
                 self.config[sec].update(c)
         else:
             raise TypeError("Invalid flag: %r" % cfg)
 
 # raw --identifier=value pattern
 # but *also* accept '-' as wordsep, for aliases
 # accepts:  --foo=a
 #           --Class.trait=value
 #           --alias-name=value
 # rejects:  -foo=value
 #           --foo
 #           --Class.trait
 kv_pattern = re.compile(r'\-\-[A-Za-z][\w\-]*(\.[\w\-]+)*\=.*')
 
 # just flags, no assignments, with two *or one* leading '-'
 # accepts:  --foo
 #           -foo-bar-again
 # rejects:  --anything=anything
 #           --two.word
 
 flag_pattern = re.compile(r'\-\-?\w+[\-\w]*$')
 
 class KeyValueConfigLoader(CommandLineConfigLoader):
     """A config loader that loads key value pairs from the command line.
 
     This allows command line options to be gives in the following form::
 
         ipython --profile="foo" --InteractiveShell.autocall=False
     """
 
     def __init__(self, argv=None, aliases=None, flags=None, **kw):
         """Create a key value pair config loader.
 
         Parameters
         ----------
         argv : list
             A list that has the form of sys.argv[1:] which has unicode
             elements of the form u"key=value". If this is None (default),
             then sys.argv[1:] will be used.
         aliases : dict
             A dict of aliases for configurable traits.
             Keys are the short aliases, Values are the resolved trait.
             Of the form: `{'alias' : 'Configurable.trait'}`
         flags : dict
             A dict of flags, keyed by str name. Vaues can be Config objects,
             dicts, or "key=value" strings.  If Config or dict, when the flag
             is triggered, The flag is loaded as `self.config.update(m)`.
 
         Returns
         -------
         config : Config
             The resulting Config object.
 
         Examples
         --------
 
             >>> from IPython.config.loader import KeyValueConfigLoader
             >>> cl = KeyValueConfigLoader()
             >>> d = cl.load_config(["--A.name='brian'","--B.number=0"])
             >>> sorted(d.items())
             [('A', {'name': 'brian'}), ('B', {'number': 0})]
         """
         super(KeyValueConfigLoader, self).__init__(**kw)
         if argv is None:
             argv = sys.argv[1:]
         self.argv = argv
         self.aliases = aliases or {}
         self.flags = flags or {}
 
 
     def clear(self):
         super(KeyValueConfigLoader, self).clear()
         self.extra_args = []
 
 
     def _decode_argv(self, argv, enc=None):
         """decode argv if bytes, using stin.encoding, falling back on default enc"""
         uargv = []
         if enc is None:
             enc = DEFAULT_ENCODING
         for arg in argv:
             if not isinstance(arg, unicode_type):
                 # only decode if not already decoded
                 arg = arg.decode(enc)
             uargv.append(arg)
         return uargv
 
 
     def load_config(self, argv=None, aliases=None, flags=None):
         """Parse the configuration and generate the Config object.
 
         After loading, any arguments that are not key-value or
         flags will be stored in self.extra_args - a list of
         unparsed command-line arguments.  This is used for
         arguments such as input files or subcommands.
 
         Parameters
         ----------
         argv : list, optional
             A list that has the form of sys.argv[1:] which has unicode
             elements of the form u"key=value". If this is None (default),
             then self.argv will be used.
         aliases : dict
             A dict of aliases for configurable traits.
             Keys are the short aliases, Values are the resolved trait.
             Of the form: `{'alias' : 'Configurable.trait'}`
         flags : dict
             A dict of flags, keyed by str name. Values can be Config objects
             or dicts.  When the flag is triggered, The config is loaded as
             `self.config.update(cfg)`.
         """
         self.clear()
         if argv is None:
             argv = self.argv
         if aliases is None:
             aliases = self.aliases
         if flags is None:
             flags = self.flags
 
         # ensure argv is a list of unicode strings:
         uargv = self._decode_argv(argv)
         for idx,raw in enumerate(uargv):
             # strip leading '-'
             item = raw.lstrip('-')
 
             if raw == '--':
                 # don't parse arguments after '--'
                 # this is useful for relaying arguments to scripts, e.g.
                 # ipython -i foo.py --matplotlib=qt -- args after '--' go-to-foo.py
                 self.extra_args.extend(uargv[idx+1:])
                 break
 
             if kv_pattern.match(raw):
                 lhs,rhs = item.split('=',1)
                 # Substitute longnames for aliases.
                 if lhs in aliases:
                     lhs = aliases[lhs]
                 if '.' not in lhs:
                     # probably a mistyped alias, but not technically illegal
                     self.log.warn("Unrecognized alias: '%s', it will probably have no effect.", raw)
                 try:
                     self._exec_config_str(lhs, rhs)
                 except Exception:
                     raise ArgumentError("Invalid argument: '%s'" % raw)
 
             elif flag_pattern.match(raw):
                 if item in flags:
                     cfg,help = flags[item]
                     self._load_flag(cfg)
                 else:
                     raise ArgumentError("Unrecognized flag: '%s'"%raw)
             elif raw.startswith('-'):
                 kv = '--'+item
                 if kv_pattern.match(kv):
                     raise ArgumentError("Invalid argument: '%s', did you mean '%s'?"%(raw, kv))
                 else:
                     raise ArgumentError("Invalid argument: '%s'"%raw)
             else:
                 # keep all args that aren't valid in a list,
                 # in case our parent knows what to do with them.
                 self.extra_args.append(item)
         return self.config
 
 class ArgParseConfigLoader(CommandLineConfigLoader):
     """A loader that uses the argparse module to load from the command line."""
 
     def __init__(self, argv=None, aliases=None, flags=None, log=None,  *parser_args, **parser_kw):
         """Create a config loader for use with argparse.
 
         Parameters
         ----------
 
         argv : optional, list
           If given, used to read command-line arguments from, otherwise
           sys.argv[1:] is used.
 
         parser_args : tuple
           A tuple of positional arguments that will be passed to the
           constructor of :class:`argparse.ArgumentParser`.
 
         parser_kw : dict
           A tuple of keyword arguments that will be passed to the
           constructor of :class:`argparse.ArgumentParser`.
 
         Returns
         -------
         config : Config
             The resulting Config object.
         """
         super(CommandLineConfigLoader, self).__init__(log=log)
         self.clear()
         if argv is None:
             argv = sys.argv[1:]
         self.argv = argv
         self.aliases = aliases or {}
         self.flags = flags or {}
 
         self.parser_args = parser_args
         self.version = parser_kw.pop("version", None)
         kwargs = dict(argument_default=argparse.SUPPRESS)
         kwargs.update(parser_kw)
         self.parser_kw = kwargs
 
     def load_config(self, argv=None, aliases=None, flags=None):
         """Parse command line arguments and return as a Config object.
 
         Parameters
         ----------
 
         args : optional, list
           If given, a list with the structure of sys.argv[1:] to parse
           arguments from. If not given, the instance's self.argv attribute
           (given at construction time) is used."""
         self.clear()
         if argv is None:
             argv = self.argv
         if aliases is None:
             aliases = self.aliases
         if flags is None:
             flags = self.flags
         self._create_parser(aliases, flags)
         self._parse_args(argv)
         self._convert_to_config()
         return self.config
 
     def get_extra_args(self):
         if hasattr(self, 'extra_args'):
             return self.extra_args
         else:
             return []
 
     def _create_parser(self, aliases=None, flags=None):
         self.parser = ArgumentParser(*self.parser_args, **self.parser_kw)
         self._add_arguments(aliases, flags)
 
     def _add_arguments(self, aliases=None, flags=None):
         raise NotImplementedError("subclasses must implement _add_arguments")
 
     def _parse_args(self, args):
         """self.parser->self.parsed_data"""
         # decode sys.argv to support unicode command-line options
         enc = DEFAULT_ENCODING
         uargs = [py3compat.cast_unicode(a, enc) for a in args]
         self.parsed_data, self.extra_args = self.parser.parse_known_args(uargs)
 
     def _convert_to_config(self):
         """self.parsed_data->self.config"""
         for k, v in iteritems(vars(self.parsed_data)):
             exec("self.config.%s = v"%k, locals(), globals())
 
 class KVArgParseConfigLoader(ArgParseConfigLoader):
     """A config loader that loads aliases and flags with argparse,
     but will use KVLoader for the rest.  This allows better parsing
     of common args, such as `ipython -c 'print 5'`, but still gets
     arbitrary config with `ipython --InteractiveShell.use_readline=False`"""
 
     def _add_arguments(self, aliases=None, flags=None):
         self.alias_flags = {}
         # print aliases, flags
         if aliases is None:
             aliases = self.aliases
         if flags is None:
             flags = self.flags
         paa = self.parser.add_argument
         for key,value in iteritems(aliases):
             if key in flags:
                 # flags
                 nargs = '?'
             else:
                 nargs = None
             if len(key) is 1:
                 paa('-'+key, '--'+key, type=unicode_type, dest=value, nargs=nargs)
             else:
                 paa('--'+key, type=unicode_type, dest=value, nargs=nargs)
         for key, (value, help) in iteritems(flags):
             if key in self.aliases:
                 #
                 self.alias_flags[self.aliases[key]] = value
                 continue
             if len(key) is 1:
                 paa('-'+key, '--'+key, action='append_const', dest='_flags', const=value)
             else:
                 paa('--'+key, action='append_const', dest='_flags', const=value)
 
     def _convert_to_config(self):
         """self.parsed_data->self.config, parse unrecognized extra args via KVLoader."""
         # remove subconfigs list from namespace before transforming the Namespace
         if '_flags' in self.parsed_data:
             subcs = self.parsed_data._flags
             del self.parsed_data._flags
         else:
             subcs = []
 
         for k, v in iteritems(vars(self.parsed_data)):
             if v is None:
                 # it was a flag that shares the name of an alias
                 subcs.append(self.alias_flags[k])
             else:
                 # eval the KV assignment
                 self._exec_config_str(k, v)
 
         for subc in subcs:
             self._load_flag(subc)
 
         if self.extra_args:
             sub_parser = KeyValueConfigLoader(log=self.log)
             sub_parser.load_config(self.extra_args)
             self.config.merge(sub_parser.config)
             self.extra_args = sub_parser.extra_args
 
 
 def load_pyconfig_files(config_files, path):
     """Load multiple Python config files, merging each of them in turn.
 
     Parameters
     ==========
     config_files : list of str
         List of config files names to load and merge into the config.
     path : unicode
         The full path to the location of the config files.
     """
     config = Config()
     for cf in config_files:
         loader = PyFileConfigLoader(cf, path=path)
         try:
             next_config = loader.load_config()
         except ConfigFileNotFound:
             pass
         except:
             raise
         else:
             config.merge(next_config)
     return config

IPython/html/base/handlers.py

@@ -1,390 +1,376 b''
-"""Base Tornado handlers for the notebook.
-
-Authors:
-
-* Brian Granger
-"""
-
-#-----------------------------------------------------------------------------
-#  Copyright (C) 2011  The IPython Development Team
-#
-#  Distributed under the terms of the BSD License.  The full license is in
-#  the file COPYING, distributed as part of this software.
-#-----------------------------------------------------------------------------
-
-#-----------------------------------------------------------------------------
-# Imports
-#-----------------------------------------------------------------------------
+"""Base Tornado handlers for the notebook."""
 
+# Copyright (c) IPython Development Team.
+# Distributed under the terms of the Modified BSD License.
 
 import functools
 import json
 import logging
 import os
 import re
 import sys
 import traceback
 try:
     # py3
     from http.client import responses
 except ImportError:
     from httplib import responses
 
 from jinja2 import TemplateNotFound
 from tornado import web
 
 try:
     from tornado.log import app_log
 except ImportError:
     app_log = logging.getLogger()
 
 from IPython.config import Application
 from IPython.utils.path import filefind
 from IPython.utils.py3compat import string_types
 from IPython.html.utils import is_hidden
 
 #-----------------------------------------------------------------------------
 # Top-level handlers
 #-----------------------------------------------------------------------------
 non_alphanum = re.compile(r'[^A-Za-z0-9]')
 
 class AuthenticatedHandler(web.RequestHandler):
     """A RequestHandler with an authenticated user."""
 
     def set_default_headers(self):
         headers = self.settings.get('headers', {})
         for header_name,value in headers.items() :
             try:
                 self.set_header(header_name, value)
             except Exception:
                 # tornado raise Exception (not a subclass)
                 # if method is unsupported (websocket and Access-Control-Allow-Origin
                 # for example, so just ignore)
                 pass
     
     def clear_login_cookie(self):
         self.clear_cookie(self.cookie_name)
     
     def get_current_user(self):
         user_id = self.get_secure_cookie(self.cookie_name)
         # For now the user_id should not return empty, but it could eventually
         if user_id == '':
             user_id = 'anonymous'
         if user_id is None:
             # prevent extra Invalid cookie sig warnings:
             self.clear_login_cookie()
             if not self.login_available:
                 user_id = 'anonymous'
         return user_id
 
     @property
     def cookie_name(self):
         default_cookie_name = non_alphanum.sub('-', 'username-{}'.format(
             self.request.host
         ))
         return self.settings.get('cookie_name', default_cookie_name)
     
     @property
     def password(self):
         """our password"""
         return self.settings.get('password', '')
     
     @property
     def logged_in(self):
         """Is a user currently logged in?
 
         """
         user = self.get_current_user()
         return (user and not user == 'anonymous')
 
     @property
     def login_available(self):
         """May a user proceed to log in?
 
         This returns True if login capability is available, irrespective of
         whether the user is already logged in or not.
 
         """
         return bool(self.settings.get('password', ''))
 
 
 class IPythonHandler(AuthenticatedHandler):
     """IPython-specific extensions to authenticated handling
     
     Mostly property shortcuts to IPython-specific settings.
     """
     
     @property
     def config(self):
         return self.settings.get('config', None)
     
     @property
     def log(self):
         """use the IPython log by default, falling back on tornado's logger"""
         if Application.initialized():
             return Application.instance().log
         else:
             return app_log
     
     #---------------------------------------------------------------
     # URLs
     #---------------------------------------------------------------
     
     @property
     def mathjax_url(self):
         return self.settings.get('mathjax_url', '')
     
     @property
     def base_url(self):
         return self.settings.get('base_url', '/')
     
     #---------------------------------------------------------------
     # Manager objects
     #---------------------------------------------------------------
     
     @property
     def kernel_manager(self):
         return self.settings['kernel_manager']
 
     @property
     def notebook_manager(self):
         return self.settings['notebook_manager']
     
     @property
     def cluster_manager(self):
         return self.settings['cluster_manager']
     
     @property
     def session_manager(self):
         return self.settings['session_manager']
     
     @property
     def kernel_spec_manager(self):
         return self.settings['kernel_spec_manager']
 
     @property
     def project_dir(self):
         return self.notebook_manager.notebook_dir
     
     #---------------------------------------------------------------
     # template rendering
     #---------------------------------------------------------------
     
     def get_template(self, name):
         """Return the jinja template object for a given name"""
         return self.settings['jinja2_env'].get_template(name)
     
     def render_template(self, name, **ns):
         ns.update(self.template_namespace)
         template = self.get_template(name)
         return template.render(**ns)
     
     @property
     def template_namespace(self):
         return dict(
             base_url=self.base_url,
             logged_in=self.logged_in,
             login_available=self.login_available,
             static_url=self.static_url,
         )
     
     def get_json_body(self):
         """Return the body of the request as JSON data."""
         if not self.request.body:
             return None
         # Do we need to call body.decode('utf-8') here?
         body = self.request.body.strip().decode(u'utf-8')
         try:
             model = json.loads(body)
         except Exception:
             self.log.debug("Bad JSON: %r", body)
             self.log.error("Couldn't parse JSON", exc_info=True)
             raise web.HTTPError(400, u'Invalid JSON in body of request')
         return model
 
     def get_error_html(self, status_code, **kwargs):
         """render custom error pages"""
         exception = kwargs.get('exception')
         message = ''
         status_message = responses.get(status_code, 'Unknown HTTP Error')
         if exception:
             # get the custom message, if defined
             try:
                 message = exception.log_message % exception.args
             except Exception:
                 pass
             
             # construct the custom reason, if defined
             reason = getattr(exception, 'reason', '')
             if reason:
                 status_message = reason
         
         # build template namespace
         ns = dict(
             status_code=status_code,
             status_message=status_message,
             message=message,
             exception=exception,
         )
         
         # render the template
         try:
             html = self.render_template('%s.html' % status_code, **ns)
         except TemplateNotFound:
             self.log.debug("No template for %d", status_code)
             html = self.render_template('error.html', **ns)
         return html
 
 
 class Template404(IPythonHandler):
     """Render our 404 template"""
     def prepare(self):
         raise web.HTTPError(404)
 
 
 class AuthenticatedFileHandler(IPythonHandler, web.StaticFileHandler):
     """static files should only be accessible when logged in"""
 
     @web.authenticated
     def get(self, path):
         if os.path.splitext(path)[1] == '.ipynb':
             name = os.path.basename(path)
             self.set_header('Content-Type', 'application/json')
             self.set_header('Content-Disposition','attachment; filename="%s"' % name)
         
         return web.StaticFileHandler.get(self, path)
     
     def compute_etag(self):
         return None
     
     def validate_absolute_path(self, root, absolute_path):
         """Validate and return the absolute path.
         
         Requires tornado 3.1
         
         Adding to tornado's own handling, forbids the serving of hidden files.
         """
         abs_path = super(AuthenticatedFileHandler, self).validate_absolute_path(root, absolute_path)
         abs_root = os.path.abspath(root)
         if is_hidden(abs_path, abs_root):
             self.log.info("Refusing to serve hidden file, via 404 Error")
             raise web.HTTPError(404)
         return abs_path
 
 
 def json_errors(method):
     """Decorate methods with this to return GitHub style JSON errors.
     
     This should be used on any JSON API on any handler method that can raise HTTPErrors.
     
     This will grab the latest HTTPError exception using sys.exc_info
     and then:
     
     1. Set the HTTP status code based on the HTTPError
     2. Create and return a JSON body with a message field describing
        the error in a human readable form.
     """
     @functools.wraps(method)
     def wrapper(self, *args, **kwargs):
         try:
             result = method(self, *args, **kwargs)
         except web.HTTPError as e:
             status = e.status_code
             message = e.log_message
             self.log.warn(message)
             self.set_status(e.status_code)
             self.finish(json.dumps(dict(message=message)))
         except Exception:
             self.log.error("Unhandled error in API request", exc_info=True)
             status = 500
             message = "Unknown server error"
             t, value, tb = sys.exc_info()
             self.set_status(status)
             tb_text = ''.join(traceback.format_exception(t, value, tb))
             reply = dict(message=message, traceback=tb_text)
             self.finish(json.dumps(reply))
         else:
             return result
     return wrapper
 
 
 
 #-----------------------------------------------------------------------------
 # File handler
 #-----------------------------------------------------------------------------
 
 # to minimize subclass changes:
 HTTPError = web.HTTPError
 
 class FileFindHandler(web.StaticFileHandler):
     """subclass of StaticFileHandler for serving files from a search path"""
     
     # cache search results, don't search for files more than once
     _static_paths = {}
     
     def initialize(self, path, default_filename=None):
         if isinstance(path, string_types):
             path = [path]
         
         self.root = tuple(
             os.path.abspath(os.path.expanduser(p)) + os.sep for p in path
         )
         self.default_filename = default_filename
     
     def compute_etag(self):
         return None
     
     @classmethod
     def get_absolute_path(cls, roots, path):
         """locate a file to serve on our static file search path"""
         with cls._lock:
             if path in cls._static_paths:
                 return cls._static_paths[path]
             try:
                 abspath = os.path.abspath(filefind(path, roots))
             except IOError:
                 # IOError means not found
                 return ''
             
             cls._static_paths[path] = abspath
             return abspath
     
     def validate_absolute_path(self, root, absolute_path):
         """check if the file should be served (raises 404, 403, etc.)"""
         if absolute_path == '':
             raise web.HTTPError(404)
         
         for root in self.root:
             if (absolute_path + os.sep).startswith(root):
                 break
         
         return super(FileFindHandler, self).validate_absolute_path(root, absolute_path)
 
 
 class TrailingSlashHandler(web.RequestHandler):
     """Simple redirect handler that strips trailing slashes
     
     This should be the first, highest priority handler.
     """
     
     SUPPORTED_METHODS = ['GET']
     
     def get(self):
         self.redirect(self.request.uri.rstrip('/'))
 
 #-----------------------------------------------------------------------------
 # URL pattern fragments for re-use
 #-----------------------------------------------------------------------------
 
 path_regex = r"(?P<path>(?:/.*)*)"
 notebook_name_regex = r"(?P<name>[^/]+\.ipynb)"
 notebook_path_regex = "%s/%s" % (path_regex, notebook_name_regex)
 
 #-----------------------------------------------------------------------------
 # URL to handler mappings
 #-----------------------------------------------------------------------------
 
 
 default_handlers = [
     (r".*/", TrailingSlashHandler)
 ]

IPython/nbformat/current.py

@@ -1,238 +1,217 b''
-"""The official API for working with notebooks in the current format version.
-
-Authors:
-
-* Brian Granger
-* Jonathan Frederic
-"""
-
-#-----------------------------------------------------------------------------
-#  Copyright (C) 2008-2011  The IPython Development Team
-#
-#  Distributed under the terms of the BSD License.  The full license is in
-#  the file COPYING, distributed as part of this software.
-#-----------------------------------------------------------------------------
-
-#-----------------------------------------------------------------------------
-# Imports
-#-----------------------------------------------------------------------------
+"""The official API for working with notebooks in the current format version."""
 
 from __future__ import print_function
 
 from xml.etree import ElementTree as ET
 import re
 
 from IPython.utils.py3compat import unicode_type
 
 from IPython.nbformat.v3 import (
     NotebookNode,
     new_code_cell, new_text_cell, new_notebook, new_output, new_worksheet,
     parse_filename, new_metadata, new_author, new_heading_cell, nbformat,
     nbformat_minor, nbformat_schema, to_notebook_json
 )
 from IPython.nbformat import v3 as _v_latest
 
 from .reader import reads as reader_reads
 from .reader import versions
 from .convert import convert
 from .validator import validate
 
-import logging
-logger = logging.getLogger('NotebookApp')
+from IPython.utils.log import get_logger
 
-#-----------------------------------------------------------------------------
-# Code
-#-----------------------------------------------------------------------------
 
 current_nbformat = nbformat
 current_nbformat_minor = nbformat_minor
 current_nbformat_module = _v_latest.__name__
 
 
 def docstring_nbformat_mod(func):
     """Decorator for docstrings referring to classes/functions accessed through
     nbformat.current.
 
     Put {nbformat_mod} in the docstring in place of 'IPython.nbformat.v3'.
     """
     func.__doc__ = func.__doc__.format(nbformat_mod=current_nbformat_module)
     return func
 
 
 class NBFormatError(ValueError):
     pass
 
 
 def parse_py(s, **kwargs):
     """Parse a string into a (nbformat, string) tuple."""
     nbf = current_nbformat
     nbm = current_nbformat_minor
     
     pattern = r'# <nbformat>(?P<nbformat>\d+[\.\d+]*)</nbformat>'
     m = re.search(pattern,s)
     if m is not None:
         digits = m.group('nbformat').split('.')
         nbf = int(digits[0])
         if len(digits) > 1:
             nbm = int(digits[1])
 
     return nbf, nbm, s
 
 
 def reads_json(nbjson, **kwargs):
     """Read a JSON notebook from a string and return the NotebookNode
     object. Report if any JSON format errors are detected.
 
     """
     nb = reader_reads(nbjson, **kwargs)
     nb_current = convert(nb, current_nbformat)
     errors = validate(nb_current)
     if errors:
-        logger.error(
+        get_logger().error(
             "Notebook JSON is invalid (%d errors detected during read)",
             len(errors))
     return nb_current
 
 
 def writes_json(nb, **kwargs):
     """Take a NotebookNode object and write out a JSON string. Report if
     any JSON format errors are detected.
 
     """
     errors = validate(nb)
     if errors:
-        logger.error(
+        get_logger().error(
             "Notebook JSON is invalid (%d errors detected during write)",
             len(errors))
     nbjson = versions[current_nbformat].writes_json(nb, **kwargs)
     return nbjson
 
 
 def reads_py(s, **kwargs):
     """Read a .py notebook from a string and return the NotebookNode object."""
     nbf, nbm, s = parse_py(s, **kwargs)
     if nbf in (2, 3):
         nb = versions[nbf].to_notebook_py(s, **kwargs)
     else:
         raise NBFormatError('Unsupported PY nbformat version: %i' % nbf)
     return nb
 
 
 def writes_py(nb, **kwargs):
     # nbformat 3 is the latest format that supports py
     return versions[3].writes_py(nb, **kwargs)
 
 
 # High level API
 
 
 def reads(s, format, **kwargs):
     """Read a notebook from a string and return the NotebookNode object.
 
     This function properly handles notebooks of any version. The notebook
     returned will always be in the current version's format.
 
     Parameters
     ----------
     s : unicode
         The raw unicode string to read the notebook from.
     format : (u'json', u'ipynb', u'py')
         The format that the string is in.
 
     Returns
     -------
     nb : NotebookNode
         The notebook that was read.
     """
     format = unicode_type(format)
     if format == u'json' or format == u'ipynb':
         return reads_json(s, **kwargs)
     elif format == u'py':
         return reads_py(s, **kwargs)
     else:
         raise NBFormatError('Unsupported format: %s' % format)
 
 
 def writes(nb, format, **kwargs):
     """Write a notebook to a string in a given format in the current nbformat version.
 
     This function always writes the notebook in the current nbformat version.
 
     Parameters
     ----------
     nb : NotebookNode
         The notebook to write.
     format : (u'json', u'ipynb', u'py')
         The format to write the notebook in.
 
     Returns
     -------
     s : unicode
         The notebook string.
     """
     format = unicode_type(format)
     if format == u'json' or format == u'ipynb':
         return writes_json(nb, **kwargs)
     elif format == u'py':
         return writes_py(nb, **kwargs)
     else:
         raise NBFormatError('Unsupported format: %s' % format)
 
 
 def read(fp, format, **kwargs):
     """Read a notebook from a file and return the NotebookNode object.
 
     This function properly handles notebooks of any version. The notebook
     returned will always be in the current version's format.
 
     Parameters
     ----------
     fp : file
         Any file-like object with a read method.
     format : (u'json', u'ipynb', u'py')
         The format that the string is in.
 
     Returns
     -------
     nb : NotebookNode
         The notebook that was read.
     """
     return reads(fp.read(), format, **kwargs)
 
 
 def write(nb, fp, format, **kwargs):
     """Write a notebook to a file in a given format in the current nbformat version.
 
     This function always writes the notebook in the current nbformat version.
 
     Parameters
     ----------
     nb : NotebookNode
         The notebook to write.
     fp : file
         Any file-like object with a write method.
     format : (u'json', u'ipynb', u'py')
         The format to write the notebook in.
 
     Returns
     -------
     s : unicode
         The notebook string.
     """
     return fp.write(writes(nb, format, **kwargs))
 
 def _convert_to_metadata():
     """Convert to a notebook having notebook metadata."""
     import glob
     for fname in glob.glob('*.ipynb'):
         print('Converting file:',fname)
         with open(fname,'r') as f:
             nb = read(f,u'json')
         md = new_metadata()
         if u'name' in nb:
             md.name = nb.name
             del nb[u'name']            
         nb.metadata = md
         with open(fname,'w') as f:
             write(nb, f, u'json')
 

IPython/parallel/controller/scheduler.py

@@ -1,859 +1,848 b''
 """The Python scheduler for rich scheduling.
 
 The Pure ZMQ scheduler does not allow routing schemes other than LRU,
 nor does it check msg_id DAG dependencies. For those, a slightly slower
 Python Scheduler exists.
-
-Authors:
-
-* Min RK
 """
-#-----------------------------------------------------------------------------
-#  Copyright (C) 2010-2011  The IPython Development Team
-#
-#  Distributed under the terms of the BSD License.  The full license is in
-#  the file COPYING, distributed as part of this software.
-#-----------------------------------------------------------------------------
 
-#----------------------------------------------------------------------
-# Imports
-#----------------------------------------------------------------------
+# Copyright (c) IPython Development Team.
+# Distributed under the terms of the Modified BSD License.
 
 import logging
 import sys
 import time
 
 from collections import deque
 from datetime import datetime
 from random import randint, random
 from types import FunctionType
 
 try:
     import numpy
 except ImportError:
     numpy = None
 
 import zmq
 from zmq.eventloop import ioloop, zmqstream
 
 # local imports
 from IPython.external.decorator import decorator
 from IPython.config.application import Application
 from IPython.config.loader import Config
 from IPython.utils.traitlets import Instance, Dict, List, Set, Integer, Enum, CBytes
 from IPython.utils.py3compat import cast_bytes
 
 from IPython.parallel import error, util
 from IPython.parallel.factory import SessionFactory
 from IPython.parallel.util import connect_logger, local_logger
 
 from .dependency import Dependency
 
 @decorator
 def logged(f,self,*args,**kwargs):
     # print ("#--------------------")
     self.log.debug("scheduler::%s(*%s,**%s)", f.__name__, args, kwargs)
     # print ("#--")
     return f(self,*args, **kwargs)
 
 #----------------------------------------------------------------------
 # Chooser functions
 #----------------------------------------------------------------------
 
 def plainrandom(loads):
     """Plain random pick."""
     n = len(loads)
     return randint(0,n-1)
 
 def lru(loads):
     """Always pick the front of the line.
 
     The content of `loads` is ignored.
 
     Assumes LRU ordering of loads, with oldest first.
     """
     return 0
 
 def twobin(loads):
     """Pick two at random, use the LRU of the two.
 
     The content of loads is ignored.
 
     Assumes LRU ordering of loads, with oldest first.
     """
     n = len(loads)
     a = randint(0,n-1)
     b = randint(0,n-1)
     return min(a,b)
 
 def weighted(loads):
     """Pick two at random using inverse load as weight.
 
     Return the less loaded of the two.
     """
     # weight 0 a million times more than 1:
     weights = 1./(1e-6+numpy.array(loads))
     sums = weights.cumsum()
     t = sums[-1]
     x = random()*t
     y = random()*t
     idx = 0
     idy = 0
     while sums[idx] < x:
         idx += 1
     while sums[idy] < y:
         idy += 1
     if weights[idy] > weights[idx]:
         return idy
     else:
         return idx
 
 def leastload(loads):
     """Always choose the lowest load.
 
     If the lowest load occurs more than once, the first
     occurance will be used.  If loads has LRU ordering, this means
     the LRU of those with the lowest load is chosen.
     """
     return loads.index(min(loads))
 
 #---------------------------------------------------------------------
 # Classes
 #---------------------------------------------------------------------
 
 
 # store empty default dependency:
 MET = Dependency([])
 
 
 class Job(object):
     """Simple container for a job"""
     def __init__(self, msg_id, raw_msg, idents, msg, header, metadata,
                     targets, after, follow, timeout):
         self.msg_id = msg_id
         self.raw_msg = raw_msg
         self.idents = idents
         self.msg = msg
         self.header = header
         self.metadata = metadata
         self.targets = targets
         self.after = after
         self.follow = follow
         self.timeout = timeout
         
         self.removed = False # used for lazy-delete from sorted queue
         self.timestamp = time.time()
         self.timeout_id = 0
         self.blacklist = set()
 
     def __lt__(self, other):
         return self.timestamp < other.timestamp
 
     def __cmp__(self, other):
         return cmp(self.timestamp, other.timestamp)
 
     @property
     def dependents(self):
         return self.follow.union(self.after)
 
 
 class TaskScheduler(SessionFactory):
     """Python TaskScheduler object.
 
     This is the simplest object that supports msg_id based
     DAG dependencies. *Only* task msg_ids are checked, not
     msg_ids of jobs submitted via the MUX queue.
 
     """
 
     hwm = Integer(1, config=True,
         help="""specify the High Water Mark (HWM) for the downstream
         socket in the Task scheduler. This is the maximum number
         of allowed outstanding tasks on each engine.
         
         The default (1) means that only one task can be outstanding on each
         engine.  Setting TaskScheduler.hwm=0 means there is no limit, and the
         engines continue to be assigned tasks while they are working,
         effectively hiding network latency behind computation, but can result
         in an imbalance of work when submitting many heterogenous tasks all at
         once.  Any positive value greater than one is a compromise between the
         two.
 
         """
     )
     scheme_name = Enum(('leastload', 'pure', 'lru', 'plainrandom', 'weighted', 'twobin'),
         'leastload', config=True, allow_none=False,
         help="""select the task scheduler scheme  [default: Python LRU]
         Options are: 'pure', 'lru', 'plainrandom', 'weighted', 'twobin','leastload'"""
     )
     def _scheme_name_changed(self, old, new):
         self.log.debug("Using scheme %r"%new)
         self.scheme = globals()[new]
 
     # input arguments:
     scheme = Instance(FunctionType) # function for determining the destination
     def _scheme_default(self):
         return leastload
     client_stream = Instance(zmqstream.ZMQStream) # client-facing stream
     engine_stream = Instance(zmqstream.ZMQStream) # engine-facing stream
     notifier_stream = Instance(zmqstream.ZMQStream) # hub-facing sub stream
     mon_stream = Instance(zmqstream.ZMQStream) # hub-facing pub stream
     query_stream = Instance(zmqstream.ZMQStream) # hub-facing DEALER stream
 
     # internals:
     queue = Instance(deque) # sorted list of Jobs
     def _queue_default(self):
         return deque()
     queue_map = Dict() # dict by msg_id of Jobs (for O(1) access to the Queue)
     graph = Dict() # dict by msg_id of [ msg_ids that depend on key ]
     retries = Dict() # dict by msg_id of retries remaining (non-neg ints)
     # waiting = List() # list of msg_ids ready to run, but haven't due to HWM
     pending = Dict() # dict by engine_uuid of submitted tasks
     completed = Dict() # dict by engine_uuid of completed tasks
     failed = Dict() # dict by engine_uuid of failed tasks
     destinations = Dict() # dict by msg_id of engine_uuids where jobs ran (reverse of completed+failed)
     clients = Dict() # dict by msg_id for who submitted the task
     targets = List() # list of target IDENTs
     loads = List() # list of engine loads
     # full = Set() # set of IDENTs that have HWM outstanding tasks
     all_completed = Set() # set of all completed tasks
     all_failed = Set() # set of all failed tasks
     all_done = Set() # set of all finished tasks=union(completed,failed)
     all_ids = Set() # set of all submitted task IDs
 
     ident = CBytes() # ZMQ identity. This should just be self.session.session
                      # but ensure Bytes
     def _ident_default(self):
         return self.session.bsession
 
     def start(self):
         self.query_stream.on_recv(self.dispatch_query_reply)
         self.session.send(self.query_stream, "connection_request", {})
         
         self.engine_stream.on_recv(self.dispatch_result, copy=False)
         self.client_stream.on_recv(self.dispatch_submission, copy=False)
 
         self._notification_handlers = dict(
             registration_notification = self._register_engine,
             unregistration_notification = self._unregister_engine
         )
         self.notifier_stream.on_recv(self.dispatch_notification)
         self.log.info("Scheduler started [%s]" % self.scheme_name)
 
     def resume_receiving(self):
         """Resume accepting jobs."""
         self.client_stream.on_recv(self.dispatch_submission, copy=False)
 
     def stop_receiving(self):
         """Stop accepting jobs while there are no engines.
         Leave them in the ZMQ queue."""
         self.client_stream.on_recv(None)
 
     #-----------------------------------------------------------------------
     # [Un]Registration Handling
     #-----------------------------------------------------------------------
     
     
     def dispatch_query_reply(self, msg):
         """handle reply to our initial connection request"""
         try:
             idents,msg = self.session.feed_identities(msg)
         except ValueError:
             self.log.warn("task::Invalid Message: %r",msg)
             return
         try:
             msg = self.session.unserialize(msg)
         except ValueError:
             self.log.warn("task::Unauthorized message from: %r"%idents)
             return
         
         content = msg['content']
         for uuid in content.get('engines', {}).values():
             self._register_engine(cast_bytes(uuid))
 
     
     @util.log_errors
     def dispatch_notification(self, msg):
         """dispatch register/unregister events."""
         try:
             idents,msg = self.session.feed_identities(msg)
         except ValueError:
             self.log.warn("task::Invalid Message: %r",msg)
             return
         try:
             msg = self.session.unserialize(msg)
         except ValueError:
             self.log.warn("task::Unauthorized message from: %r"%idents)
             return
 
         msg_type = msg['header']['msg_type']
 
         handler = self._notification_handlers.get(msg_type, None)
         if handler is None:
             self.log.error("Unhandled message type: %r"%msg_type)
         else:
             try:
                 handler(cast_bytes(msg['content']['uuid']))
             except Exception:
                 self.log.error("task::Invalid notification msg: %r", msg, exc_info=True)
 
     def _register_engine(self, uid):
         """New engine with ident `uid` became available."""
         # head of the line:
         self.targets.insert(0,uid)
         self.loads.insert(0,0)
 
         # initialize sets
         self.completed[uid] = set()
         self.failed[uid] = set()
         self.pending[uid] = {}
 
         # rescan the graph:
         self.update_graph(None)
 
     def _unregister_engine(self, uid):
         """Existing engine with ident `uid` became unavailable."""
         if len(self.targets) == 1:
             # this was our only engine
             pass
 
         # handle any potentially finished tasks:
         self.engine_stream.flush()
 
         # don't pop destinations, because they might be used later
         # map(self.destinations.pop, self.completed.pop(uid))
         # map(self.destinations.pop, self.failed.pop(uid))
 
         # prevent this engine from receiving work
         idx = self.targets.index(uid)
         self.targets.pop(idx)
         self.loads.pop(idx)
 
         # wait 5 seconds before cleaning up pending jobs, since the results might
         # still be incoming
         if self.pending[uid]:
             dc = ioloop.DelayedCallback(lambda : self.handle_stranded_tasks(uid), 5000, self.loop)
             dc.start()
         else:
             self.completed.pop(uid)
             self.failed.pop(uid)
 
 
     def handle_stranded_tasks(self, engine):
         """Deal with jobs resident in an engine that died."""
         lost = self.pending[engine]
         for msg_id in lost.keys():
             if msg_id not in self.pending[engine]:
                 # prevent double-handling of messages
                 continue
 
             raw_msg = lost[msg_id].raw_msg
             idents,msg = self.session.feed_identities(raw_msg, copy=False)
             parent = self.session.unpack(msg[1].bytes)
             idents = [engine, idents[0]]
 
             # build fake error reply
             try:
                 raise error.EngineError("Engine %r died while running task %r"%(engine, msg_id))
             except:
                 content = error.wrap_exception()
             # build fake metadata
             md = dict(
                 status=u'error',
                 engine=engine.decode('ascii'),
                 date=datetime.now(),
             )
             msg = self.session.msg('apply_reply', content, parent=parent, metadata=md)
             raw_reply = list(map(zmq.Message, self.session.serialize(msg, ident=idents)))
             # and dispatch it
             self.dispatch_result(raw_reply)
 
         # finally scrub completed/failed lists
         self.completed.pop(engine)
         self.failed.pop(engine)
 
 
     #-----------------------------------------------------------------------
     # Job Submission
     #-----------------------------------------------------------------------
     
 
     @util.log_errors
     def dispatch_submission(self, raw_msg):
         """Dispatch job submission to appropriate handlers."""
         # ensure targets up to date:
         self.notifier_stream.flush()
         try:
             idents, msg = self.session.feed_identities(raw_msg, copy=False)
             msg = self.session.unserialize(msg, content=False, copy=False)
         except Exception:
             self.log.error("task::Invaid task msg: %r"%raw_msg, exc_info=True)
             return
 
 
         # send to monitor
         self.mon_stream.send_multipart([b'intask']+raw_msg, copy=False)
 
         header = msg['header']
         md = msg['metadata']
         msg_id = header['msg_id']
         self.all_ids.add(msg_id)
 
         # get targets as a set of bytes objects
         # from a list of unicode objects
         targets = md.get('targets', [])
         targets = set(map(cast_bytes, targets))
 
         retries = md.get('retries', 0)
         self.retries[msg_id] = retries
 
         # time dependencies
         after = md.get('after', None)
         if after:
             after = Dependency(after)
             if after.all:
                 if after.success:
                     after = Dependency(after.difference(self.all_completed),
                                 success=after.success,
                                 failure=after.failure,
                                 all=after.all,
                     )
                 if after.failure:
                     after = Dependency(after.difference(self.all_failed),
                                 success=after.success,
                                 failure=after.failure,
                                 all=after.all,
                     )
             if after.check(self.all_completed, self.all_failed):
                 # recast as empty set, if `after` already met,
                 # to prevent unnecessary set comparisons
                 after = MET
         else:
             after = MET
 
         # location dependencies
         follow = Dependency(md.get('follow', []))
 
         timeout = md.get('timeout', None)
         if timeout:
             timeout = float(timeout)
 
         job = Job(msg_id=msg_id, raw_msg=raw_msg, idents=idents, msg=msg,
                  header=header, targets=targets, after=after, follow=follow,
                  timeout=timeout, metadata=md,
         )
         # validate and reduce dependencies:
         for dep in after,follow:
             if not dep: # empty dependency
                 continue
             # check valid:
             if msg_id in dep or dep.difference(self.all_ids):
                 self.queue_map[msg_id] = job
                 return self.fail_unreachable(msg_id, error.InvalidDependency)
             # check if unreachable:
             if dep.unreachable(self.all_completed, self.all_failed):
                 self.queue_map[msg_id] = job
                 return self.fail_unreachable(msg_id)
 
         if after.check(self.all_completed, self.all_failed):
             # time deps already met, try to run
             if not self.maybe_run(job):
                 # can't run yet
                 if msg_id not in self.all_failed:
                     # could have failed as unreachable
                     self.save_unmet(job)
         else:
             self.save_unmet(job)
 
     def job_timeout(self, job, timeout_id):
         """callback for a job's timeout.
         
         The job may or may not have been run at this point.
         """
         if job.timeout_id != timeout_id:
             # not the most recent call
             return
         now = time.time()
         if job.timeout >= (now + 1):
             self.log.warn("task %s timeout fired prematurely: %s > %s",
                 job.msg_id, job.timeout, now
             )
         if job.msg_id in self.queue_map:
             # still waiting, but ran out of time
             self.log.info("task %r timed out", job.msg_id)
             self.fail_unreachable(job.msg_id, error.TaskTimeout)
 
     def fail_unreachable(self, msg_id, why=error.ImpossibleDependency):
         """a task has become unreachable, send a reply with an ImpossibleDependency
         error."""
         if msg_id not in self.queue_map:
             self.log.error("task %r already failed!", msg_id)
             return
         job = self.queue_map.pop(msg_id)
         # lazy-delete from the queue
         job.removed = True
         for mid in job.dependents:
             if mid in self.graph:
                 self.graph[mid].remove(msg_id)
 
         try:
             raise why()
         except:
             content = error.wrap_exception()
         self.log.debug("task %r failing as unreachable with: %s", msg_id, content['ename'])
 
         self.all_done.add(msg_id)
         self.all_failed.add(msg_id)
 
         msg = self.session.send(self.client_stream, 'apply_reply', content,
                                                 parent=job.header, ident=job.idents)
         self.session.send(self.mon_stream, msg, ident=[b'outtask']+job.idents)
 
         self.update_graph(msg_id, success=False)
 
     def available_engines(self):
         """return a list of available engine indices based on HWM"""
         if not self.hwm:
             return list(range(len(self.targets)))
         available = []
         for idx in range(len(self.targets)):
             if self.loads[idx] < self.hwm:
                 available.append(idx)
         return available
 
     def maybe_run(self, job):
         """check location dependencies, and run if they are met."""
         msg_id = job.msg_id
         self.log.debug("Attempting to assign task %s", msg_id)
         available = self.available_engines()
         if not available:
             # no engines, definitely can't run
             return False
         
         if job.follow or job.targets or job.blacklist or self.hwm:
             # we need a can_run filter
             def can_run(idx):
                 # check hwm
                 if self.hwm and self.loads[idx] == self.hwm:
                     return False
                 target = self.targets[idx]
                 # check blacklist
                 if target in job.blacklist:
                     return False
                 # check targets
                 if job.targets and target not in job.targets:
                     return False
                 # check follow
                 return job.follow.check(self.completed[target], self.failed[target])
 
             indices = list(filter(can_run, available))
 
             if not indices:
                 # couldn't run
                 if job.follow.all:
                     # check follow for impossibility
                     dests = set()
                     relevant = set()
                     if job.follow.success:
                         relevant = self.all_completed
                     if job.follow.failure:
                         relevant = relevant.union(self.all_failed)
                     for m in job.follow.intersection(relevant):
                         dests.add(self.destinations[m])
                     if len(dests) > 1:
                         self.queue_map[msg_id] = job
                         self.fail_unreachable(msg_id)
                         return False
                 if job.targets:
                     # check blacklist+targets for impossibility
                     job.targets.difference_update(job.blacklist)
                     if not job.targets or not job.targets.intersection(self.targets):
                         self.queue_map[msg_id] = job
                         self.fail_unreachable(msg_id)
                         return False
                 return False
         else:
             indices = None
 
         self.submit_task(job, indices)
         return True
 
     def save_unmet(self, job):
         """Save a message for later submission when its dependencies are met."""
         msg_id = job.msg_id
         self.log.debug("Adding task %s to the queue", msg_id)
         self.queue_map[msg_id] = job
         self.queue.append(job)
         # track the ids in follow or after, but not those already finished
         for dep_id in job.after.union(job.follow).difference(self.all_done):
             if dep_id not in self.graph:
                 self.graph[dep_id] = set()
             self.graph[dep_id].add(msg_id)
         
         # schedule timeout callback
         if job.timeout:
             timeout_id = job.timeout_id = job.timeout_id + 1
             self.loop.add_timeout(time.time() + job.timeout,
                 lambda : self.job_timeout(job, timeout_id)
             )
         
 
     def submit_task(self, job, indices=None):
         """Submit a task to any of a subset of our targets."""
         if indices:
             loads = [self.loads[i] for i in indices]
         else:
             loads = self.loads
         idx = self.scheme(loads)
         if indices:
             idx = indices[idx]
         target = self.targets[idx]
         # print (target, map(str, msg[:3]))
         # send job to the engine
         self.engine_stream.send(target, flags=zmq.SNDMORE, copy=False)
         self.engine_stream.send_multipart(job.raw_msg, copy=False)
         # update load
         self.add_job(idx)
         self.pending[target][job.msg_id] = job
         # notify Hub
         content = dict(msg_id=job.msg_id, engine_id=target.decode('ascii'))
         self.session.send(self.mon_stream, 'task_destination', content=content,
                         ident=[b'tracktask',self.ident])
 
 
     #-----------------------------------------------------------------------
     # Result Handling
     #-----------------------------------------------------------------------
     
     
     @util.log_errors
     def dispatch_result(self, raw_msg):
         """dispatch method for result replies"""
         try:
             idents,msg = self.session.feed_identities(raw_msg, copy=False)
             msg = self.session.unserialize(msg, content=False, copy=False)
             engine = idents[0]
             try:
                 idx = self.targets.index(engine)
             except ValueError:
                 pass # skip load-update for dead engines
             else:
                 self.finish_job(idx)
         except Exception:
             self.log.error("task::Invalid result: %r", raw_msg, exc_info=True)
             return
 
         md = msg['metadata']
         parent = msg['parent_header']
         if md.get('dependencies_met', True):
             success = (md['status'] == 'ok')
             msg_id = parent['msg_id']
             retries = self.retries[msg_id]
             if not success and retries > 0:
                 # failed
                 self.retries[msg_id] = retries - 1
                 self.handle_unmet_dependency(idents, parent)
             else:
                 del self.retries[msg_id]
                 # relay to client and update graph
                 self.handle_result(idents, parent, raw_msg, success)
                 # send to Hub monitor
                 self.mon_stream.send_multipart([b'outtask']+raw_msg, copy=False)
         else:
             self.handle_unmet_dependency(idents, parent)
 
     def handle_result(self, idents, parent, raw_msg, success=True):
         """handle a real task result, either success or failure"""
         # first, relay result to client
         engine = idents[0]
         client = idents[1]
         # swap_ids for ROUTER-ROUTER mirror
         raw_msg[:2] = [client,engine]
         # print (map(str, raw_msg[:4]))
         self.client_stream.send_multipart(raw_msg, copy=False)
         # now, update our data structures
         msg_id = parent['msg_id']
         self.pending[engine].pop(msg_id)
         if success:
             self.completed[engine].add(msg_id)
             self.all_completed.add(msg_id)
         else:
             self.failed[engine].add(msg_id)
             self.all_failed.add(msg_id)
         self.all_done.add(msg_id)
         self.destinations[msg_id] = engine
 
         self.update_graph(msg_id, success)
 
     def handle_unmet_dependency(self, idents, parent):
         """handle an unmet dependency"""
         engine = idents[0]
         msg_id = parent['msg_id']
 
         job = self.pending[engine].pop(msg_id)
         job.blacklist.add(engine)
 
         if job.blacklist == job.targets:
             self.queue_map[msg_id] = job
             self.fail_unreachable(msg_id)
         elif not self.maybe_run(job):
             # resubmit failed
             if msg_id not in self.all_failed:
                 # put it back in our dependency tree
                 self.save_unmet(job)
 
         if self.hwm:
             try:
                 idx = self.targets.index(engine)
             except ValueError:
                 pass # skip load-update for dead engines
             else:
                 if self.loads[idx] == self.hwm-1:
                     self.update_graph(None)
 
     def update_graph(self, dep_id=None, success=True):
         """dep_id just finished. Update our dependency
         graph and submit any jobs that just became runnable.
 
         Called with dep_id=None to update entire graph for hwm, but without finishing a task.
         """
         # print ("\n\n***********")
         # pprint (dep_id)
         # pprint (self.graph)
         # pprint (self.queue_map)
         # pprint (self.all_completed)
         # pprint (self.all_failed)
         # print ("\n\n***********\n\n")
         # update any jobs that depended on the dependency
         msg_ids = self.graph.pop(dep_id, [])
 
         # recheck *all* jobs if
         # a) we have HWM and an engine just become no longer full
         # or b) dep_id was given as None
         
         if dep_id is None or self.hwm and any( [ load==self.hwm-1 for load in self.loads ]):
             jobs = self.queue
             using_queue = True
         else:
             using_queue = False
             jobs = deque(sorted( self.queue_map[msg_id] for msg_id in msg_ids ))
         
         to_restore = []
         while jobs:
             job = jobs.popleft()
             if job.removed:
                 continue
             msg_id = job.msg_id
             
             put_it_back = True
             
             if job.after.unreachable(self.all_completed, self.all_failed)\
                     or job.follow.unreachable(self.all_completed, self.all_failed):
                 self.fail_unreachable(msg_id)
                 put_it_back = False
 
             elif job.after.check(self.all_completed, self.all_failed): # time deps met, maybe run
                 if self.maybe_run(job):
                     put_it_back = False
                     self.queue_map.pop(msg_id)
                     for mid in job.dependents:
                         if mid in self.graph:
                             self.graph[mid].remove(msg_id)
                     
                     # abort the loop if we just filled up all of our engines.
                     # avoids an O(N) operation in situation of full queue,
                     # where graph update is triggered as soon as an engine becomes
                     # non-full, and all tasks after the first are checked,
                     # even though they can't run.
                     if not self.available_engines():
                         break
             
             if using_queue and put_it_back:
                 # popped a job from the queue but it neither ran nor failed,
                 # so we need to put it back when we are done
                 # make sure to_restore preserves the same ordering
                 to_restore.append(job)
         
         # put back any tasks we popped but didn't run
         if using_queue:
             self.queue.extendleft(to_restore)
     
     #----------------------------------------------------------------------
     # methods to be overridden by subclasses
     #----------------------------------------------------------------------
 
     def add_job(self, idx):
         """Called after self.targets[idx] just got the job with header.
         Override with subclasses.  The default ordering is simple LRU.
         The default loads are the number of outstanding jobs."""
         self.loads[idx] += 1
         for lis in (self.targets, self.loads):
             lis.append(lis.pop(idx))
 
 
     def finish_job(self, idx):
         """Called after self.targets[idx] just finished a job.
         Override with subclasses."""
         self.loads[idx] -= 1
 
 
 
 def launch_scheduler(in_addr, out_addr, mon_addr, not_addr, reg_addr, config=None,
                         logname='root', log_url=None, loglevel=logging.DEBUG,
                         identity=b'task', in_thread=False):
 
     ZMQStream = zmqstream.ZMQStream
 
     if config:
         # unwrap dict back into Config
         config = Config(config)
 
     if in_thread:
         # use instance() to get the same Context/Loop as our parent
         ctx = zmq.Context.instance()
         loop = ioloop.IOLoop.instance()
     else:
         # in a process, don't use instance()
         # for safety with multiprocessing
         ctx = zmq.Context()
         loop = ioloop.IOLoop()
     ins = ZMQStream(ctx.socket(zmq.ROUTER),loop)
     util.set_hwm(ins, 0)
     ins.setsockopt(zmq.IDENTITY, identity + b'_in')
     ins.bind(in_addr)
 
     outs = ZMQStream(ctx.socket(zmq.ROUTER),loop)
     util.set_hwm(outs, 0)
     outs.setsockopt(zmq.IDENTITY, identity + b'_out')
     outs.bind(out_addr)
     mons = zmqstream.ZMQStream(ctx.socket(zmq.PUB),loop)
     util.set_hwm(mons, 0)
     mons.connect(mon_addr)
     nots = zmqstream.ZMQStream(ctx.socket(zmq.SUB),loop)
     nots.setsockopt(zmq.SUBSCRIBE, b'')
     nots.connect(not_addr)
     
     querys = ZMQStream(ctx.socket(zmq.DEALER),loop)
     querys.connect(reg_addr)
     
     # setup logging.
     if in_thread:
         log = Application.instance().log
     else:
         if log_url:
             log = connect_logger(logname, ctx, log_url, root="scheduler", loglevel=loglevel)
         else:
             log = local_logger(logname, loglevel)
 
     scheduler = TaskScheduler(client_stream=ins, engine_stream=outs,
                             mon_stream=mons, notifier_stream=nots,
                             query_stream=querys,
                             loop=loop, log=log,
                             config=config)
     scheduler.start()
     if not in_thread:
         try:
             loop.start()
         except KeyboardInterrupt:
             scheduler.log.critical("Interrupted, exiting...")
 

IPython/parallel/util.py

@@ -1,388 +1,389 b''
 """Some generic utilities for dealing with classes, urls, and serialization."""
 
 # Copyright (c) IPython Development Team.
 # Distributed under the terms of the Modified BSD License.
 
 import logging
 import os
 import re
 import stat
 import socket
 import sys
 import warnings
 from signal import signal, SIGINT, SIGABRT, SIGTERM
 try:
     from signal import SIGKILL
 except ImportError:
     SIGKILL=None
 from types import FunctionType
 
 try:
     import cPickle
     pickle = cPickle
 except:
     cPickle = None
     import pickle
 
 import zmq
 from zmq.log import handlers
 
+from IPython.utils.log import get_logger
 from IPython.external.decorator import decorator
 
 from IPython.config.application import Application
 from IPython.utils.localinterfaces import localhost, is_public_ip, public_ips
 from IPython.utils.py3compat import string_types, iteritems, itervalues
 from IPython.kernel.zmq.log import EnginePUBHandler
 
 
 #-----------------------------------------------------------------------------
 # Classes
 #-----------------------------------------------------------------------------
 
 class Namespace(dict):
     """Subclass of dict for attribute access to keys."""
     
     def __getattr__(self, key):
         """getattr aliased to getitem"""
         if key in self:
             return self[key]
         else:
             raise NameError(key)
 
     def __setattr__(self, key, value):
         """setattr aliased to setitem, with strict"""
         if hasattr(dict, key):
             raise KeyError("Cannot override dict keys %r"%key)
         self[key] = value
     
 
 class ReverseDict(dict):
     """simple double-keyed subset of dict methods."""
     
     def __init__(self, *args, **kwargs):
         dict.__init__(self, *args, **kwargs)
         self._reverse = dict()
         for key, value in iteritems(self):
             self._reverse[value] = key
     
     def __getitem__(self, key):
         try:
             return dict.__getitem__(self, key)
         except KeyError:
             return self._reverse[key]
     
     def __setitem__(self, key, value):
         if key in self._reverse:
             raise KeyError("Can't have key %r on both sides!"%key)
         dict.__setitem__(self, key, value)
         self._reverse[value] = key
     
     def pop(self, key):
         value = dict.pop(self, key)
         self._reverse.pop(value)
         return value
     
     def get(self, key, default=None):
         try:
             return self[key]
         except KeyError:
             return default
 
 #-----------------------------------------------------------------------------
 # Functions
 #-----------------------------------------------------------------------------
 
 @decorator
 def log_errors(f, self, *args, **kwargs):
     """decorator to log unhandled exceptions raised in a method.
     
     For use wrapping on_recv callbacks, so that exceptions
     do not cause the stream to be closed.
     """
     try:
         return f(self, *args, **kwargs)
     except Exception:
         self.log.error("Uncaught exception in %r" % f, exc_info=True)
     
 
 def is_url(url):
     """boolean check for whether a string is a zmq url"""
     if '://' not in url:
         return False
     proto, addr = url.split('://', 1)
     if proto.lower() not in ['tcp','pgm','epgm','ipc','inproc']:
         return False
     return True
 
 def validate_url(url):
     """validate a url for zeromq"""
     if not isinstance(url, string_types):
         raise TypeError("url must be a string, not %r"%type(url))
     url = url.lower()
     
     proto_addr = url.split('://')
     assert len(proto_addr) == 2, 'Invalid url: %r'%url
     proto, addr = proto_addr
     assert proto in ['tcp','pgm','epgm','ipc','inproc'], "Invalid protocol: %r"%proto
     
     # domain pattern adapted from http://www.regexlib.com/REDetails.aspx?regexp_id=391
     # author: Remi Sabourin
     pat = re.compile(r'^([\w\d]([\w\d\-]{0,61}[\w\d])?\.)*[\w\d]([\w\d\-]{0,61}[\w\d])?$')
     
     if proto == 'tcp':
         lis = addr.split(':')
         assert len(lis) == 2, 'Invalid url: %r'%url
         addr,s_port = lis
         try:
             port = int(s_port)
         except ValueError:
             raise AssertionError("Invalid port %r in url: %r"%(port, url))
         
         assert addr == '*' or pat.match(addr) is not None, 'Invalid url: %r'%url
         
     else:
         # only validate tcp urls currently
         pass
     
     return True
 
 
 def validate_url_container(container):
     """validate a potentially nested collection of urls."""
     if isinstance(container, string_types):
         url = container
         return validate_url(url)
     elif isinstance(container, dict):
         container = itervalues(container)
     
     for element in container:
         validate_url_container(element)
 
 
 def split_url(url):
     """split a zmq url (tcp://ip:port) into ('tcp','ip','port')."""
     proto_addr = url.split('://')
     assert len(proto_addr) == 2, 'Invalid url: %r'%url
     proto, addr = proto_addr
     lis = addr.split(':')
     assert len(lis) == 2, 'Invalid url: %r'%url
     addr,s_port = lis
     return proto,addr,s_port
 
 
 def disambiguate_ip_address(ip, location=None):
     """turn multi-ip interfaces '0.0.0.0' and '*' into a connectable address
     
     Explicit IP addresses are returned unmodified.
     
     Parameters
     ----------
     
     ip : IP address
         An IP address, or the special values 0.0.0.0, or *
     location: IP address, optional
         A public IP of the target machine.
         If location is an IP of the current machine,
         localhost will be returned,
         otherwise location will be returned.
     """
     if ip in {'0.0.0.0', '*'}:
         if not location:
             # unspecified location, localhost is the only choice
             ip = localhost()
         elif is_public_ip(location):
             # location is a public IP on this machine, use localhost
             ip = localhost()
         elif not public_ips():
             # this machine's public IPs cannot be determined,
             # assume `location` is not this machine
             warnings.warn("IPython could not determine public IPs", RuntimeWarning)
             ip = location
         else:
             # location is not this machine, do not use loopback
             ip = location
     return ip
 
 
 def disambiguate_url(url, location=None):
     """turn multi-ip interfaces '0.0.0.0' and '*' into connectable
     ones, based on the location (default interpretation is localhost).
     
     This is for zeromq urls, such as ``tcp://*:10101``.
     """
     try:
         proto,ip,port = split_url(url)
     except AssertionError:
         # probably not tcp url; could be ipc, etc.
         return url
     
     ip = disambiguate_ip_address(ip,location)
     
     return "%s://%s:%s"%(proto,ip,port)
 
 
 #--------------------------------------------------------------------------
 # helpers for implementing old MEC API via view.apply
 #--------------------------------------------------------------------------
 
 def interactive(f):
     """decorator for making functions appear as interactively defined.
     This results in the function being linked to the user_ns as globals()
     instead of the module globals().
     """
     
     # build new FunctionType, so it can have the right globals
     # interactive functions never have closures, that's kind of the point
     if isinstance(f, FunctionType):
         mainmod = __import__('__main__')
         f = FunctionType(f.__code__, mainmod.__dict__,
             f.__name__, f.__defaults__,
         )
     # associate with __main__ for uncanning
     f.__module__ = '__main__'
     return f
 
 @interactive
 def _push(**ns):
     """helper method for implementing `client.push` via `client.apply`"""
     user_ns = globals()
     tmp = '_IP_PUSH_TMP_'
     while tmp in user_ns:
         tmp = tmp + '_'
     try:
         for name, value in ns.items():
             user_ns[tmp] = value
             exec("%s = %s" % (name, tmp), user_ns)
     finally:
         user_ns.pop(tmp, None)
 
 @interactive
 def _pull(keys):
     """helper method for implementing `client.pull` via `client.apply`"""
     if isinstance(keys, (list,tuple, set)):
         return [eval(key, globals()) for key in keys]
     else:
         return eval(keys, globals())
 
 @interactive
 def _execute(code):
     """helper method for implementing `client.execute` via `client.apply`"""
     exec(code, globals())
 
 #--------------------------------------------------------------------------
 # extra process management utilities
 #--------------------------------------------------------------------------
 
 _random_ports = set()
 
 def select_random_ports(n):
     """Selects and return n random ports that are available."""
     ports = []
     for i in range(n):
         sock = socket.socket()
         sock.bind(('', 0))
         while sock.getsockname()[1] in _random_ports:
             sock.close()
             sock = socket.socket()
             sock.bind(('', 0))
         ports.append(sock)
     for i, sock in enumerate(ports):
         port = sock.getsockname()[1]
         sock.close()
         ports[i] = port
         _random_ports.add(port)
     return ports
 
 def signal_children(children):
     """Relay interupt/term signals to children, for more solid process cleanup."""
     def terminate_children(sig, frame):
-        log = Application.instance().log
+        log = get_logger()
         log.critical("Got signal %i, terminating children..."%sig)
         for child in children:
             child.terminate()
         
         sys.exit(sig != SIGINT)
         # sys.exit(sig)
     for sig in (SIGINT, SIGABRT, SIGTERM):
         signal(sig, terminate_children)
 
 def generate_exec_key(keyfile):
     import uuid
     newkey = str(uuid.uuid4())
     with open(keyfile, 'w') as f:
         # f.write('ipython-key ')
         f.write(newkey+'\n')
     # set user-only RW permissions (0600)
     # this will have no effect on Windows
     os.chmod(keyfile, stat.S_IRUSR|stat.S_IWUSR)
 
 
 def integer_loglevel(loglevel):
     try:
         loglevel = int(loglevel)
     except ValueError:
         if isinstance(loglevel, str):
             loglevel = getattr(logging, loglevel)
     return loglevel
 
 def connect_logger(logname, context, iface, root="ip", loglevel=logging.DEBUG):
     logger = logging.getLogger(logname)
     if any([isinstance(h, handlers.PUBHandler) for h in logger.handlers]):
         # don't add a second PUBHandler
         return
     loglevel = integer_loglevel(loglevel)
     lsock = context.socket(zmq.PUB)
     lsock.connect(iface)
     handler = handlers.PUBHandler(lsock)
     handler.setLevel(loglevel)
     handler.root_topic = root
     logger.addHandler(handler)
     logger.setLevel(loglevel)
 
 def connect_engine_logger(context, iface, engine, loglevel=logging.DEBUG):
     logger = logging.getLogger()
     if any([isinstance(h, handlers.PUBHandler) for h in logger.handlers]):
         # don't add a second PUBHandler
         return
     loglevel = integer_loglevel(loglevel)
     lsock = context.socket(zmq.PUB)
     lsock.connect(iface)
     handler = EnginePUBHandler(engine, lsock)
     handler.setLevel(loglevel)
     logger.addHandler(handler)
     logger.setLevel(loglevel)
     return logger
 
 def local_logger(logname, loglevel=logging.DEBUG):
     loglevel = integer_loglevel(loglevel)
     logger = logging.getLogger(logname)
     if any([isinstance(h, logging.StreamHandler) for h in logger.handlers]):
         # don't add a second StreamHandler
         return
     handler = logging.StreamHandler()
     handler.setLevel(loglevel)
     formatter = logging.Formatter("%(asctime)s.%(msecs).03d [%(name)s] %(message)s",
                 datefmt="%Y-%m-%d %H:%M:%S")
     handler.setFormatter(formatter)
 
     logger.addHandler(handler)
     logger.setLevel(loglevel)
     return logger
 
 def set_hwm(sock, hwm=0):
     """set zmq High Water Mark on a socket
     
     in a way that always works for various pyzmq / libzmq versions.
     """
     import zmq
     
     for key in ('HWM', 'SNDHWM', 'RCVHWM'):
         opt = getattr(zmq, key, None)
         if opt is None:
             continue
         try:
             sock.setsockopt(opt, hwm)
         except zmq.ZMQError:
             pass
 
         

IPython/utils/pickleutil.py

@@ -1,433 +1,420 b''
 # encoding: utf-8
 """Pickle related utilities. Perhaps this should be called 'can'."""
 
 # Copyright (c) IPython Development Team.
 # Distributed under the terms of the Modified BSD License.
 
 import copy
 import logging
 import sys
 from types import FunctionType
 
 try:
     import cPickle as pickle
 except ImportError:
     import pickle
 
 from . import codeutil  # This registers a hook when it's imported
 from . import py3compat
 from .importstring import import_item
 from .py3compat import string_types, iteritems
 
 from IPython.config import Application
+from IPython.utils.log import get_logger
 
 if py3compat.PY3:
     buffer = memoryview
     class_type = type
 else:
     from types import ClassType
     class_type = (type, ClassType)
 
 def _get_cell_type(a=None):
     """the type of a closure cell doesn't seem to be importable,
     so just create one
     """
     def inner():
         return a
     return type(py3compat.get_closure(inner)[0])
 
 cell_type = _get_cell_type()
 
 #-------------------------------------------------------------------------------
 # Functions
 #-------------------------------------------------------------------------------
 
 
 def use_dill():
     """use dill to expand serialization support
     
     adds support for object methods and closures to serialization.
     """
     # import dill causes most of the magic
     import dill
     
     # dill doesn't work with cPickle,
     # tell the two relevant modules to use plain pickle
     
     global pickle
     pickle = dill
 
     try:
         from IPython.kernel.zmq import serialize
     except ImportError:
         pass
     else:
         serialize.pickle = dill
     
     # disable special function handling, let dill take care of it
     can_map.pop(FunctionType, None)
 
 def use_cloudpickle():
     """use cloudpickle to expand serialization support
     
     adds support for object methods and closures to serialization.
     """
     from cloud.serialization import cloudpickle
     
     global pickle
     pickle = cloudpickle
 
     try:
         from IPython.kernel.zmq import serialize
     except ImportError:
         pass
     else:
         serialize.pickle = cloudpickle
     
     # disable special function handling, let cloudpickle take care of it
     can_map.pop(FunctionType, None)
 
 
 #-------------------------------------------------------------------------------
 # Classes
 #-------------------------------------------------------------------------------
 
 
 class CannedObject(object):
     def __init__(self, obj, keys=[], hook=None):
         """can an object for safe pickling
         
         Parameters
         ==========
         
         obj:
             The object to be canned
         keys: list (optional)
             list of attribute names that will be explicitly canned / uncanned
         hook: callable (optional)
             An optional extra callable,
             which can do additional processing of the uncanned object.
         
         large data may be offloaded into the buffers list,
         used for zero-copy transfers.
         """
         self.keys = keys
         self.obj = copy.copy(obj)
         self.hook = can(hook)
         for key in keys:
             setattr(self.obj, key, can(getattr(obj, key)))
         
         self.buffers = []
 
     def get_object(self, g=None):
         if g is None:
             g = {}
         obj = self.obj
         for key in self.keys:
             setattr(obj, key, uncan(getattr(obj, key), g))
         
         if self.hook:
             self.hook = uncan(self.hook, g)
             self.hook(obj, g)
         return self.obj
     
 
 class Reference(CannedObject):
     """object for wrapping a remote reference by name."""
     def __init__(self, name):
         if not isinstance(name, string_types):
             raise TypeError("illegal name: %r"%name)
         self.name = name
         self.buffers = []
 
     def __repr__(self):
         return "<Reference: %r>"%self.name
 
     def get_object(self, g=None):
         if g is None:
             g = {}
         
         return eval(self.name, g)
 
 
 class CannedCell(CannedObject):
     """Can a closure cell"""
     def __init__(self, cell):
         self.cell_contents = can(cell.cell_contents)
     
     def get_object(self, g=None):
         cell_contents = uncan(self.cell_contents, g)
         def inner():
             return cell_contents
         return py3compat.get_closure(inner)[0]
 
 
 class CannedFunction(CannedObject):
 
     def __init__(self, f):
         self._check_type(f)
         self.code = f.__code__
         if f.__defaults__:
             self.defaults = [ can(fd) for fd in f.__defaults__ ]
         else:
             self.defaults = None
         
         closure = py3compat.get_closure(f)
         if closure:
             self.closure = tuple( can(cell) for cell in closure )
         else:
             self.closure = None
         
         self.module = f.__module__ or '__main__'
         self.__name__ = f.__name__
         self.buffers = []
 
     def _check_type(self, obj):
         assert isinstance(obj, FunctionType), "Not a function type"
 
     def get_object(self, g=None):
         # try to load function back into its module:
         if not self.module.startswith('__'):
             __import__(self.module)
             g = sys.modules[self.module].__dict__
 
         if g is None:
             g = {}
         if self.defaults:
             defaults = tuple(uncan(cfd, g) for cfd in self.defaults)
         else:
             defaults = None
         if self.closure:
             closure = tuple(uncan(cell, g) for cell in self.closure)
         else:
             closure = None
         newFunc = FunctionType(self.code, g, self.__name__, defaults, closure)
         return newFunc
 
 class CannedClass(CannedObject):
 
     def __init__(self, cls):
         self._check_type(cls)
         self.name = cls.__name__
         self.old_style = not isinstance(cls, type)
         self._canned_dict = {}
         for k,v in cls.__dict__.items():
             if k not in ('__weakref__', '__dict__'):
                 self._canned_dict[k] = can(v)
         if self.old_style:
             mro = []
         else:
             mro = cls.mro()
         
         self.parents = [ can(c) for c in mro[1:] ]
         self.buffers = []
 
     def _check_type(self, obj):
         assert isinstance(obj, class_type), "Not a class type"
 
     def get_object(self, g=None):
         parents = tuple(uncan(p, g) for p in self.parents)
         return type(self.name, parents, uncan_dict(self._canned_dict, g=g))
 
 class CannedArray(CannedObject):
     def __init__(self, obj):
         from numpy import ascontiguousarray
         self.shape = obj.shape
         self.dtype = obj.dtype.descr if obj.dtype.fields else obj.dtype.str
         self.pickled = False
         if sum(obj.shape) == 0:
             self.pickled = True
         elif obj.dtype == 'O':
             # can't handle object dtype with buffer approach
             self.pickled = True
         elif obj.dtype.fields and any(dt == 'O' for dt,sz in obj.dtype.fields.values()):
             self.pickled = True
         if self.pickled:
             # just pickle it
             self.buffers = [pickle.dumps(obj, -1)]
         else:
             # ensure contiguous
             obj = ascontiguousarray(obj, dtype=None)
             self.buffers = [buffer(obj)]
     
     def get_object(self, g=None):
         from numpy import frombuffer
         data = self.buffers[0]
         if self.pickled:
             # no shape, we just pickled it
             return pickle.loads(data)
         else:
             return frombuffer(data, dtype=self.dtype).reshape(self.shape)
 
 
 class CannedBytes(CannedObject):
     wrap = bytes
     def __init__(self, obj):
         self.buffers = [obj]
     
     def get_object(self, g=None):
         data = self.buffers[0]
         return self.wrap(data)
 
 def CannedBuffer(CannedBytes):
     wrap = buffer
 
 #-------------------------------------------------------------------------------
 # Functions
 #-------------------------------------------------------------------------------
 
-def _logger():
-    """get the logger for the current Application
-    
-    the root logger will be used if no Application is running
-    """
-    if Application.initialized():
-        logger = Application.instance().log
-    else:
-        logger = logging.getLogger()
-        if not logger.handlers:
-            logging.basicConfig()
-    
-    return logger
-
 def _import_mapping(mapping, original=None):
     """import any string-keys in a type mapping
     
     """
-    log = _logger()
+    log = get_logger()
     log.debug("Importing canning map")
     for key,value in list(mapping.items()):
         if isinstance(key, string_types):
             try:
                 cls = import_item(key)
             except Exception:
                 if original and key not in original:
                     # only message on user-added classes
                     log.error("canning class not importable: %r", key, exc_info=True)
                 mapping.pop(key)
             else:
                 mapping[cls] = mapping.pop(key)
 
 def istype(obj, check):
     """like isinstance(obj, check), but strict
     
     This won't catch subclasses.
     """
     if isinstance(check, tuple):
         for cls in check:
             if type(obj) is cls:
                 return True
         return False
     else:
         return type(obj) is check
 
 def can(obj):
     """prepare an object for pickling"""
     
     import_needed = False
     
     for cls,canner in iteritems(can_map):
         if isinstance(cls, string_types):
             import_needed = True
             break
         elif istype(obj, cls):
             return canner(obj)
     
     if import_needed:
         # perform can_map imports, then try again
         # this will usually only happen once
         _import_mapping(can_map, _original_can_map)
         return can(obj)
     
     return obj
 
 def can_class(obj):
     if isinstance(obj, class_type) and obj.__module__ == '__main__':
         return CannedClass(obj)
     else:
         return obj
 
 def can_dict(obj):
     """can the *values* of a dict"""
     if istype(obj, dict):
         newobj = {}
         for k, v in iteritems(obj):
             newobj[k] = can(v)
         return newobj
     else:
         return obj
 
 sequence_types = (list, tuple, set)
 
 def can_sequence(obj):
     """can the elements of a sequence"""
     if istype(obj, sequence_types):
         t = type(obj)
         return t([can(i) for i in obj])
     else:
         return obj
 
 def uncan(obj, g=None):
     """invert canning"""
     
     import_needed = False
     for cls,uncanner in iteritems(uncan_map):
         if isinstance(cls, string_types):
             import_needed = True
             break
         elif isinstance(obj, cls):
             return uncanner(obj, g)
     
     if import_needed:
         # perform uncan_map imports, then try again
         # this will usually only happen once
         _import_mapping(uncan_map, _original_uncan_map)
         return uncan(obj, g)
     
     return obj
 
 def uncan_dict(obj, g=None):
     if istype(obj, dict):
         newobj = {}
         for k, v in iteritems(obj):
             newobj[k] = uncan(v,g)
         return newobj
     else:
         return obj
 
 def uncan_sequence(obj, g=None):
     if istype(obj, sequence_types):
         t = type(obj)
         return t([uncan(i,g) for i in obj])
     else:
         return obj
 
 def _uncan_dependent_hook(dep, g=None):
     dep.check_dependency()
     
 def can_dependent(obj):
     return CannedObject(obj, keys=('f', 'df'), hook=_uncan_dependent_hook)
 
 #-------------------------------------------------------------------------------
 # API dictionaries
 #-------------------------------------------------------------------------------
 
 # These dicts can be extended for custom serialization of new objects
 
 can_map = {
     'IPython.parallel.dependent' : can_dependent,
     'numpy.ndarray' : CannedArray,
     FunctionType : CannedFunction,
     bytes : CannedBytes,
     buffer : CannedBuffer,
     cell_type : CannedCell,
     class_type : can_class,
 }
 
 uncan_map = {
     CannedObject : lambda obj, g: obj.get_object(g),
 }
 
 # for use in _import_mapping:
 _original_can_map = can_map.copy()
 _original_uncan_map = uncan_map.copy()