change docs/examples refs to be just examples

this closes  #3915

docs/source/attic/parallel_task_old.rst

@@ -1,240 +1,240 b''
 .. _paralleltask:
 
 ==========================
 The IPython task interface
 ==========================
 
 .. contents::
 
 The ``Task`` interface to the controller presents the engines as a fault tolerant, dynamic load-balanced system or workers. Unlike the ``MultiEngine`` interface, in the ``Task`` interface, the user have no direct access to individual engines. In some ways, this interface is simpler, but in other ways it is more powerful. Best of all the user can use both of these interfaces at the same time to take advantage or both of their strengths.  When the user can break up the user's work into segments that do not depend on previous execution, the ``Task`` interface is ideal.  But it also has more power and flexibility, allowing the user to guide the distribution of jobs, without having to assign Tasks to engines explicitly.
 
 Starting the IPython controller and engines
 ===========================================
 
 To follow along with this tutorial, the user will need to start the IPython
 controller and four IPython engines. The simplest way of doing this is to
 use the ``ipcluster`` command::
 
 	$ ipcluster -n 4
 	
 For more detailed information about starting the controller and engines, see our :ref:`introduction <parallel_overview>` to using IPython for parallel computing.
 
 The magic here is that this single controller and set of engines is running both the MultiEngine and ``Task`` interfaces simultaneously.
 
 QuickStart Task Farming
 =======================
 
 First, a quick example of how to start running the most basic Tasks.
 The first step is to import the IPython ``client`` module and then create a ``TaskClient`` instance::
     
     In [1]: from IPython.kernel import client
     
     In [2]: tc = client.TaskClient()
 
 Then the user wrap the commands the user want to run in Tasks::
 
     In [3]: tasklist = []
     In [4]: for n in range(1000):
         ...     tasklist.append(client.Task("a = %i"%n, pull="a"))
 
 The first argument of the ``Task`` constructor is a string, the command to be executed.  The most important optional keyword argument is ``pull``, which can be a string or list of strings, and it specifies the variable names to be saved as results of the ``Task``.
 
 Next, the user need to submit the Tasks to the ``TaskController`` with the ``TaskClient``::
 
     In [5]: taskids = [ tc.run(t) for t in tasklist ]
 
 This will give the user a list of the TaskIDs used by the controller to keep track of the Tasks and their results.  Now at some point the user are going to want to get those results back.  The ``barrier`` method allows the user to wait for the Tasks to finish running::
 
     In [6]: tc.barrier(taskids)
 
 This command will block until all the Tasks in ``taskids`` have finished.  Now, the user probably want to look at the user's results::
 
     In [7]: task_results = [ tc.get_task_result(taskid) for taskid in taskids ]
 
 Now the user have a list of ``TaskResult`` objects, which have the actual result as a dictionary, but also keep track of some useful metadata about the ``Task``::
 
     In [8]: tr = ``Task``_results[73]
     
     In [9]: tr
     Out[9]: ``TaskResult``[ID:73]:{'a':73}
     
     In [10]: tr.engineid
     Out[10]: 1
     
     In [11]: tr.submitted, tr.completed, tr.duration
     Out[11]: ("2008/03/08 03:41:42", "2008/03/08 03:41:44", 2.12345)
 
 The actual results are stored in a dictionary, ``tr.results``, and a namespace object ``tr.ns`` which accesses the result keys by attribute::
 
     In [12]: tr.results['a']
     Out[12]: 73
     
     In [13]: tr.ns.a
     Out[13]: 73
 
 That should cover the basics of running simple Tasks.  There are several more powerful things the user can do with Tasks covered later.  The most useful probably being using a ``MutiEngineClient`` interface to initialize all the engines with the import dependencies necessary to run the user's Tasks.
 
-There are many options for running and managing Tasks. The best way to learn further about the ``Task`` interface is to study the examples in ``docs/examples``. If the user do so and learn a lots about this interface, we encourage the user to expand this documentation about the ``Task`` system.
+There are many options for running and managing Tasks. The best way to learn further about the ``Task`` interface is to study the examples in ``examples``. If the user do so and learn a lots about this interface, we encourage the user to expand this documentation about the ``Task`` system.
 
 Overview of the Task System
 ===========================
 
 The user's view of the ``Task`` system has three basic objects:  The ``TaskClient``, the ``Task``, and the ``TaskResult``.  The names of these three objects well indicate their role.
 
 The ``TaskClient`` is the user's ``Task`` farming connection to the IPython cluster.  Unlike the ``MultiEngineClient``, the ``TaskControler`` handles all the scheduling and distribution of work, so the ``TaskClient`` has no notion of engines, it just submits Tasks and requests their results.  The Tasks are described as ``Task`` objects, and their results are wrapped in ``TaskResult`` objects.  Thus, there are very few necessary methods for the user to manage.
 
 Inside the task system is a Scheduler object, which assigns tasks to workers.  The default scheduler is a simple FIFO queue.  Subclassing the Scheduler should be easy, just implementing your own priority system.
 
 The TaskClient
 ==============
 
 The ``TaskClient`` is the object the user use to connect to the ``Controller`` that is managing the user's Tasks.  It is the analog of the ``MultiEngineClient`` for the standard IPython multiplexing interface.  As with all client interfaces, the first step is to import the IPython Client Module::
 
     In [1]: from IPython.kernel import client
 
 Just as with the ``MultiEngineClient``, the user create the ``TaskClient`` with a tuple, containing the ip-address and port of the ``Controller``.  the ``client`` module conveniently has the default address of the ``Task`` interface of the controller.  Creating a default ``TaskClient`` object would be done with this::
 
     In [2]: tc = client.TaskClient(client.default_task_address)
 
 or, if the user want to specify a non default location of the ``Controller``, the user can specify explicitly::
 
     In [3]: tc = client.TaskClient(("192.168.1.1", 10113))
 
 As discussed earlier, the ``TaskClient`` only has a few basic methods.
 
  * ``tc.run(task)``
     ``run`` is the method by which the user submits Tasks.  It takes exactly one argument, a ``Task`` object.  All the advanced control of ``Task`` behavior is handled by properties of the ``Task`` object, rather than the submission command, so they will be discussed later in the `Task`_ section.  ``run`` returns an integer, the ``Task``ID by which the ``Task`` and its results can be tracked and retrieved::
     
         In [4]: ``Task``ID = tc.run(``Task``)
     
  * ``tc.get_task_result(taskid, block=``False``)``
     ``get_task_result`` is the method by which results are retrieved.  It takes a single integer argument, the ``Task``ID`` of the result the user wish to retrieve.  ``get_task_result`` also takes a keyword argument ``block``.  ``block`` specifies whether the user actually want to wait for the result.  If ``block`` is false, as it is by default, ``get_task_result`` will return immediately.  If the ``Task`` has completed, it will return the ``TaskResult`` object for that ``Task``.  But if the ``Task`` has not completed, it will return ``None``.  If the user specify ``block=``True``, then ``get_task_result`` will wait for the ``Task`` to complete, and always return the ``TaskResult`` for the requested ``Task``.
  * ``tc.barrier(taskid(s))``
     ``barrier`` is a synchronization method.  It takes exactly one argument, a ``Task``ID or list of taskIDs.  ``barrier`` will block until all the specified Tasks have completed.  In practice, a barrier is often called between the ``Task`` submission section of the code and the result gathering section::
         
         In [5]: taskIDs = [ tc.run(``Task``) for ``Task`` in myTasks ]
         
         In [6]: tc.get_task_result(taskIDs[-1]) is None
         Out[6]: ``True``
         
         In [7]: tc.barrier(``Task``ID)
         
         In [8]: results = [ tc.get_task_result(tid) for tid in taskIDs ]
         
  * ``tc.queue_status(verbose=``False``)``
     ``queue_status`` is a method for querying the state of the ``TaskControler``.  ``queue_status`` returns a dict of the form::
         
         {'scheduled': Tasks that have been submitted but yet run
          'pending'  : Tasks that are currently running
          'succeeded': Tasks that have completed successfully
          'failed'   : Tasks that have finished with a failure
         }
     
     if @verbose is not specified (or is ``False``), then the values of the dict are integers - the number of Tasks in each state.  if @verbose is ``True``, then each element in the dict is a list of the taskIDs in that state::
     
         In [8]: tc.queue_status()
         Out[8]: {'scheduled': 4,
                 'pending'  : 2,
                 'succeeded': 5,
                 'failed'   : 1
                 }
         
         In [9]: tc.queue_status(verbose=True)
         Out[9]: {'scheduled': [8,9,10,11],
                 'pending'  : [6,7],
                 'succeeded': [0,1,2,4,5],
                 'failed'   : [3]
                 }
         
  * ``tc.abort(taskid)``
     ``abort`` allows the user to abort Tasks that have already been submitted.  ``abort`` will always return immediately.  If the ``Task`` has completed, ``abort`` will raise an ``IndexError ``Task`` Already Completed``.  An obvious case for ``abort`` would be where the user submits a long-running ``Task`` with a number of retries (see ``Task``_ section for how to specify retries) in an interactive session, but realizes there has been a typo.  The user can then abort the ``Task``, preventing certain failures from cluttering up the queue.  It can also be used for parallel search-type problems, where only one ``Task`` will give the solution, so once the user find the solution, the user would want to abort all remaining Tasks to prevent wasted work.
  * ``tc.spin()``
     ``spin`` simply triggers the scheduler in the ``TaskControler``.  Under most normal circumstances, this will do nothing.  The primary known usage case involves the ``Task`` dependency (see `Dependencies`_).  The dependency is a function of an Engine's ``properties``, but changing the ``properties`` via the ``MutliEngineClient`` does not trigger a reschedule event.  The main example case for this requires the following event sequence:
      * ``engine`` is available, ``Task`` is submitted, but ``engine`` does not have ``Task``'s dependencies.
      * ``engine`` gets necessary dependencies while no new Tasks are submitted or completed.
      * now ``engine`` can run ``Task``, but a ``Task`` event is required for the ``TaskControler`` to try scheduling ``Task`` again.
      
     ``spin`` is just an empty ping method to ensure that the Controller has scheduled all available Tasks, and should not be needed under most normal circumstances.
 
 That covers the ``TaskClient``, a simple interface to the cluster.  With this, the user can submit jobs (and abort if necessary), request their results, synchronize on arbitrary subsets of jobs.
 
 .. _task: The Task Object
 
 The Task Object
 ===============
 
 The ``Task`` is the basic object for describing a job.  It can be used in a very simple manner, where the user just specifies a command string to be executed as the ``Task``.  The usage of this first argument is exactly the same as the ``execute`` method of the ``MultiEngine`` (in fact, ``execute`` is called to run the code)::
 
     In [1]: t = client.Task("a = str(id)")
     
 This ``Task`` would run, and store the string representation of the ``id`` element in ``a`` in each worker's namespace, but it is fairly useless because the user does not know anything about the state of the ``worker`` on which it ran at the time of retrieving results.  It is important that each ``Task`` not expect the state of the ``worker`` to persist after the ``Task`` is completed.  
 There are many different situations for using ``Task`` Farming, and the ``Task`` object has many attributes for use in customizing the ``Task`` behavior.  All of a ``Task``'s attributes may be specified in the constructor, through keyword arguments, or after ``Task`` construction through attribute assignment.
 
 Data Attributes
 ***************
 It is likely that the user may want to move data around before or after executing the ``Task``.  We provide methods of sending data to initialize the worker's namespace, and specifying what data to bring back as the ``Task``'s results.
 
  * pull = []
     The obvious case is as above, where ``t`` would execute and store the result of ``myfunc`` in ``a``, it is likely that the user would want to bring ``a`` back to their namespace.  This is done through the ``pull`` attribute.  ``pull`` can be a string or list of strings, and it specifies the names of variables to be retrieved.  The ``TaskResult`` object retrieved by ``get_task_result`` will have a dictionary of keys and values, and the ``Task``'s ``pull`` attribute determines what goes into it::
         
         In [2]: t = client.Task("a = str(id)", pull = "a")
         
         In [3]: t = client.Task("a = str(id)", pull = ["a", "id"])
         
  * push = {}
     A user might also want to initialize some data into the namespace before the code part of the ``Task`` is run.  Enter ``push``.  ``push`` is a dictionary of key/value pairs to be loaded from the user's namespace into the worker's immediately before execution::
 
         In [4]: t = client.Task("a = f(submitted)", push=dict(submitted=time.time()), pull="a")
 
 push and pull result directly in calling an ``engine``'s ``push`` and ``pull`` methods before and after ``Task`` execution respectively, and thus their api is the same.
 
 Namespace Cleaning
 ******************
 When a user is running a large number of Tasks, it is likely that the namespace of the worker's could become cluttered.  Some Tasks might be sensitive to clutter, while others might be known to cause namespace pollution.  For these reasons, Tasks have two boolean attributes for cleaning up the namespace.
 
  * ``clear_after``
     if clear_after is specified ``True``, the worker on which the ``Task`` was run will be reset (via ``engine.reset``) upon completion of the ``Task``.  This can be useful for both Tasks that produce clutter or Tasks whose intermediate data one might wish to be kept private::
     
         In [5]: t = client.Task("a = range(1e10)", pull = "a",clear_after=True)
         
     
  * ``clear_before``
     as one might guess, clear_before is identical to ``clear_after``, but it takes place before the ``Task`` is run.  This ensures that the ``Task`` runs on a fresh worker::
 
         In [6]: t = client.Task("a = globals()", pull = "a",clear_before=True)
 
 Of course, a user can both at the same time, ensuring that all workers are clear except when they are currently running a job.  Both of these default to ``False``.
 
 Fault Tolerance
 ***************
 It is possible that Tasks might fail, and there are a variety of reasons this could happen.  One might be that the worker it was running on disconnected, and there was nothing wrong with the ``Task`` itself.  With the fault tolerance attributes of the ``Task``, the user can specify how many times to resubmit the ``Task``, and what to do if it never succeeds.
 
  * ``retries``
     ``retries`` is an integer, specifying the number of times a ``Task`` is to be retried.  It defaults to zero.  It is often a good idea for this number to be 1 or 2, to protect the ``Task`` from disconnecting engines, but not a large number.  If a ``Task`` is failing 100 times, there is probably something wrong with the ``Task``.  The canonical bad example:
     
         In [7]: t = client.Task("os.kill(os.getpid(), 9)", retries=99)
         
     This would actually take down 100 workers.
     
  * ``recovery_task``
     ``recovery_task`` is another ``Task`` object, to be run in the event of the original ``Task`` still failing after running out of retries.  Since ``recovery_task`` is another ``Task`` object, it can have its own ``recovery_task``.  The chain of Tasks is limitless, except loops are not allowed (that would be bad!).
 
 Dependencies
 ************
 Dependencies are the most powerful part of the ``Task`` farming system, because it allows the user to do some classification of the workers, and guide the ``Task`` distribution without meddling with the controller directly.  It makes use of two objects - the ``Task``'s ``depend`` attribute, and the engine's ``properties``.  See the `MultiEngine`_ reference for how to use engine properties.  The engine properties api exists for extending IPython, allowing conditional execution and new controllers that make decisions based on properties of its engines.  Currently the ``Task`` dependency is the only internal use of the properties api.
 
 .. _MultiEngine: ./parallel_multiengine
 
 The ``depend`` attribute of a ``Task`` must be a function of exactly one argument, the worker's properties dictionary, and it should return ``True`` if the ``Task`` should be allowed to run on the worker and ``False`` if not.  The usage in the controller is fault tolerant, so exceptions raised by ``Task.depend`` will be ignored and functionally equivalent to always returning ``False``.  Tasks`` with invalid ``depend`` functions will never be assigned to a worker::
 
     In [8]: def dep(properties):
         ...     return properties["RAM"] > 2**32 # have at least 4GB
     In [9]: t = client.Task("a = bigfunc()", depend=dep)
     
 It is important to note that assignment of values to the properties dict is done entirely by the user, either locally (in the engine) using the EngineAPI, or remotely, through the ``MultiEngineClient``'s get/set_properties methods.
 
 
 
 
 
     

docs/source/interactive/reference.rst

@@ -1,1165 +1,1165 b''
 =================
 IPython reference
 =================
 
 .. _command_line_options:
 
 Command-line usage
 ==================
 
 You start IPython with the command::
 
     $ ipython [options] files
 
 .. note::
 
   For IPython on Python 3, use ``ipython3`` in place of ``ipython``.
 
 If invoked with no options, it executes all the files listed in sequence
 and drops you into the interpreter while still acknowledging any options
 you may have set in your ipython_config.py. This behavior is different from
 standard Python, which when called as python -i will only execute one
 file and ignore your configuration setup.
 
 Please note that some of the configuration options are not available at
 the command line, simply because they are not practical here. Look into
 your configuration files for details on those. There are separate configuration 
 files for each profile, and the files look like "ipython_config.py" or 
 "ipython_config_<frontendname>.py".  Profile directories look like 
 "profile_profilename" and are typically installed in the IPYTHONDIR directory.
 For Linux users, this will be $HOME/.config/ipython, and for other users it 
 will be $HOME/.ipython.  For Windows users, $HOME resolves to C:\\Documents and
 Settings\\YourUserName in most instances.
 
 
 Eventloop integration
 ---------------------
 
 Previously IPython had command line options for controlling GUI event loop
 integration (-gthread, -qthread, -q4thread, -wthread, -pylab). As of IPython
 version 0.11, these have been removed. Please see the new ``%gui``
 magic command or :ref:`this section <gui_support>` for details on the new
 interface, or specify the gui at the commandline::
 
     $ ipython --gui=qt
 
 
 Command-line Options
 --------------------
 
 To see the options IPython accepts, use ``ipython --help`` (and you probably
 should run the output through a pager such as ``ipython --help | less`` for
 more convenient reading).  This shows all the options that have a single-word
 alias to control them, but IPython lets you configure all of its objects from
 the command-line by passing the full class name and a corresponding value; type
 ``ipython --help-all`` to see this full list.  For example::
 
   ipython --matplotlib qt
 
 is equivalent to::
 
   ipython --TerminalIPythonApp.matplotlib='qt'
 
 Note that in the second form, you *must* use the equal sign, as the expression
 is evaluated as an actual Python assignment.  While in the above example the
 short form is more convenient, only the most common options have a short form,
 while any configurable variable in IPython can be set at the command-line by
 using the long form.  This long form is the same syntax used in the
 configuration files, if you want to set these options permanently.
 
 
 Interactive use
 ===============
 
 IPython is meant to work as a drop-in replacement for the standard interactive
 interpreter. As such, any code which is valid python should execute normally
 under IPython (cases where this is not true should be reported as bugs). It
 does, however, offer many features which are not available at a standard python
 prompt. What follows is a list of these.
 
 
 Caution for Windows users
 -------------------------
 
 Windows, unfortunately, uses the '\\' character as a path separator. This is a
 terrible choice, because '\\' also represents the escape character in most
 modern programming languages, including Python. For this reason, using '/'
 character is recommended if you have problems with ``\``.  However, in Windows
 commands '/' flags options, so you can not use it for the root directory. This
 means that paths beginning at the root must be typed in a contrived manner
 like: ``%copy \opt/foo/bar.txt \tmp``
 
 .. _magic:
 
 Magic command system
 --------------------
 
 IPython will treat any line whose first character is a % as a special
 call to a 'magic' function. These allow you to control the behavior of
 IPython itself, plus a lot of system-type features. They are all
 prefixed with a % character, but parameters are given without
 parentheses or quotes.
 
 Lines that begin with ``%%`` signal a *cell magic*: they take as arguments not
 only the rest of the current line, but all lines below them as well, in the
 current execution block.  Cell magics can in fact make arbitrary modifications
 to the input they receive, which need not even be valid Python code at all.
 They receive the whole block as a single string.
 
 As a line magic example, the ``%cd`` magic works just like the OS command of
 the same name::
 
       In [8]: %cd
       /home/fperez
 
 The following uses the builtin ``timeit`` in cell mode::
   
   In [10]: %%timeit x = range(10000)
       ...: min(x)
       ...: max(x)
       ...: 
   1000 loops, best of 3: 438 us per loop
 
 In this case, ``x = range(10000)`` is called as the line argument, and the
 block with ``min(x)`` and ``max(x)`` is called as the cell body.  The
 ``timeit`` magic receives both.
   
 If you have 'automagic' enabled (as it by default), you don't need to type in
 the single ``%`` explicitly for line magics; IPython will scan its internal
 list of magic functions and call one if it exists. With automagic on you can
 then just type ``cd mydir`` to go to directory 'mydir'::
 
       In [9]: cd mydir
       /home/fperez/mydir
 
 Note that cell magics *always* require an explicit ``%%`` prefix, automagic
 calling only works for line magics.
       
 The automagic system has the lowest possible precedence in name searches, so
 defining an identifier with the same name as an existing magic function will
 shadow it for automagic use. You can still access the shadowed magic function
 by explicitly using the ``%`` character at the beginning of the line.
 
 An example (with automagic on) should clarify all this:
 
 .. sourcecode:: ipython
 
     In [1]: cd ipython     # %cd is called by automagic
     /home/fperez/ipython
 
     In [2]: cd=1 	   # now cd is just a variable
 
     In [3]: cd .. 	   # and doesn't work as a function anymore
     File "<ipython-input-3-9fedb3aff56c>", line 1
       cd ..
           ^
     SyntaxError: invalid syntax
 
 
     In [4]: %cd .. 	   # but %cd always works
     /home/fperez
 
     In [5]: del cd     # if you remove the cd variable, automagic works again
 
     In [6]: cd ipython
 
     /home/fperez/ipython
 
 Defining your own magics
 ++++++++++++++++++++++++
 
 There are two main ways to define your own magic functions: from standalone
 functions and by inheriting from a base class provided by IPython:
 :class:`IPython.core.magic.Magics`.  Below we show code you can place in a file
 that you load from your configuration, such as any file in the ``startup``
 subdirectory of your default IPython profile.
 
 First, let us see the simplest case.  The following shows how to create a line
 magic, a cell one and one that works in both modes, using just plain functions:
 
 .. sourcecode:: python
 
     from IPython.core.magic import (register_line_magic, register_cell_magic,
                                     register_line_cell_magic)
 		
     @register_line_magic
     def lmagic(line):
         "my line magic"
 	return line
 
     @register_cell_magic
     def cmagic(line, cell):
         "my cell magic"
         return line, cell
 
     @register_line_cell_magic
     def lcmagic(line, cell=None):
         "Magic that works both as %lcmagic and as %%lcmagic"
 	if cell is None:
 	    print "Called as line magic"
 	    return line
 	else:
 	    print "Called as cell magic"
             return line, cell
 
     # We delete these to avoid name conflicts for automagic to work
     del lmagic, lcmagic
 
 
 You can also create magics of all three kinds by inheriting from the
 :class:`IPython.core.magic.Magics` class.  This lets you create magics that can
 potentially hold state in between calls, and that have full access to the main
 IPython object:
     
 .. sourcecode:: python
 		
     # This code can be put in any Python module, it does not require IPython
     # itself to be running already.  It only creates the magics subclass but
     # doesn't instantiate it yet.
     from IPython.core.magic import (Magics, magics_class, line_magic,
                                     cell_magic, line_cell_magic)
 
     # The class MUST call this class decorator at creation time
     @magics_class
     class MyMagics(Magics):
 
         @line_magic
         def lmagic(self, line):
             "my line magic"
             print "Full access to the main IPython object:", self.shell
             print "Variables in the user namespace:", self.shell.user_ns.keys()
             return line
 
         @cell_magic
         def cmagic(self, line, cell):
             "my cell magic"
             return line, cell
 
         @line_cell_magic
         def lcmagic(self, line, cell=None):
             "Magic that works both as %lcmagic and as %%lcmagic"
             if cell is None:
                 print "Called as line magic"
                 return line
             else:
                 print "Called as cell magic"
                 return line, cell
 
 
     # In order to actually use these magics, you must register them with a
     # running IPython.  This code must be placed in a file that is loaded once
     # IPython is up and running:
     ip = get_ipython()
     # You can register the class itself without instantiating it.  IPython will
     # call the default constructor on it.
     ip.register_magics(MyMagics)
 
 If you want to create a class with a different constructor that holds
 additional state, then you should always call the parent constructor and
 instantiate the class yourself before registration:
 
 .. sourcecode:: python
 		
     @magics_class
     class StatefulMagics(Magics):
         "Magics that hold additional state"
 	
         def __init__(self, shell, data):
             # You must call the parent constructor
             super(StatefulMagics, self).__init__(shell)
             self.data = data
         
         # etc...
 
     # This class must then be registered with a manually created instance,
     # since its constructor has different arguments from the default:
     ip = get_ipython()
     magics = StatefulMagics(ip, some_data)
     ip.register_magics(magics)
     
 
 In earlier versions, IPython had an API for the creation of line magics (cell
 magics did not exist at the time) that required you to create functions with a
 method-looking signature and to manually pass both the function and the name.
 While this API is no longer recommended, it remains indefinitely supported for
 backwards compatibility purposes.  With the old API, you'd create a magic as
 follows:
 
 .. sourcecode:: python
 
     def func(self, line):
         print "Line magic called with line:", line
 	print "IPython object:", self.shell
 
     ip = get_ipython()
     # Declare this function as the magic %mycommand
     ip.define_magic('mycommand', func)
 
 Type ``%magic`` for more information, including a list of all available magic
 functions at any time and their docstrings. You can also type
 ``%magic_function_name?`` (see :ref:`below <dynamic_object_info>` for
 information on the '?' system) to get information about any particular magic
 function you are interested in.
 
 The API documentation for the :mod:`IPython.core.magic` module contains the full
 docstrings of all currently available magic commands.
 
 
 Access to the standard Python help
 ----------------------------------
 
 Simply type ``help()`` to access Python's standard help system. You can
 also type ``help(object)`` for information about a given object, or
 ``help('keyword')`` for information on a keyword. You may need to configure your
 PYTHONDOCS environment variable for this feature to work correctly.
 
 .. _dynamic_object_info:
 
 Dynamic object information
 --------------------------
 
 Typing ``?word`` or ``word?`` prints detailed information about an object. If
 certain strings in the object are too long (e.g. function signatures) they get
 snipped in the center for brevity. This system gives access variable types and
 values, docstrings, function prototypes and other useful information.
 
 If the information will not fit in the terminal, it is displayed in a pager
 (``less`` if available, otherwise a basic internal pager).
 
 Typing ``??word`` or ``word??`` gives access to the full information, including
 the source code where possible. Long strings are not snipped.
 
 The following magic functions are particularly useful for gathering
 information about your working environment. You can get more details by
 typing ``%magic`` or querying them individually (``%function_name?``);
 this is just a summary:
 
     * **%pdoc <object>**: Print (or run through a pager if too long) the
       docstring for an object. If the given object is a class, it will
       print both the class and the constructor docstrings.
     * **%pdef <object>**: Print the call signature for any callable
       object. If the object is a class, print the constructor information.
     * **%psource <object>**: Print (or run through a pager if too long)
       the source code for an object.
     * **%pfile <object>**: Show the entire source file where an object was
       defined via a pager, opening it at the line where the object
       definition begins.
     * **%who/%whos**: These functions give information about identifiers
       you have defined interactively (not things you loaded or defined
       in your configuration files). %who just prints a list of
       identifiers and %whos prints a table with some basic details about
       each identifier.
 
 Note that the dynamic object information functions (?/??, ``%pdoc``,
 ``%pfile``, ``%pdef``, ``%psource``) work on object attributes, as well as
 directly on variables. For example, after doing ``import os``, you can use
 ``os.path.abspath??``.
 
 .. _readline:
 
 Readline-based features
 -----------------------
 
 These features require the GNU readline library, so they won't work if your
 Python installation lacks readline support. We will first describe the default
 behavior IPython uses, and then how to change it to suit your preferences.
 
 
 Command line completion
 +++++++++++++++++++++++
 
 At any time, hitting TAB will complete any available python commands or
 variable names, and show you a list of the possible completions if
 there's no unambiguous one. It will also complete filenames in the
 current directory if no python names match what you've typed so far.
 
 
 Search command history
 ++++++++++++++++++++++
 
 IPython provides two ways for searching through previous input and thus
 reduce the need for repetitive typing:
 
    1. Start typing, and then use Ctrl-p (previous,up) and Ctrl-n
       (next,down) to search through only the history items that match
       what you've typed so far. If you use Ctrl-p/Ctrl-n at a blank
       prompt, they just behave like normal arrow keys.
    2. Hit Ctrl-r: opens a search prompt. Begin typing and the system
       searches your history for lines that contain what you've typed so
       far, completing as much as it can.
 
 
 Persistent command history across sessions
 ++++++++++++++++++++++++++++++++++++++++++
 
 IPython will save your input history when it leaves and reload it next
 time you restart it. By default, the history file is named
 $IPYTHONDIR/profile_<name>/history.sqlite. This allows you to keep
 separate histories related to various tasks: commands related to
 numerical work will not be clobbered by a system shell history, for
 example.
 
 
 Autoindent
 ++++++++++
 
 IPython can recognize lines ending in ':' and indent the next line,
 while also un-indenting automatically after 'raise' or 'return'.
 
 This feature uses the readline library, so it will honor your
 :file:`~/.inputrc` configuration (or whatever file your INPUTRC variable points
 to). Adding the following lines to your :file:`.inputrc` file can make
 indenting/unindenting more convenient (M-i indents, M-u unindents)::
 
     # if you don't already have a ~/.inputrc file, you need this include:
     $include /etc/inputrc
     
     $if Python 
     "\M-i": "    "  
     "\M-u": "\d\d\d\d"  
     $endif
 
 Note that there are 4 spaces between the quote marks after "M-i" above.
 
 .. warning::
 
     Setting the above indents will cause problems with unicode text entry in
     the terminal.
 
 .. warning::
 
     Autoindent is ON by default, but it can cause problems with the pasting of
     multi-line indented code (the pasted code gets re-indented on each line). A
     magic function %autoindent allows you to toggle it on/off at runtime. You
     can also disable it permanently on in your :file:`ipython_config.py` file
     (set TerminalInteractiveShell.autoindent=False).
     
     If you want to paste multiple lines in the terminal, it is recommended that
     you use ``%paste``.
 
 
 Customizing readline behavior
 +++++++++++++++++++++++++++++
 
 All these features are based on the GNU readline library, which has an
 extremely customizable interface. Normally, readline is configured via a
 file which defines the behavior of the library; the details of the
 syntax for this can be found in the readline documentation available
 with your system or on the Internet. IPython doesn't read this file (if
 it exists) directly, but it does support passing to readline valid
 options via a simple interface. In brief, you can customize readline by
 setting the following options in your configuration file (note
 that these options can not be specified at the command line):
 
     * **readline_parse_and_bind**: this holds a list of strings to be executed
       via a readline.parse_and_bind() command. The syntax for valid commands
       of this kind can be found by reading the documentation for the GNU
       readline library, as these commands are of the kind which readline
       accepts in its configuration file.
     * **readline_remove_delims**: a string of characters to be removed
       from the default word-delimiters list used by readline, so that
       completions may be performed on strings which contain them. Do not
       change the default value unless you know what you're doing.
 
 You will find the default values in your configuration file.
 
 
 Session logging and restoring
 -----------------------------
 
 You can log all input from a session either by starting IPython with the
 command line switch ``--logfile=foo.py`` (see :ref:`here <command_line_options>`)
 or by activating the logging at any moment with the magic function %logstart.
 
 Log files can later be reloaded by running them as scripts and IPython
 will attempt to 'replay' the log by executing all the lines in it, thus
 restoring the state of a previous session. This feature is not quite
 perfect, but can still be useful in many cases.
 
 The log files can also be used as a way to have a permanent record of
 any code you wrote while experimenting. Log files are regular text files
 which you can later open in your favorite text editor to extract code or
 to 'clean them up' before using them to replay a session.
 
 The `%logstart` function for activating logging in mid-session is used as
 follows::
 
     %logstart [log_name [log_mode]]
 
 If no name is given, it defaults to a file named 'ipython_log.py' in your
 current working directory, in 'rotate' mode (see below).
 
 '%logstart name' saves to file 'name' in 'backup' mode. It saves your
 history up to that point and then continues logging.
 
 %logstart takes a second optional parameter: logging mode. This can be
 one of (note that the modes are given unquoted):
 
     * [over:] overwrite existing log_name.
     * [backup:] rename (if exists) to log_name~ and start log_name.
     * [append:] well, that says it.
     * [rotate:] create rotating logs log_name.1~, log_name.2~, etc.
 
 The %logoff and %logon functions allow you to temporarily stop and
 resume logging to a file which had previously been started with
 %logstart. They will fail (with an explanation) if you try to use them
 before logging has been started.
 
 .. _system_shell_access:
 
 System shell access
 -------------------
 
 Any input line beginning with a ! character is passed verbatim (minus
 the !, of course) to the underlying operating system. For example,
 typing ``!ls`` will run 'ls' in the current directory.
 
 Manual capture of command output
 --------------------------------
 
 You can assign the result of a system command to a Python variable with the
 syntax ``myfiles = !ls``. This gets machine readable output from stdout 
 (e.g. without colours), and splits on newlines. To explicitly get this sort of
 output without assigning to a variable, use two exclamation marks (``!!ls``) or
 the ``%sx`` magic command.
 
 The captured list has some convenience features. ``myfiles.n`` or ``myfiles.s``
 returns a string delimited by newlines or spaces, respectively. ``myfiles.p``
 produces `path objects <http://pypi.python.org/pypi/path.py>`_ from the list items.
 See :ref:`string_lists` for details.
 
 IPython also allows you to expand the value of python variables when
 making system calls. Wrap variables or expressions in {braces}::
 
     In [1]: pyvar = 'Hello world' 
     In [2]: !echo "A python variable: {pyvar}"  
     A python variable: Hello world
     In [3]: import math
     In [4]: x = 8
     In [5]: !echo {math.factorial(x)}
     40320
 
 For simple cases, you can alternatively prepend $ to a variable name::
 
     In [6]: !echo $sys.argv  
     [/home/fperez/usr/bin/ipython]
     In [7]: !echo "A system variable: $$HOME"  # Use $$ for literal $
     A system variable: /home/fperez
 
 System command aliases
 ----------------------
 
 The %alias magic function allows you to define magic functions which are in fact
 system shell commands. These aliases can have parameters.
 
 ``%alias alias_name cmd`` defines 'alias_name' as an alias for 'cmd'
 
 Then, typing ``alias_name params`` will execute the system command 'cmd
 params' (from your underlying operating system).
 
 You can also define aliases with parameters using %s specifiers (one per
 parameter). The following example defines the parts function as an
 alias to the command 'echo first %s second %s' where each %s will be
 replaced by a positional parameter to the call to %parts::
 
     In [1]: %alias parts echo first %s second %s
     In [2]: parts A B
     first A second B
     In [3]: parts A  
     ERROR: Alias <parts> requires 2 arguments, 1 given.
 
 If called with no parameters, %alias prints the table of currently
 defined aliases.
 
 The %rehashx magic allows you to load your entire $PATH as
 ipython aliases. See its docstring for further details.
 
 
 .. _dreload:
 
 Recursive reload
 ----------------
 
 The :mod:`IPython.lib.deepreload` module allows you to recursively reload a
 module: changes made to any of its dependencies will be reloaded without
 having to exit. To start using it, do::
 
     from IPython.lib.deepreload import reload as dreload
 
 
 Verbose and colored exception traceback printouts
 -------------------------------------------------
 
 IPython provides the option to see very detailed exception tracebacks,
 which can be especially useful when debugging large programs. You can
 run any Python file with the %run function to benefit from these
 detailed tracebacks. Furthermore, both normal and verbose tracebacks can
 be colored (if your terminal supports it) which makes them much easier
 to parse visually.
 
 See the magic xmode and colors functions for details (just type %magic).
 
 These features are basically a terminal version of Ka-Ping Yee's cgitb
 module, now part of the standard Python library.
 
 
 .. _input_caching:
 
 Input caching system
 --------------------
 
 IPython offers numbered prompts (In/Out) with input and output caching
 (also referred to as 'input history'). All input is saved and can be 
 retrieved as variables (besides the usual arrow key recall), in 
 addition to the %rep magic command that brings a history entry
 up for editing on the next  command line.
 
 The following GLOBAL variables always exist (so don't overwrite them!):
 
 * _i, _ii, _iii: store previous, next previous and next-next previous inputs.
 * In, _ih : a list of all inputs; _ih[n] is the input from line n. If you
   overwrite In with a variable of your own, you can remake the assignment to the
   internal list with a simple ``In=_ih``.
 
 Additionally, global variables named _i<n> are dynamically created (<n>
 being the prompt counter), so ``_i<n> == _ih[<n>] == In[<n>]``.
 
 For example, what you typed at prompt 14 is available as _i14, _ih[14]
 and In[14].
 
 This allows you to easily cut and paste multi line interactive prompts
 by printing them out: they print like a clean string, without prompt
 characters. You can also manipulate them like regular variables (they
 are strings), modify or exec them (typing ``exec _i9`` will re-execute the
 contents of input prompt 9.
 
 You can also re-execute multiple lines of input easily by using the
 magic %rerun or %macro functions. The macro system also allows you to re-execute
 previous lines which include magic function calls (which require special
 processing). Type %macro? for more details on the macro system.
 
 A history function %hist allows you to see any part of your input
 history by printing a range of the _i variables.
 
 You can also search ('grep') through your history by typing 
 ``%hist -g somestring``. This is handy for searching for URLs, IP addresses,
 etc. You can bring history entries listed by '%hist -g' up for editing
 with the %recall command, or run them immediately with %rerun.
 
 .. _output_caching:
 
 Output caching system
 ---------------------
 
 For output that is returned from actions, a system similar to the input
 cache exists but using _ instead of _i. Only actions that produce a
 result (NOT assignments, for example) are cached. If you are familiar
 with Mathematica, IPython's _ variables behave exactly like
 Mathematica's % variables.
 
 The following GLOBAL variables always exist (so don't overwrite them!):
 
     * [_] (a single underscore) : stores previous output, like Python's
       default interpreter.
     * [__] (two underscores): next previous.
     * [___] (three underscores): next-next previous.
 
 Additionally, global variables named _<n> are dynamically created (<n>
 being the prompt counter), such that the result of output <n> is always
 available as _<n> (don't use the angle brackets, just the number, e.g.
 _21).
 
 These variables are also stored in a global dictionary (not a
 list, since it only has entries for lines which returned a result)
 available under the names _oh and Out (similar to _ih and In). So the
 output from line 12 can be obtained as _12, Out[12] or _oh[12]. If you
 accidentally overwrite the Out variable you can recover it by typing
 'Out=_oh' at the prompt.
 
 This system obviously can potentially put heavy memory demands on your
 system, since it prevents Python's garbage collector from removing any
 previously computed results. You can control how many results are kept
 in memory with the option (at the command line or in your configuration
 file) cache_size. If you set it to 0, the whole system is completely
 disabled and the prompts revert to the classic '>>>' of normal Python.
 
 
 Directory history
 -----------------
 
 Your history of visited directories is kept in the global list _dh, and
 the magic %cd command can be used to go to any entry in that list. The
 %dhist command allows you to view this history. Do ``cd -<TAB>`` to
 conveniently view the directory history.
 
 
 Automatic parentheses and quotes
 --------------------------------
 
 These features were adapted from Nathan Gray's LazyPython. They are
 meant to allow less typing for common situations.
 
 
 Automatic parentheses
 +++++++++++++++++++++
 
 Callable objects (i.e. functions, methods, etc) can be invoked like this
 (notice the commas between the arguments)::
 
     In [1]: callable_ob arg1, arg2, arg3
     ------> callable_ob(arg1, arg2, arg3)
 
 You can force automatic parentheses by using '/' as the first character
 of a line. For example::
 
     In [2]: /globals # becomes 'globals()'
 
 Note that the '/' MUST be the first character on the line! This won't work::
 
     In [3]: print /globals # syntax error
 
 In most cases the automatic algorithm should work, so you should rarely
 need to explicitly invoke /. One notable exception is if you are trying
 to call a function with a list of tuples as arguments (the parenthesis
 will confuse IPython)::
 
     In [4]: zip (1,2,3),(4,5,6) # won't work
 
 but this will work::
 
     In [5]: /zip (1,2,3),(4,5,6) 
     ------> zip ((1,2,3),(4,5,6))  
     Out[5]: [(1, 4), (2, 5), (3, 6)]
 
 IPython tells you that it has altered your command line by displaying
 the new command line preceded by ->. e.g.::
 
     In [6]: callable list 
     ------> callable(list)
 
 
 Automatic quoting
 +++++++++++++++++
 
 You can force automatic quoting of a function's arguments by using ','
 or ';' as the first character of a line. For example::
 
     In [1]: ,my_function /home/me  # becomes my_function("/home/me")
 
 If you use ';' the whole argument is quoted as a single string, while ',' splits
 on whitespace::
 
     In [2]: ,my_function a b c    # becomes my_function("a","b","c")
 
     In [3]: ;my_function a b c    # becomes my_function("a b c")
 
 Note that the ',' or ';' MUST be the first character on the line! This
 won't work::
 
     In [4]: x = ,my_function /home/me # syntax error
 
 IPython as your default Python environment
 ==========================================
 
 Python honors the environment variable PYTHONSTARTUP and will execute at
 startup the file referenced by this variable. If you put the following code at
 the end of that file, then IPython will be your working environment anytime you
 start Python::
 
     from IPython.frontend.terminal.ipapp import launch_new_instance
     launch_new_instance()
     raise SystemExit
 
 The ``raise SystemExit`` is needed to exit Python when
 it finishes, otherwise you'll be back at the normal Python '>>>'
 prompt.
 
 This is probably useful to developers who manage multiple Python
 versions and don't want to have correspondingly multiple IPython
 versions. Note that in this mode, there is no way to pass IPython any
 command-line options, as those are trapped first by Python itself.
 
 .. _Embedding:
 
 Embedding IPython
 =================
 
 You can start a regular IPython session with
 
 .. sourcecode:: python
 
     import IPython
     IPython.start_ipython()
 
 at any point in your program.  This will load IPython configuration,
 startup files, and everything, just as if it were a normal IPython session.
 In addition to this,
 it is possible to embed an IPython instance inside your own Python programs.
 This allows you to evaluate dynamically the state of your code,
 operate with your variables, analyze them, etc. Note however that
 any changes you make to values while in the shell do not propagate back
 to the running code, so it is safe to modify your values because you
 won't break your code in bizarre ways by doing so.
 
 .. note::
 
   At present, embedding IPython cannot be done from inside IPython.
   Run the code samples below outside IPython.
 
 This feature allows you to easily have a fully functional python
 environment for doing object introspection anywhere in your code with a
 simple function call. In some cases a simple print statement is enough,
 but if you need to do more detailed analysis of a code fragment this
 feature can be very valuable.
 
 It can also be useful in scientific computing situations where it is
 common to need to do some automatic, computationally intensive part and
 then stop to look at data, plots, etc.
 Opening an IPython instance will give you full access to your data and
 functions, and you can resume program execution once you are done with
 the interactive part (perhaps to stop again later, as many times as
 needed).
 
 The following code snippet is the bare minimum you need to include in
 your Python programs for this to work (detailed examples follow later)::
 
     from IPython import embed
 
     embed() # this call anywhere in your program will start IPython
 
 .. note::
 
     As of 0.13, you can embed an IPython *kernel*, for use with qtconsole,
     etc. via ``IPython.embed_kernel()`` instead of ``IPython.embed()``.
     It should function just the same as regular embed, but you connect
     an external frontend rather than IPython starting up in the local
     terminal.
 
 You can run embedded instances even in code which is itself being run at
 the IPython interactive prompt with '%run <filename>'. Since it's easy
 to get lost as to where you are (in your top-level IPython or in your
 embedded one), it's a good idea in such cases to set the in/out prompts
 to something different for the embedded instances. The code examples
 below illustrate this.
 
 You can also have multiple IPython instances in your program and open
 them separately, for example with different options for data
 presentation. If you close and open the same instance multiple times,
 its prompt counters simply continue from each execution to the next.
 
 Please look at the docstrings in the :mod:`~IPython.frontend.terminal.embed` 
 module for more details on the use of this system.
 
 The following sample file illustrating how to use the embedding
 functionality is provided in the examples directory as example-embed.py.
 It should be fairly self-explanatory:
 
 .. literalinclude:: ../../../examples/core/example-embed.py
     :language: python
 
 Once you understand how the system functions, you can use the following
 code fragments in your programs which are ready for cut and paste:
 
 .. literalinclude:: ../../../examples/core/example-embed-short.py
     :language: python
 
 Using the Python debugger (pdb)
 ===============================
 
 Running entire programs via pdb
 -------------------------------
 
 pdb, the Python debugger, is a powerful interactive debugger which
 allows you to step through code, set breakpoints, watch variables,
 etc.  IPython makes it very easy to start any script under the control
 of pdb, regardless of whether you have wrapped it into a 'main()'
 function or not. For this, simply type '%run -d myscript' at an
 IPython prompt. See the %run command's documentation (via '%run?' or
 in Sec. magic_ for more details, including how to control where pdb
 will stop execution first.
 
 For more information on the use of the pdb debugger, read the included
 pdb.doc file (part of the standard Python distribution). On a stock
 Linux system it is located at /usr/lib/python2.3/pdb.doc, but the
 easiest way to read it is by using the help() function of the pdb module
 as follows (in an IPython prompt)::
 
     In [1]: import pdb
     In [2]: pdb.help()
 
 This will load the pdb.doc document in a file viewer for you automatically.
 
 
 Automatic invocation of pdb on exceptions
 -----------------------------------------
 
 IPython, if started with the ``--pdb`` option (or if the option is set in
 your config file) can call the Python pdb debugger every time your code
 triggers an uncaught exception. This feature
 can also be toggled at any time with the %pdb magic command. This can be
 extremely useful in order to find the origin of subtle bugs, because pdb
 opens up at the point in your code which triggered the exception, and
 while your program is at this point 'dead', all the data is still
 available and you can walk up and down the stack frame and understand
 the origin of the problem.
 
 Furthermore, you can use these debugging facilities both with the
 embedded IPython mode and without IPython at all. For an embedded shell
 (see sec. Embedding_), simply call the constructor with
 ``--pdb`` in the argument string and pdb will automatically be called if an
 uncaught exception is triggered by your code.
 
 For stand-alone use of the feature in your programs which do not use
 IPython at all, put the following lines toward the top of your 'main'
 routine::
 
     import sys
     from IPython.core import ultratb
     sys.excepthook = ultratb.FormattedTB(mode='Verbose',
     color_scheme='Linux', call_pdb=1)
 
 The mode keyword can be either 'Verbose' or 'Plain', giving either very
 detailed or normal tracebacks respectively. The color_scheme keyword can
 be one of 'NoColor', 'Linux' (default) or 'LightBG'. These are the same
 options which can be set in IPython with ``--colors`` and ``--xmode``.
 
 This will give any of your programs detailed, colored tracebacks with
 automatic invocation of pdb.
 
 
 Extensions for syntax processing
 ================================
 
 This isn't for the faint of heart, because the potential for breaking
 things is quite high. But it can be a very powerful and useful feature.
 In a nutshell, you can redefine the way IPython processes the user input
 line to accept new, special extensions to the syntax without needing to
 change any of IPython's own code.
 
 In the IPython/extensions directory you will find some examples
 supplied, which we will briefly describe now. These can be used 'as is'
 (and both provide very useful functionality), or you can use them as a
 starting point for writing your own extensions.
 
 .. _pasting_with_prompts:
 
 Pasting of code starting with Python or IPython prompts
 -------------------------------------------------------
 
 IPython is smart enough to filter out input prompts, be they plain Python ones
 (``>>>`` and ``...``) or IPython ones (``In [N]:`` and ``...:``).  You can
 therefore copy and paste from existing interactive sessions without worry.
 
 The following is a 'screenshot' of how things work, copying an example from the
 standard Python tutorial::
 
     In [1]: >>> # Fibonacci series:
 
     In [2]: ... # the sum of two elements defines the next
 
     In [3]: ... a, b = 0, 1
 
     In [4]: >>> while b < 10:
        ...:     ...     print b
        ...:     ...     a, b = b, a+b
        ...:     
     1
     1
     2
     3
     5
     8
 
 And pasting from IPython sessions works equally well::
 
     In [1]: In [5]: def f(x):
        ...:        ...:     "A simple function"
        ...:        ...:     return x**2
        ...:    ...: 
 
     In [2]: f(3)
     Out[2]: 9
 
 .. _gui_support:
 
 GUI event loop support
 ======================
 
 .. versionadded:: 0.11
     The ``%gui`` magic and :mod:`IPython.lib.inputhook`.
 
 IPython has excellent support for working interactively with Graphical User
 Interface (GUI) toolkits, such as wxPython, PyQt4/PySide, PyGTK and Tk. This is
 implemented using Python's builtin ``PyOSInputHook`` hook. This implementation
 is extremely robust compared to our previous thread-based version. The
 advantages of this are:
 
 * GUIs can be enabled and disabled dynamically at runtime.
 * The active GUI can be switched dynamically at runtime.
 * In some cases, multiple GUIs can run simultaneously with no problems.
 * There is a developer API in :mod:`IPython.lib.inputhook` for customizing 
   all of these things.
 
 For users, enabling GUI event loop integration is simple.  You simple use the
 ``%gui`` magic as follows::
 
     %gui [GUINAME]
 
 With no arguments, ``%gui`` removes all GUI support.  Valid ``GUINAME``
 arguments are ``wx``, ``qt``, ``gtk`` and ``tk``.
 
 Thus, to use wxPython interactively and create a running :class:`wx.App`
 object, do::
 
     %gui wx
 
 For information on IPython's Matplotlib integration (and the ``matplotlib``
 mode) see :ref:`this section <matplotlib_support>`.
 
 For developers that want to use IPython's GUI event loop integration in the
 form of a library, these capabilities are exposed in library form in the
 :mod:`IPython.lib.inputhook` and :mod:`IPython.lib.guisupport` modules.
 Interested developers should see the module docstrings for more information,
 but there are a few points that should be mentioned here.
 
 First, the ``PyOSInputHook`` approach only works in command line settings 
 where readline is activated.  The integration with various eventloops
 is handled somewhat differently (and more simply) when using the standalone
 kernel, as in the qtconsole and notebook.
 
 Second, when using the ``PyOSInputHook`` approach, a GUI application should
 *not* start its event loop. Instead all of this is handled by the
 ``PyOSInputHook``. This means that applications that are meant to be used both
 in IPython and as standalone apps need to have special code to detects how the
 application is being run. We highly recommend using IPython's support for this.
 Since the details vary slightly between toolkits, we point you to the various
-examples in our source directory :file:`docs/examples/lib` that demonstrate
+examples in our source directory :file:`examples/lib` that demonstrate
 these capabilities.
 
 Third, unlike previous versions of IPython, we no longer "hijack" (replace
 them with no-ops) the event loops. This is done to allow applications that
 actually need to run the real event loops to do so. This is often needed to
 process pending events at critical points.
 
 Finally, we also have a number of examples in our source directory
-:file:`docs/examples/lib` that demonstrate these capabilities.
+:file:`examples/lib` that demonstrate these capabilities.
 
 PyQt and PySide
 ---------------
 
 .. attempt at explanation of the complete mess that is Qt support
 
 When you use ``--gui=qt`` or ``--matplotlib=qt``, IPython can work with either
 PyQt4 or PySide.  There are three options for configuration here, because
 PyQt4 has two APIs for QString and QVariant - v1, which is the default on
 Python 2, and the more natural v2, which is the only API supported by PySide.
 v2 is also the default for PyQt4 on Python 3.  IPython's code for the QtConsole
 uses v2, but you can still use any interface in your code, since the
 Qt frontend is in a different process.
 
 The default will be to import PyQt4 without configuration of the APIs, thus
 matching what most applications would expect. It will fall back of PySide if
 PyQt4 is unavailable.
 
 If specified, IPython will respect the environment variable ``QT_API`` used
 by ETS.  ETS 4.0 also works with both PyQt4 and PySide, but it requires
 PyQt4 to use its v2 API.  So if ``QT_API=pyside`` PySide will be used,
 and if ``QT_API=pyqt`` then PyQt4 will be used *with the v2 API* for
 QString and QVariant, so ETS codes like MayaVi will also work with IPython.
 
 If you launch IPython in matplotlib mode with ``ipython --matplotlib=qt``,
 then IPython will ask matplotlib which Qt library to use (only if QT_API is
 *not set*), via the 'backend.qt4' rcParam.  If matplotlib is version 1.0.1 or
 older, then IPython will always use PyQt4 without setting the v2 APIs, since
 neither v2 PyQt nor PySide work.
 
 .. warning::
 
     Note that this means for ETS 4 to work with PyQt4, ``QT_API`` *must* be set
     to work with IPython's qt integration, because otherwise PyQt4 will be
     loaded in an incompatible mode.
     
     It also means that you must *not* have ``QT_API`` set if you want to
     use ``--gui=qt`` with code that requires PyQt4 API v1.
 
 
 .. _matplotlib_support:
 
 Plotting with matplotlib
 ========================
 
 `Matplotlib`_ provides high quality 2D and 3D plotting for Python. Matplotlib
 can produce plots on screen using a variety of GUI toolkits, including Tk,
 PyGTK, PyQt4 and wxPython. It also provides a number of commands useful for
 scientific computing, all with a syntax compatible with that of the popular
 Matlab program.
 
 To start IPython with matplotlib support, use the ``--matplotlib`` switch. If
 IPython is already running, you can run the ``%matplotlib`` magic.  If no
 arguments are given, IPython will automatically detect your choice of
 matplotlib backend.  You can also request a specific backend with
 ``%matplotlib backend``, where ``backend`` must be one of: 'tk', 'qt', 'wx',
 'gtk', 'osx'.  In the web notebook and Qt console, 'inline' is also a valid
 backend value, which produces static figures inlined inside the application
 window instead of matplotlib's interactive figures that live in separate
 windows.
 
 .. _Matplotlib: http://matplotlib.sourceforge.net
 
 .. _interactive_demos:
 
 Interactive demos with IPython
 ==============================
 
 IPython ships with a basic system for running scripts interactively in
 sections, useful when presenting code to audiences. A few tags embedded
 in comments (so that the script remains valid Python code) divide a file
 into separate blocks, and the demo can be run one block at a time, with
 IPython printing (with syntax highlighting) the block before executing
 it, and returning to the interactive prompt after each block. The
 interactive namespace is updated after each block is run with the
 contents of the demo's namespace.
 
 This allows you to show a piece of code, run it and then execute
 interactively commands based on the variables just created. Once you
 want to continue, you simply execute the next block of the demo. The
 following listing shows the markup necessary for dividing a script into
 sections for execution as a demo:
 
 .. literalinclude:: ../../../examples/lib/example-demo.py
     :language: python
 
 In order to run a file as a demo, you must first make a Demo object out
 of it. If the file is named myscript.py, the following code will make a
 demo::
 
     from IPython.lib.demo import Demo
 
     mydemo = Demo('myscript.py')
 
 This creates the mydemo object, whose blocks you run one at a time by
 simply calling the object with no arguments. If you have autocall active
 in IPython (the default), all you need to do is type::
 
     mydemo 
 
 and IPython will call it, executing each block. Demo objects can be
 restarted, you can move forward or back skipping blocks, re-execute the
 last block, etc. Simply use the Tab key on a demo object to see its
 methods, and call '?' on them to see their docstrings for more usage
 details. In addition, the demo module itself contains a comprehensive
 docstring, which you can access via::
 
     from IPython.lib import demo
 
     demo?
 
 Limitations: It is important to note that these demos are limited to
 fairly simple uses. In particular, you cannot break up sections within
 indented code (loops, if statements, function definitions, etc.)
 Supporting something like this would basically require tracking the
 internal execution state of the Python interpreter, so only top-level
 divisions are allowed. If you want to be able to open an IPython
 instance at an arbitrary point in a program, you can use IPython's
 embedding facilities, see :func:`IPython.embed` for details.
 

docs/source/parallel/asyncresult.rst

@@ -1,150 +1,150 b''
 .. _parallel_asyncresult:
 
 ======================
 The AsyncResult object
 ======================
 
 In non-blocking mode, :meth:`apply` submits the command to be executed and
 then returns a :class:`~.AsyncResult` object immediately. The
 AsyncResult object gives you a way of getting a result at a later
 time through its :meth:`get` method, but it also collects metadata
 on execution.
 
 
 Beyond multiprocessing's AsyncResult
 ====================================
 
 .. Note::
 
     The :class:`~.AsyncResult` object provides a superset of the interface in 
     :py:class:`multiprocessing.pool.AsyncResult`.  See the 
     `official Python documentation <http://docs.python.org/library/multiprocessing#multiprocessing.pool.AsyncResult>`_
     for more on the basics of this interface.
 
 Our AsyncResult objects add a number of convenient features for working with
 parallel results, beyond what is provided by the original AsyncResult.
 
 
 get_dict
 --------
 
 First, is :meth:`.AsyncResult.get_dict`, which pulls results as a dictionary
 keyed by engine_id, rather than a flat list.  This is useful for quickly
 coordinating or distributing information about all of the engines.
 
 As an example, here is a quick call that gives every engine a dict showing
 the PID of every other engine:
 
 .. sourcecode:: ipython
 
     In [10]: ar = rc[:].apply_async(os.getpid)
     In [11]: pids = ar.get_dict()
     In [12]: rc[:]['pid_map'] = pids
 
 This trick is particularly useful when setting up inter-engine communication,
 as in IPython's :file:`examples/parallel/interengine` examples.
 
 
 Metadata
 ========
 
 IPython.parallel tracks some metadata about the tasks, which is stored
 in the :attr:`.Client.metadata` dict.  The AsyncResult object gives you an
 interface for this information as well, including timestamps stdout/err,
 and engine IDs.
 
 
 Timing
 ------
 
 IPython tracks various timestamps as :py:class:`.datetime` objects,
 and the AsyncResult object has a few properties that turn these into useful
 times (in seconds as floats).
 
 For use while the tasks are still pending:
 
 * :attr:`ar.elapsed` is just the elapsed seconds since submission, for use
   before the AsyncResult is complete.
 * :attr:`ar.progress` is the number of tasks that have completed.  Fractional progress
   would be::
 
     1.0 * ar.progress / len(ar)
 
 * :meth:`AsyncResult.wait_interactive` will wait for the result to finish, but
   print out status updates on progress and elapsed time while it waits.
 
 For use after the tasks are done:
 
 * :attr:`ar.serial_time` is the sum of the computation time of all of the tasks
   done in parallel.
 * :attr:`ar.wall_time` is the time between the first task submitted and last result
   received.  This is the actual cost of computation, including IPython overhead. 
   
 
 .. note::
 
   wall_time is only precise if the Client is waiting for results when
   the task finished, because the `received` timestamp is made when the result is
   unpacked by the Client, triggered by the :meth:`~Client.spin` call. If you
   are doing work in the Client, and not waiting/spinning, then `received` might
   be artificially high.
 
 An often interesting metric is the time it actually cost to do the work in parallel
 relative to the serial computation, and this can be given simply with
 
 .. sourcecode:: python
 
     speedup = ar.serial_time / ar.wall_time
 
 
 Map results are iterable!
 =========================
 
 When an AsyncResult object has multiple results (e.g. the :class:`~AsyncMapResult`
 object), you can actually iterate through results themselves, and act on them as they arrive:
 
 .. literalinclude:: ../../../examples/parallel/itermapresult.py
     :language: python
     :lines: 20-67
     
 That is to say, if you treat an AsyncMapResult as if it were a list of your actual
 results, it should behave as you would expect, with the only difference being
 that you can start iterating through the results before they have even been computed.
 
 This lets you do a dumb version of map/reduce with the builtin Python functions,
 and the only difference between doing this locally and doing it remotely in parallel
 is using the asynchronous view.map instead of the builtin map.
 
 
 Here is a simple one-line RMS (root-mean-square) implemented with Python's builtin map/reduce.
 
 .. sourcecode:: ipython
 
     In [38]: X = np.linspace(0,100)
 
     In [39]: from math import sqrt
 
     In [40]: add = lambda a,b: a+b
 
     In [41]: sq = lambda x: x*x
 
     In [42]: sqrt(reduce(add, map(sq, X)) / len(X))
     Out[42]: 58.028845747399714
 
     In [43]: sqrt(reduce(add, view.map(sq, X)) / len(X))
     Out[43]: 58.028845747399714
 
 To break that down:
 
 1. ``map(sq, X)`` Compute the square of each element in the list (locally, or in parallel)
 2. ``reduce(add, sqX) / len(X)`` compute the mean by summing over the list (or AsyncMapResult)
    and dividing by the size
 3. take the square root of the resulting number
 
 .. seealso::
 
     When AsyncResult or the AsyncMapResult don't provide what you need (for instance,
     handling individual results as they arrive, but with metadata), you can always
     just split the original result's ``msg_ids`` attribute, and handle them as you like.
     
-    For an example of this, see :file:`docs/examples/parallel/customresult.py`
+    For an example of this, see :file:`examples/parallel/customresult.py`

docs/source/parallel/dag_dependencies.rst

@@ -1,177 +1,177 b''
 .. _dag_dependencies:
 
 ================
 DAG Dependencies
 ================
 
 Often, parallel workflow is described in terms of a `Directed Acyclic Graph
 <http://en.wikipedia.org/wiki/Directed_acyclic_graph>`_ or DAG.  A popular library
 for working with Graphs is NetworkX_.  Here, we will walk through a demo mapping
 a nx DAG to task dependencies.
 
 The full script that runs this demo can be found in
-:file:`docs/examples/parallel/dagdeps.py`.
+:file:`examples/parallel/dagdeps.py`.
 
 Why are DAGs good for task dependencies?
 ----------------------------------------
 
 The 'G' in DAG is 'Graph'. A Graph is a collection of **nodes** and **edges** that connect
 the nodes. For our purposes, each node would be a task, and each edge would be a
 dependency. The 'D' in DAG stands for 'Directed'. This means that each edge has a
 direction associated with it. So we can interpret the edge (a,b) as meaning that b depends
 on a, whereas the edge (b,a) would mean a depends on b. The 'A' is 'Acyclic', meaning that
 there must not be any closed loops in the graph. This is important for dependencies,
 because if a loop were closed, then a task could ultimately depend on itself, and never be
 able to run. If your workflow can be described as a DAG, then it is impossible for your
 dependencies to cause a deadlock.
 
 A Sample DAG
 ------------
 
 Here, we have a very simple 5-node DAG:
 
 .. figure:: figs/simpledag.*
     :width: 600px
 
 With NetworkX, an arrow is just a fattened bit on the edge. Here, we can see that task 0
 depends on nothing, and can run immediately. 1 and 2 depend on 0; 3 depends on
 1 and 2; and 4 depends only on 1.
 
 A possible sequence of events for this workflow:
 
 0. Task 0 can run right away
 1. 0 finishes, so 1,2 can start
 2. 1 finishes, 3 is still waiting on 2, but 4 can start right away
 3. 2 finishes, and 3 can finally start
 
 
 Further, taking failures into account, assuming all dependencies are run with the default
 `success=True,failure=False`, the following cases would occur for each node's failure:
 
 0. fails: all other tasks fail as Impossible
 1. 2 can still succeed, but 3,4 are unreachable
 2. 3 becomes unreachable, but 4 is unaffected
 3. and 4. are terminal, and can have no effect on other nodes
 
 The code to generate the simple DAG:
 
 .. sourcecode:: python
 
     import networkx as nx
     
     G = nx.DiGraph()
     
     # add 5 nodes, labeled 0-4:
     map(G.add_node, range(5))
     # 1,2 depend on 0:
     G.add_edge(0,1)
     G.add_edge(0,2)
     # 3 depends on 1,2
     G.add_edge(1,3)
     G.add_edge(2,3)
     # 4 depends on 1
     G.add_edge(1,4)
     
     # now draw the graph:
     pos = { 0 : (0,0), 1 : (1,1), 2 : (-1,1),
             3 : (0,2), 4 : (2,2)}
     nx.draw(G, pos, edge_color='r')
 
 
 For demonstration purposes, we have a function that generates a random DAG with a given
 number of nodes and edges.
 
 .. literalinclude:: ../../../examples/parallel/dagdeps.py
     :language: python
     :lines: 20-36
 
 So first, we start with a graph of 32 nodes, with 128 edges:
 
 .. sourcecode:: ipython
 
     In [2]: G = random_dag(32,128)
 
 Now, we need to build our dict of jobs corresponding to the nodes on the graph:
 
 .. sourcecode:: ipython
 
     In [3]: jobs = {}
     
     # in reality, each job would presumably be different
     # randomwait is just a function that sleeps for a random interval
     In [4]: for node in G:
        ...:     jobs[node] = randomwait 
 
 Once we have a dict of jobs matching the nodes on the graph, we can start submitting jobs,
 and linking up the dependencies. Since we don't know a job's msg_id until it is submitted,
 which is necessary for building dependencies, it is critical that we don't submit any jobs
 before other jobs it may depend on. Fortunately, NetworkX provides a
 :meth:`topological_sort` method which ensures exactly this. It presents an iterable, that
 guarantees that when you arrive at a node, you have already visited all the nodes it
 on which it depends:
 
 .. sourcecode:: ipython
 
     In [5]: rc = Client()
     In [5]: view = rc.load_balanced_view()
     
     In [6]: results = {}
     
     In [7]: for node in G.topological_sort():
        ...:    # get list of AsyncResult objects from nodes
        ...:    # leading into this one as dependencies
        ...:    deps = [ results[n] for n in G.predecessors(node) ]
        ...:    # submit and store AsyncResult object
        ...:    with view.temp_flags(after=deps, block=False):
        ...:         results[node] = view.apply_with_flags(jobs[node])
 
 
 Now that we have submitted all the jobs, we can wait for the results:
 
 .. sourcecode:: ipython
 
     In [8]: view.wait(results.values())
 
 Now, at least we know that all the jobs ran and did not fail (``r.get()`` would have
 raised an error if a task failed).  But we don't know that the ordering was properly
 respected.  For this, we can use the :attr:`metadata` attribute of each AsyncResult.
 
 These objects store a variety of metadata about each task, including various timestamps.
 We can validate that the dependencies were respected by checking that each task was
 started after all of its predecessors were completed:
 
 .. literalinclude:: ../../../examples/parallel/dagdeps.py
     :language: python
     :lines: 64-70
 
 We can also validate the graph visually. By drawing the graph with each node's x-position
 as its start time, all arrows must be pointing to the right if dependencies were respected.
 For spreading, the y-position will be the runtime of the task, so long tasks
 will be at the top, and quick, small tasks will be at the bottom.
 
 .. sourcecode:: ipython
 
     In [10]: from matplotlib.dates import date2num
     
     In [11]: from matplotlib.cm import gist_rainbow
     
     In [12]: pos = {}; colors = {}
     
     In [12]: for node in G:
        ....:    md = results[node].metadata
        ....:    start = date2num(md.started)
        ....:    runtime = date2num(md.completed) - start
        ....:    pos[node] = (start, runtime)
        ....:    colors[node] = md.engine_id
     
     In [13]: nx.draw(G, pos, node_list=colors.keys(), node_color=colors.values(),
        ....:    cmap=gist_rainbow)
 
 .. figure:: figs/dagdeps.*
     :width: 600px
 
     Time started on x, runtime on y, and color-coded by engine-id (in this case there
     were four engines). Edges denote dependencies.
 
 
 .. _NetworkX: http://networkx.lanl.gov/

docs/source/parallel/parallel_demos.rst

@@ -1,208 +1,208 b''
 .. _parallel_examples:
 
 =================
 Parallel examples
 =================
 
 In this section we describe two more involved examples of using an IPython
 cluster to perform a parallel computation. We will be doing some plotting,
 so we start IPython with matplotlib integration by typing::
 
     ipython --matplotlib
 
 at the system command line.
 Or you can enable matplotlib integration at any point with:
 
 .. sourcecode:: ipython
 
     In [1]: %matplotlib
 
 
 150 million digits of pi
 ========================
 
 In this example we would like to study the distribution of digits in the
 number pi (in base 10). While it is not known if pi is a normal number (a
 number is normal in base 10 if 0-9 occur with equal likelihood) numerical
 investigations suggest that it is. We will begin with a serial calculation on
 10,000 digits of pi and then perform a parallel calculation involving 150
 million digits.
 
 In both the serial and parallel calculation we will be using functions defined
 in the :file:`pidigits.py` file, which is available in the
-:file:`docs/examples/parallel` directory of the IPython source distribution.
+:file:`examples/parallel` directory of the IPython source distribution.
 These functions provide basic facilities for working with the digits of pi and
 can be loaded into IPython by putting :file:`pidigits.py` in your current
 working directory and then doing:
 
 .. sourcecode:: ipython
 
     In [1]: run pidigits.py
 
 Serial calculation
 ------------------
 
 For the serial calculation, we will use `SymPy <http://www.sympy.org>`_ to
 calculate 10,000 digits of pi and then look at the frequencies of the digits
 0-9. Out of 10,000 digits, we expect each digit to occur 1,000 times. While
 SymPy is capable of calculating many more digits of pi, our purpose here is to
 set the stage for the much larger parallel calculation.
 
 In this example, we use two functions from :file:`pidigits.py`: 
 :func:`one_digit_freqs` (which calculates how many times each digit occurs)
 and :func:`plot_one_digit_freqs` (which uses Matplotlib to plot the result).
 Here is an interactive IPython session that uses these functions with
 SymPy:
 
 .. sourcecode:: ipython
 
     In [7]: import sympy
 
     In [8]: pi = sympy.pi.evalf(40)
 
     In [9]: pi
     Out[9]: 3.141592653589793238462643383279502884197
 
     In [10]: pi = sympy.pi.evalf(10000)
 
     In [11]: digits = (d for d in str(pi)[2:])  # create a sequence of digits
 
     In [13]: freqs = one_digit_freqs(digits)
 
     In [14]: plot_one_digit_freqs(freqs)
     Out[14]: [<matplotlib.lines.Line2D object at 0x18a55290>]
 
 The resulting plot of the single digit counts shows that each digit occurs
 approximately 1,000 times, but that with only 10,000 digits the
 statistical fluctuations are still rather large:
 
 .. image:: figs/single_digits.*
 
 It is clear that to reduce the relative fluctuations in the counts, we need
 to look at many more digits of pi. That brings us to the parallel calculation.
 
 Parallel calculation
 --------------------
 
 Calculating many digits of pi is a challenging computational problem in itself. 
 Because we want to focus on the distribution of digits in this example, we
 will use pre-computed digit of pi from the website of Professor Yasumasa
 Kanada at the University of Tokyo (http://www.super-computing.org). These
 digits come in a set of text files (ftp://pi.super-computing.org/.2/pi200m/)
 that each have 10 million digits of pi.
 
 For the parallel calculation, we have copied these files to the local hard
 drives of the compute nodes. A total of 15 of these files will be used, for a
 total of 150 million digits of pi. To make things a little more interesting we
 will calculate the frequencies of all 2 digits sequences (00-99) and then plot
 the result using a 2D matrix in Matplotlib.
 
 The overall idea of the calculation is simple: each IPython engine will
 compute the two digit counts for the digits in a single file. Then in a final
 step the counts from each engine will be added up. To perform this
 calculation, we will need two top-level functions from :file:`pidigits.py`:
 
 .. literalinclude:: ../../../examples/parallel/pi/pidigits.py
    :language: python
    :lines: 47-62
 
 We will also use the :func:`plot_two_digit_freqs` function to plot the
 results. The code to run this calculation in parallel is contained in
-:file:`docs/examples/parallel/parallelpi.py`. This code can be run in parallel
+:file:`examples/parallel/parallelpi.py`. This code can be run in parallel
 using IPython by following these steps:
 
 1. Use :command:`ipcluster` to start 15 engines. We used 16 cores of an SGE linux
    cluster (1 controller + 15 engines).
 2. With the file :file:`parallelpi.py` in your current working directory, open
    up IPython, enable matplotlib, and type ``run parallelpi.py``.  This will download
    the pi files via ftp the first time you run it, if they are not
    present in the Engines' working directory.
 
 When run on our 16 cores, we observe a speedup of 14.2x. This is slightly
 less than linear scaling (16x) because the controller is also running on one of
 the cores.
 
 To emphasize the interactive nature of IPython, we now show how the
 calculation can also be run by simply typing the commands from
 :file:`parallelpi.py` interactively into IPython:
 
 .. sourcecode:: ipython
 
     In [1]: from IPython.parallel import Client
 
     # The Client allows us to use the engines interactively.
     # We simply pass Client the name of the cluster profile we
     # are using.
     In [2]: c = Client(profile='mycluster')
     In [3]: v = c[:]
 
     In [3]: c.ids
     Out[3]: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]
 
     In [4]: run pidigits.py
 
     In [5]: filestring = 'pi200m.ascii.%(i)02dof20'
 
     # Create the list of files to process.
     In [6]: files = [filestring % {'i':i} for i in range(1,16)]
 
     In [7]: files
     Out[7]: 
     ['pi200m.ascii.01of20',
      'pi200m.ascii.02of20',
      'pi200m.ascii.03of20',
      'pi200m.ascii.04of20',
      'pi200m.ascii.05of20',
      'pi200m.ascii.06of20',
      'pi200m.ascii.07of20',
      'pi200m.ascii.08of20',
      'pi200m.ascii.09of20',
      'pi200m.ascii.10of20',
      'pi200m.ascii.11of20',
      'pi200m.ascii.12of20',
      'pi200m.ascii.13of20',
      'pi200m.ascii.14of20',
      'pi200m.ascii.15of20']
     
     # download the data files if they don't already exist:
     In [8]: v.map(fetch_pi_file, files)
     
     # This is the parallel calculation using the Client.map method 
     # which applies compute_two_digit_freqs to each file in files in parallel.
     In [9]: freqs_all = v.map(compute_two_digit_freqs, files)
 
     # Add up the frequencies from each engine.
     In [10]: freqs = reduce_freqs(freqs_all)
 
     In [11]: plot_two_digit_freqs(freqs)
     Out[11]: <matplotlib.image.AxesImage object at 0x18beb110>
 
     In [12]: plt.title('2 digit counts of 150m digits of pi')
     Out[12]: <matplotlib.text.Text object at 0x18d1f9b0>
 
 The resulting plot generated by Matplotlib is shown below. The colors indicate
 which two digit sequences are more (red) or less (blue) likely to occur in the
 first 150 million digits of pi. We clearly see that the sequence "41" is 
 most likely and that "06" and "07" are least likely. Further analysis would
 show that the relative size of the statistical fluctuations have decreased
 compared to the 10,000 digit calculation.
 
 .. image:: figs/two_digit_counts.*
 
 Conclusion
 ==========
 
 To conclude these examples, we summarize the key features of IPython's
 parallel architecture that have been demonstrated:
 
 * Serial code can be parallelized often with only a few extra lines of code.
   We have used the :class:`DirectView` and :class:`LoadBalancedView` classes
   for this purpose.
 * The resulting parallel code can be run without ever leaving the IPython's
   interactive shell.
 * Any data computed in parallel can be explored interactively through 
   visualization or further numerical calculations.
 * We have run these examples on a cluster running RHEL 5 and Sun GridEngine.
   IPython's built in support for SGE (and other batch systems) makes it easy
   to get started with IPython's parallel capabilities.
 

docs/source/parallel/parallel_intro.rst

@@ -1,306 +1,306 b''
 .. _parallel_overview:
 
 ============================
 Overview and getting started
 ============================
 
 
 Examples
 ========
 
 We have various example scripts and notebooks for using IPython.parallel in our
-:file:`docs/examples/parallel` directory, or they can be found `on GitHub`__.
+:file:`examples/parallel` directory, or they can be found `on GitHub`__.
 Some of these are covered in more detail in the :ref:`examples
 <parallel_examples>` section.
 
-.. __: https://github.com/ipython/ipython/tree/master/docs/examples/parallel
+.. __: https://github.com/ipython/ipython/tree/master/examples/parallel
 
 Introduction
 ============
 
 This section gives an overview of IPython's sophisticated and powerful
 architecture for parallel and distributed computing. This architecture
 abstracts out parallelism in a very general way, which enables IPython to
 support many different styles of parallelism including:
 
 * Single program, multiple data (SPMD) parallelism.
 * Multiple program, multiple data (MPMD) parallelism.
 * Message passing using MPI.
 * Task farming.
 * Data parallel.
 * Combinations of these approaches.
 * Custom user defined approaches.
 
 Most importantly, IPython enables all types of parallel applications to
 be developed, executed, debugged and monitored *interactively*. Hence,
 the ``I`` in IPython.  The following are some example usage cases for IPython:
 
 * Quickly parallelize algorithms that are embarrassingly parallel
   using a number of simple approaches.  Many simple things can be
   parallelized interactively in one or two lines of code.
 
 * Steer traditional MPI applications on a supercomputer from an
   IPython session on your laptop.
 
 * Analyze and visualize large datasets (that could be remote and/or
   distributed) interactively using IPython and tools like
   matplotlib/TVTK.
 
 * Develop, test and debug new parallel algorithms 
   (that may use MPI) interactively.
 
 * Tie together multiple MPI jobs running on different systems into
   one giant distributed and parallel system.
 
 * Start a parallel job on your cluster and then have a remote 	
   collaborator connect to it and pull back data into their 
   local IPython session for plotting and analysis.
 
 * Run a set of tasks on a set of CPUs using dynamic load balancing.
 
 .. tip::
 
    At the SciPy 2011 conference in Austin, Min Ragan-Kelley presented a
    complete 4-hour tutorial on the use of these features, and all the materials
    for the tutorial are now `available online`__.  That tutorial provides an
    excellent, hands-on oriented complement to the reference documentation
    presented here.
 
 .. __: http://minrk.github.com/scipy-tutorial-2011
 
 Architecture overview
 =====================
 
 .. figure:: figs/wideView.png
     :width: 300px
 
 
 The IPython architecture consists of four components:
 
 * The IPython engine.
 * The IPython hub.
 * The IPython schedulers.
 * The controller client.
 
 These components live in the :mod:`IPython.parallel` package and are
 installed with IPython.  They do, however, have additional dependencies
 that must be installed.  For more information, see our
 :ref:`installation documentation <install_index>`.
 
 .. TODO: include zmq in install_index
 
 IPython engine
 ---------------
 
 The IPython engine is a Python instance that takes Python commands over a
 network connection. Eventually, the IPython engine will be a full IPython
 interpreter, but for now, it is a regular Python interpreter. The engine
 can also handle incoming and outgoing Python objects sent over a network
 connection.  When multiple engines are started, parallel and distributed
 computing becomes possible. An important feature of an IPython engine is
 that it blocks while user code is being executed. Read on for how the
 IPython controller solves this problem to expose a clean asynchronous API
 to the user.
 
 IPython controller
 ------------------
 
 The IPython controller processes provide an interface for working with a set of engines.
 At a general level, the controller is a collection of processes to which IPython engines
 and clients can connect. The controller is composed of a :class:`Hub` and a collection of
 :class:`Schedulers`. These Schedulers are typically run in separate processes but on the
 same machine as the Hub, but can be run anywhere from local threads or on remote machines.
 
 The controller also provides a single point of contact for users who wish to
 utilize the engines connected to the controller. There are different ways of
 working with a controller. In IPython, all of these models are implemented via
 the :meth:`.View.apply` method, after
 constructing :class:`.View` objects to represent subsets of engines. The two
 primary models for interacting with engines are:
 
 * A **Direct** interface, where engines are addressed explicitly.
 * A **LoadBalanced** interface, where the Scheduler is trusted with assigning work to
   appropriate engines.
 
 Advanced users can readily extend the View models to enable other
 styles of parallelism. 
 
 .. note:: 
 
     A single controller and set of engines can be used with multiple models
     simultaneously. This opens the door for lots of interesting things.
 
 
 The Hub
 *******
 
 The center of an IPython cluster is the Hub. This is the process that keeps
 track of engine connections, schedulers, clients, as well as all task requests and
 results. The primary role of the Hub is to facilitate queries of the cluster state, and
 minimize the necessary information required to establish the many connections involved in
 connecting new clients and engines.
 
 
 Schedulers
 **********
 
 All actions that can be performed on the engine go through a Scheduler. While the engines
 themselves block when user code is run, the schedulers hide that from the user to provide
 a fully asynchronous interface to a set of engines.
 
 
 IPython client and views
 ------------------------
 
 There is one primary object, the :class:`~.parallel.Client`, for connecting to a cluster.
 For each execution model, there is a corresponding :class:`~.parallel.View`. These views
 allow users to interact with a set of engines through the interface. Here are the two default
 views:
 
 * The :class:`DirectView` class for explicit addressing.
 * The :class:`LoadBalancedView` class for destination-agnostic scheduling.
 
 Security
 --------
 
 IPython uses ZeroMQ for networking, which has provided many advantages, but
 one of the setbacks is its utter lack of security [ZeroMQ]_. By default, no IPython
 connections are encrypted, but open ports only listen on localhost. The only
 source of security for IPython is via ssh-tunnel. IPython supports both shell
 (`openssh`) and `paramiko` based tunnels for connections.  There is a key necessary
 to submit requests, but due to the lack of encryption, it does not provide
 significant security if loopback traffic is compromised.
 
 In our architecture, the controller is the only process that listens on
 network ports, and is thus the main point of vulnerability. The standard model
 for secure connections is to designate that the controller listen on
 localhost, and use ssh-tunnels to connect clients and/or
 engines.
 
 To connect and authenticate to the controller an engine or client needs
 some information that the controller has stored in a JSON file.
 Thus, the JSON files need to be copied to a location where
 the clients and engines can find them. Typically, this is the
 :file:`~/.ipython/profile_default/security` directory on the host where the 
 client/engine is running (which could be a different host than the controller). 
 Once the JSON files are copied over, everything should work fine.
 
 Currently, there are two JSON files that the controller creates:
 
 ipcontroller-engine.json
     This JSON file has the information necessary for an engine to connect
     to a controller.
 
 ipcontroller-client.json
     The client's connection information.  This may not differ from the engine's,
     but since the controller may listen on different ports for clients and
     engines, it is stored separately.
 
 ipcontroller-client.json will look something like this, under default localhost
 circumstances:
 
 .. sourcecode:: python
 
     {
       "url":"tcp:\/\/127.0.0.1:54424",
       "exec_key":"a361fe89-92fc-4762-9767-e2f0a05e3130",
       "ssh":"",
       "location":"10.19.1.135"
     }
 
 If, however, you are running the controller on a work node on a cluster, you will likely
 need to use ssh tunnels to connect clients from your laptop to it.  You will also
 probably need to instruct the controller to listen for engines coming from other work nodes
 on the cluster.  An example of ipcontroller-client.json, as created by::
 
     $> ipcontroller --ip=* --ssh=login.mycluster.com
 
 
 .. sourcecode:: python
 
     {
       "url":"tcp:\/\/*:54424",
       "exec_key":"a361fe89-92fc-4762-9767-e2f0a05e3130",
       "ssh":"login.mycluster.com",
       "location":"10.0.0.2"
     }
 More details of how these JSON files are used are given below.
 
 A detailed description of the security model and its implementation in IPython
 can be found :ref:`here <parallelsecurity>`.
 
 .. warning::
 
     Even at its most secure, the Controller listens on ports on localhost, and
     every time you make a tunnel, you open a localhost port on the connecting
     machine that points to the Controller. If localhost on the Controller's
     machine, or the machine of any client or engine, is untrusted, then your
     Controller is insecure. There is no way around this with ZeroMQ.
 
 
 
 Getting Started
 ===============
 
 To use IPython for parallel computing, you need to start one instance of the
 controller and one or more instances of the engine. Initially, it is best to
 simply start a controller and engines on a single host using the
 :command:`ipcluster` command. To start a controller and 4 engines on your
 localhost, just do::
 
     $ ipcluster start -n 4
 
 More details about starting the IPython controller and engines can be found
 :ref:`here <parallel_process>`
 
 Once you have started the IPython controller and one or more engines, you
 are ready to use the engines to do something useful. To make sure
 everything is working correctly, try the following commands:
 
 .. sourcecode:: ipython
 
 	In [1]: from IPython.parallel import Client
 	
 	In [2]: c = Client()
 	
 	In [4]: c.ids
 	Out[4]: set([0, 1, 2, 3])
 	
 	In [5]: c[:].apply_sync(lambda : "Hello, World")
 	Out[5]: [ 'Hello, World', 'Hello, World', 'Hello, World', 'Hello, World' ]
 
 
 When a client is created with no arguments, the client tries to find the corresponding JSON file
 in the local `~/.ipython/profile_default/security` directory. Or if you specified a profile,
 you can use that with the Client.  This should cover most cases:
 
 .. sourcecode:: ipython
 
     In [2]: c = Client(profile='myprofile')
 
 If you have put the JSON file in a different location or it has a different name, create the
 client like this:
 
 .. sourcecode:: ipython
 
     In [2]: c = Client('/path/to/my/ipcontroller-client.json')
 
 Remember, a client needs to be able to see the Hub's ports to connect. So if they are on a
 different machine, you may need to use an ssh server to tunnel access to that machine,
 then you would connect to it with:
 
 .. sourcecode:: ipython
 
     In [2]: c = Client('/path/to/my/ipcontroller-client.json', sshserver='[e-mail]')
 
 Where 'myhub.example.com' is the url or IP address of the machine on
 which the Hub process is running (or another machine that has direct access to the Hub's ports).
 
 The SSH server may already be specified in ipcontroller-client.json, if the controller was
 instructed at its launch time.
 
 You are now ready to learn more about the :ref:`Direct
 <parallel_multiengine>` and :ref:`LoadBalanced <parallel_task>` interfaces to the
 controller.
 
 .. [ZeroMQ] ZeroMQ.  http://www.zeromq.org

docs/source/whatsnew/version0.11.rst

@@ -1,765 +1,765 b''
 =============
  0.11 Series
 =============
 
 Release 0.11
 ============
 
 IPython 0.11 is a *major* overhaul of IPython, two years in the making.  Most
 of the code base has been rewritten or at least reorganized, breaking backward
 compatibility with several APIs in previous versions.  It is the first major
 release in two years, and probably the most significant change to IPython since
 its inception.  We plan to have a relatively quick succession of releases, as
 people discover new bugs and regressions.  Once we iron out any significant
 bugs in this process and settle down the new APIs, this series will become
 IPython 1.0.  We encourage feedback now on the core APIs, which we hope to
 maintain stable during the 1.0 series.
 
 Since the internal APIs have changed so much, projects using IPython as a
 library (as opposed to end-users of the application) are the most likely to
 encounter regressions or changes that break their existing use patterns.  We
 will make every effort to provide updated versions of the APIs to facilitate
 the transition, and we encourage you to contact us on the `development mailing
 list`__ with questions and feedback.
 
 .. __: http://mail.scipy.org/mailman/listinfo/ipython-dev
 
 Chris Fonnesbeck recently wrote an `excellent post`__ that highlights some of
 our major new features, with examples and screenshots.  We encourage you to
 read it as it provides an illustrated, high-level overview complementing the
 detailed feature breakdown in this document.
 
 .. __: http://fonnesbeck.calepin.co/innovations-in-ipython.html
 
 A quick summary of the major changes (see below for details):
 
 * **Standalone Qt console**: a new rich console has been added to IPython,
   started with `ipython qtconsole`.  In this application we have tried to
   retain the feel of a terminal for fast and efficient workflows, while adding
   many features that a line-oriented terminal simply can not support, such as
   inline figures, full multiline editing with syntax highlighting, graphical
   tooltips for function calls and much more.  This development was sponsored by
   `Enthought Inc.`__. See :ref:`below <qtconsole_011>` for details.
 
 .. __: http://enthought.com
 
 * **High-level parallel computing with ZeroMQ**. Using the same architecture
   that our Qt console is based on, we have completely rewritten our high-level
   parallel computing machinery that in prior versions used the Twisted
   networking framework.  While this change will require users to update their
   codes, the improvements in performance, memory control and internal
   consistency across our codebase convinced us it was a price worth paying.  We
   have tried to explain how to best proceed with this update, and will be happy
   to answer questions that may arise.  A full tutorial describing these
   features `was presented at SciPy'11`__, more details :ref:`below
   <parallel_011>`.
 
 .. __: http://minrk.github.com/scipy-tutorial-2011
 
 * **New model for GUI/plotting support in the terminal**.  Now instead of the
   various `-Xthread` flags we had before, GUI support is provided without the
   use of any threads, by directly integrating GUI event loops with Python's
   `PyOS_InputHook` API.  A new command-line flag `--gui` controls GUI support,
   and it can also be enabled after IPython startup via the new `%gui` magic.
   This requires some changes if you want to execute GUI-using scripts inside
   IPython, see :ref:`the GUI support section <gui_support>` for more details.
 
 * **A two-process architecture.** The Qt console is the first use of a new
   model that splits IPython between a kernel process where code is executed and
   a client that handles user interaction.  We plan on also providing terminal
   and web-browser based clients using this infrastructure in future releases.
   This model allows multiple clients to interact with an IPython process
   through a :ref:`well-documented messaging protocol <messaging>` using the
   ZeroMQ networking library.
 
 * **Refactoring.** the entire codebase has been refactored, in order to make it
   more modular and easier to contribute to.  IPython has traditionally been a
   hard project to participate because the old codebase was very monolithic.  We
   hope this (ongoing) restructuring will make it easier for new developers to
   join us.
 
 * **Vim integration**. Vim can be configured to seamlessly control an IPython
   kernel, see the files in :file:`docs/examples/vim` for the full details.
   This work was done by Paul Ivanov, who prepared a nice `video
   demonstration`__ of the features it provides.
 
 .. __: http://pirsquared.org/blog/2011/07/28/vim-ipython/
   
 * **Integration into Microsoft Visual Studio**. Thanks to the work of the
   Microsoft `Python Tools for Visual Studio`__ team, this version of IPython
   has been integrated into Microsoft Visual Studio's Python tools open source
   plug-in.  `Details below`_
 
 .. __: http://pytools.codeplex.com
 .. _details below: ms_visual_studio_011_
 
 * **Improved unicode support**. We closed many bugs related to unicode input.
 
 * **Python 3**. IPython now runs on Python 3.x. See :ref:`python3_011` for
   details.
 
 * **New profile model**. Profiles are now directories that contain all relevant
   information for that session, and thus better isolate IPython use-cases.
 
 * **SQLite storage for history**. All history is now stored in a SQLite
   database, providing support for multiple simultaneous sessions that won't
   clobber each other as well as the ability to perform queries on all stored
   data.
 
 * **New configuration system**. All parts of IPython are now configured via a
   mechanism inspired by the Enthought Traits library.  Any configurable element
   can have its attributes set either via files that now use real Python syntax
   or from the command-line.
 
 * **Pasting of code with prompts**. IPython now intelligently strips out input
   prompts , be they plain Python ones (``>>>`` and ``...``) or IPython ones
   (``In [N]:`` and ``...:``).  More details :ref:`here <pasting_with_prompts>`.
   
 
 Authors and support
 -------------------
 
 Over 60 separate authors have contributed to this release, see :ref:`below
 <credits_011>` for a full list.  In particular, we want to highlight the
 extremely active participation of two new core team members: Evan Patterson
 implemented the Qt console, and Thomas Kluyver started with our Python 3 port
 and by now has made major contributions to just about every area of IPython.
 
 We are also grateful for the support we have received during this development
 cycle from several institutions:
 
 - `Enthought Inc`__ funded the development of our new Qt console, an effort that
   required developing major pieces of underlying infrastructure, which now
   power not only the Qt console but also our new parallel machinery.  We'd like
   to thank Eric Jones and Travis Oliphant for their support, as well as Ilan
   Schnell for his tireless work integrating and testing IPython in the
   `Enthought Python Distribution`_.
 
 .. __: http://enthought.com
 .. _Enthought Python Distribution: http://www.enthought.com/products/epd.php
 
 - Nipy/NIH: funding via the `NiPy project`__ (NIH grant 5R01MH081909-02) helped
   us jumpstart the development of this series by restructuring the entire
   codebase two years ago in a way that would make modular development and
   testing more approachable.  Without this initial groundwork, all the new
   features we have added would have been impossible to develop.
 
 .. __: http://nipy.org
 
 - Sage/NSF: funding via the grant `Sage: Unifying Mathematical Software for
   Scientists, Engineers, and Mathematicians`__ (NSF grant DMS-1015114)
   supported a meeting in spring 2011 of several of the core IPython developers
   where major progress was made integrating the last key pieces leading to this
   release.
 
 .. __: http://modular.math.washington.edu/grants/compmath09
 
 - Microsoft's team working on `Python Tools for Visual Studio`__ developed the 
   integraton of IPython into the Python plugin for Visual Studio 2010.
 
 .. __: http://pytools.codeplex.com
 
 - Google Summer of Code: in 2010, we had two students developing prototypes of
   the new machinery that is now maturing in this release: `Omar Zapata`_ and
   `Gerardo Gutiérrez`_.
 
 .. _Omar Zapata: http://ipythonzmq.blogspot.com/2010/08/ipython-zmq-status.html
 .. _Gerardo Gutiérrez: http://ipythonqt.blogspot.com/2010/04/ipython-qt-interface-gsoc-2010-proposal.html>
 
 
 Development summary: moving to Git and Github
 ---------------------------------------------
 
 In April 2010, after `one breakage too many with bzr`__, we decided to move our
 entire development process to Git and Github.com.  This has proven to be one of
 the best decisions in the project's history, as the combination of git and
 github have made us far, far more productive than we could be with our previous
 tools.  We first converted our bzr repo to a git one without losing history,
 and a few weeks later ported all open Launchpad bugs to github issues with
 their comments mostly intact (modulo some formatting changes).  This ensured a
 smooth transition where no development history or submitted bugs were lost.
 Feel free to use our little Launchpad to Github issues `porting script`_ if you
 need to make a similar transition.
 
 .. __: http://mail.scipy.org/pipermail/ipython-dev/2010-April/005944.html
 .. _porting script: https://gist.github.com/835577
 
 These simple statistics show how much work has been done on the new release, by
 comparing the current code to the last point it had in common with the 0.10
 series.  A huge diff and ~2200 commits make up this cycle::
 
     git diff $(git merge-base 0.10.2 HEAD)  | wc -l
     288019
 
     git log $(git merge-base 0.10.2 HEAD)..HEAD --oneline | wc -l
     2200
 
 Since our move to github, 511 issues were closed, 226 of which were pull
 requests and 285 regular issues (:ref:`a full list with links
 <issues_list_011>` is available for those interested in the details).  Github's
 pull requests are a fantastic mechanism for reviewing code and building a
 shared ownership of the project, and we are making enthusiastic use of it.
 
 .. Note::
 
    This undercounts the number of issues closed in this development cycle,
    since we only moved to github for issue tracking in May 2010, but we have no
    way of collecting statistics on the number of issues closed in the old
    Launchpad bug tracker prior to that.
 
    
 .. _qtconsole_011:
 
 Qt Console
 ----------
 
 IPython now ships with a Qt application that feels very much like a terminal,
 but is in fact a rich GUI that runs an IPython client but supports inline
 figures, saving sessions to PDF and HTML, multiline editing with syntax
 highlighting, graphical calltips and much more:
 
 .. figure:: ../_images/qtconsole.png
     :width: 400px
     :alt: IPython Qt console with embedded plots
     :align: center
     :target: ../_images/qtconsole.png
 
     The Qt console for IPython, using inline matplotlib plots.
 
 We hope that many projects will embed this widget, which we've kept
 deliberately very lightweight, into their own environments.  In the future we
 may also offer a slightly more featureful application (with menus and other GUI
 elements), but we remain committed to always shipping this easy to embed
 widget.
 
 See the :ref:`Qt console section <qtconsole>` of the docs for a detailed
 description of the console's features and use.
 
 
 .. _parallel_011:
 
 High-level parallel computing with ZeroMQ
 -----------------------------------------
 
 We have completely rewritten the Twisted-based code for high-level parallel
 computing to work atop our new ZeroMQ architecture.  While we realize this will
 break compatibility for a number of users, we hope to make the transition as
 easy as possible with our docs, and we are convinced the change is worth it.
 ZeroMQ provides us with much tighter control over memory, higher performance,
 and its communications are impervious to the Python Global Interpreter Lock
 because they take place in a system-level C++ thread.  The impact of the GIL in
 our previous code was something we could simply not work around, given that
 Twisted is itself a Python library.  So while Twisted is a very capable
 framework, we think ZeroMQ fits our needs much better and we hope you will find
 the change to be a significant improvement in the long run.
 
 Our manual contains :ref:`a full description of how to use IPython for parallel
 computing <parallel_overview>`, and the `tutorial`__ presented by Min
 Ragan-Kelley at the SciPy 2011 conference provides a hands-on complement to the
 reference docs.
 
 .. __: http://minrk.github.com/scipy-tutorial-2011
 
 
 Refactoring
 -----------
 
 As of this release, a signifiant portion of IPython has been refactored.  This
 refactoring is founded on a number of new abstractions.  The main new classes
 that implement these abstractions are:
 
 * :class:`IPython.utils.traitlets.HasTraits`.
 * :class:`IPython.config.configurable.Configurable`.
 * :class:`IPython.config.application.Application`.
 * :class:`IPython.config.loader.ConfigLoader`.
 * :class:`IPython.config.loader.Config`
 
 We are still in the process of writing developer focused documentation about
 these classes, but for now our :ref:`configuration documentation
 <config_overview>` contains a high level overview of the concepts that these
 classes express.
 
 The biggest user-visible change is likely the move to using the config system
 to determine the command-line arguments for IPython applications. The benefit
 of this is that *all* configurable values in IPython are exposed on the
 command-line, but the syntax for specifying values has changed. The gist is
 that assigning values is pure Python assignment.  Simple flags exist for
 commonly used options, these are always prefixed with '--'.
 
 The IPython command-line help has the details of all the options (via
 ``ipythyon --help``), but a simple example should clarify things; the ``pylab``
 flag can be used to start in pylab mode with the qt4 backend::
 
   ipython --pylab=qt
 
 which is equivalent to using the fully qualified form::
 
   ipython --TerminalIPythonApp.pylab=qt
 
 The long-form options can be listed via ``ipython --help-all``.
 
 
 ZeroMQ architecture
 -------------------
 
 There is a new GUI framework for IPython, based on a client-server model in
 which multiple clients can communicate with one IPython kernel, using the
 ZeroMQ messaging framework. There is already a Qt console client, which can
 be started by calling ``ipython qtconsole``. The protocol is :ref:`documented
 <messaging>`.
 
 The parallel computing framework has also been rewritten using ZMQ. The
 protocol is described :ref:`here <parallel_messages>`, and the code is in the
 new :mod:`IPython.parallel` module.
 
 .. _python3_011:
 
 Python 3 support
 ----------------
 
 A Python 3 version of IPython has been prepared. For the time being, this is
 maintained separately and updated from the main codebase. Its code can be found
 `here <https://github.com/ipython/ipython-py3k>`_. The parallel computing
 components are not perfect on Python3, but most functionality appears to be
 working.  As this work is evolving quickly, the best place to find updated
 information about it is our `Python 3 wiki page`__.
 
 .. __: http://wiki.ipython.org/index.php?title=Python_3
 
 
 Unicode
 -------
 
 Entering non-ascii characters in unicode literals (``u"€ø"``) now works
 properly on all platforms. However, entering these in byte/string literals
 (``"€ø"``) will not work as expected on Windows (or any platform where the
 terminal encoding is not UTF-8, as it typically is for Linux & Mac OS X). You
 can use escape sequences (``"\xe9\x82"``) to get bytes above 128, or use
 unicode literals and encode them. This is a limitation of Python 2 which we
 cannot easily work around.
 
 .. _ms_visual_studio_011:
 
 Integration with Microsoft Visual Studio
 ----------------------------------------
 
 IPython can be used as the interactive shell in the `Python plugin for
 Microsoft Visual Studio`__, as seen here:
 
 .. figure:: ../_images/ms_visual_studio.png
     :width: 500px
     :alt: IPython console embedded in Microsoft Visual Studio.
     :align: center
     :target: ../_images/ms_visual_studio.png
 
     IPython console embedded in Microsoft Visual Studio.
 
 The Microsoft team developing this currently has a release candidate out using
 IPython 0.11. We will continue to collaborate with them to ensure that as they
 approach their final release date, the integration with IPython remains smooth.
 We'd like to thank Dino Viehland and Shahrokh Mortazavi for the work they have
 done towards this feature, as well as Wenming Ye for his support of our WinHPC
 capabilities.
 
 .. __: http://pytools.codeplex.com
 
 
 Additional new features
 -----------------------
 
 * Added ``Bytes`` traitlet, removing ``Str``.  All 'string' traitlets should
   either be ``Unicode`` if a real string, or ``Bytes`` if a C-string. This
   removes ambiguity and helps the Python 3 transition.
 
 * New magic ``%loadpy`` loads a python file from disk or web URL into
   the current input buffer.
 
 * New magic ``%pastebin`` for sharing code via the 'Lodge it' pastebin.
 
 * New magic ``%precision`` for controlling float and numpy pretty printing.
 
 * IPython applications initiate logging, so any object can gain access to
   a the logger of the currently running Application with:
 
 .. sourcecode:: python
 
     from IPython.config.application import Application
     logger = Application.instance().log
 
 * You can now get help on an object halfway through typing a command. For
   instance, typing ``a = zip?`` shows the details of :func:`zip`. It also
   leaves the command at the next prompt so you can carry on with it.
 
 * The input history is now written to an SQLite database. The API for
   retrieving items from the history has also been redesigned.
 
 * The :mod:`IPython.extensions.pretty` extension has been moved out of
   quarantine and fully updated to the new extension API.
 
 * New magics for loading/unloading/reloading extensions have been added:
   ``%load_ext``, ``%unload_ext`` and ``%reload_ext``.
 
 * The configuration system and configuration files are brand new. See the
   configuration system :ref:`documentation <config_index>` for more details.
 
 * The :class:`~IPython.core.interactiveshell.InteractiveShell` class is now a
   :class:`~IPython.config.configurable.Configurable` subclass and has traitlets
   that determine the defaults and runtime environment. The ``__init__`` method
   has also been refactored so this class can be instantiated and run without
   the old :mod:`ipmaker` module.
 
 * The methods of :class:`~IPython.core.interactiveshell.InteractiveShell` have
   been organized into sections to make it easier to turn more sections
   of functionality into components.
 
 * The embedded shell has been refactored into a truly standalone subclass of
   :class:`InteractiveShell` called :class:`InteractiveShellEmbed`.  All
   embedding logic has been taken out of the base class and put into the 
   embedded subclass.
 
 * Added methods of :class:`~IPython.core.interactiveshell.InteractiveShell` to
   help it cleanup after itself. The :meth:`cleanup` method controls this. We
   couldn't do this in :meth:`__del__` because we have cycles in our object
   graph that prevent it from being called.
 
 * Created a new module :mod:`IPython.utils.importstring` for resolving
   strings like ``foo.bar.Bar`` to the actual class.
 
 * Completely refactored the :mod:`IPython.core.prefilter` module into
   :class:`~IPython.config.configurable.Configurable` subclasses. Added a new
   layer into the prefilter system, called "transformations" that all new
   prefilter logic should use (rather than the older "checker/handler"
   approach).
 
 * Aliases are now components (:mod:`IPython.core.alias`).
 
 * New top level :func:`~IPython.frontend.terminal.embed.embed` function that can
   be called to embed IPython at any place in user's code. On the first call it
   will create an :class:`~IPython.frontend.terminal.embed.InteractiveShellEmbed`
   instance and call it. In later calls, it just calls the previously created
   :class:`~IPython.frontend.terminal.embed.InteractiveShellEmbed`.
 
 * Created a configuration system (:mod:`IPython.config.configurable`) that is
   based on :mod:`IPython.utils.traitlets`. Configurables are arranged into a
   runtime containment tree (not inheritance) that i) automatically propagates
   configuration information and ii) allows singletons to discover each other in
   a loosely coupled manner. In the future all parts of IPython will be
   subclasses of :class:`~IPython.config.configurable.Configurable`. All IPython
   developers should become familiar with the config system.
 
 * Created a new :class:`~IPython.config.loader.Config` for holding
   configuration information. This is a dict like class with a few extras: i)
   it supports attribute style access, ii) it has a merge function that merges
   two :class:`~IPython.config.loader.Config` instances recursively and iii) it
   will automatically create sub-:class:`~IPython.config.loader.Config`
   instances for attributes that start with an uppercase character.
 
 * Created new configuration loaders in :mod:`IPython.config.loader`. These
   loaders provide a unified loading interface for all configuration
   information including command line arguments and configuration files. We
   have two default implementations based on :mod:`argparse` and plain python
   files.  These are used to implement the new configuration system.
 
 * Created a top-level :class:`Application` class in
   :mod:`IPython.core.application` that is designed to encapsulate the starting
   of any basic Python program. An application loads and merges all the
   configuration objects, constructs the main application, configures and
   initiates logging, and creates and configures any :class:`Configurable`
   instances and then starts the application running. An extended
   :class:`BaseIPythonApplication` class adds logic for handling the
   IPython directory as well as profiles, and all IPython entry points
   extend it.
 
 * The :class:`Type` and :class:`Instance` traitlets now handle classes given
   as strings, like ``foo.bar.Bar``. This is needed for forward declarations.
   But, this was implemented in a careful way so that string to class
   resolution is done at a single point, when the parent
   :class:`~IPython.utils.traitlets.HasTraitlets` is instantiated.
 
 * :mod:`IPython.utils.ipstruct` has been refactored to be a subclass of 
   dict.  It also now has full docstrings and doctests.
 
 * Created a Traits like implementation in :mod:`IPython.utils.traitlets`.  This
   is a pure Python, lightweight version of a library that is similar to
   Enthought's Traits project, but has no dependencies on Enthought's code.  We
   are using this for validation, defaults and notification in our new component
   system.  Although it is not 100% API compatible with Enthought's Traits, we
   plan on moving in this direction so that eventually our implementation could
   be replaced by a (yet to exist) pure Python version of Enthought Traits.
 
 * Added a new module :mod:`IPython.lib.inputhook` to manage the integration
   with GUI event loops using `PyOS_InputHook`.  See the docstrings in this
   module or the main IPython docs for details.
 
 * For users, GUI event loop integration is now handled through the new
   :command:`%gui` magic command.  Type ``%gui?`` at an IPython prompt for
   documentation.
 
 * For developers :mod:`IPython.lib.inputhook` provides a simple interface
   for managing the event loops in their interactive GUI applications.
-  Examples can be found in our :file:`docs/examples/lib` directory.
+  Examples can be found in our :file:`examples/lib` directory.
 
 Backwards incompatible changes
 ------------------------------
 
 * The Twisted-based :mod:`IPython.kernel` has been removed, and completely
   rewritten as :mod:`IPython.parallel`, using ZeroMQ.
 
 * Profiles are now directories. Instead of a profile being a single config file,
   profiles are now self-contained directories. By default, profiles get their
   own IPython history, log files, and everything. To create a new profile, do
   ``ipython profile create <name>``.
 
 * All IPython applications have been rewritten to use
   :class:`~IPython.config.loader.KeyValueConfigLoader`. This means that
   command-line options have changed. Now, all configurable values are accessible
   from the command-line with the same syntax as in a configuration file.
 
 * The command line options ``-wthread``, ``-qthread`` and
   ``-gthread`` have been removed. Use ``--gui=wx``, ``--gui=qt``, ``--gui=gtk``
   instead.
 
 * The extension loading functions have been renamed to
   :func:`load_ipython_extension` and :func:`unload_ipython_extension`.
 
 * :class:`~IPython.core.interactiveshell.InteractiveShell` no longer takes an
   ``embedded`` argument. Instead just use the
   :class:`~IPython.core.interactiveshell.InteractiveShellEmbed` class.
 
 * ``__IPYTHON__`` is no longer injected into ``__builtin__``.
 
 * :meth:`Struct.__init__` no longer takes `None` as its first argument.  It
   must be a :class:`dict` or :class:`Struct`.
 
 * :meth:`~IPython.core.interactiveshell.InteractiveShell.ipmagic` has been
   renamed :meth:`~IPython.core.interactiveshell.InteractiveShell.magic.`
 
 * The functions :func:`ipmagic` and :func:`ipalias` have been removed from
   :mod:`__builtins__`.
 
 * The references to the global
   :class:`~IPython.core.interactivehell.InteractiveShell` instance (``_ip``, and
   ``__IP``) have been removed from the user's namespace. They are replaced by a
   new function called :func:`get_ipython` that returns the current
   :class:`~IPython.core.interactiveshell.InteractiveShell` instance. This
   function is injected into the user's namespace and is now the main way of
   accessing the running IPython.
 
 * Old style configuration files :file:`ipythonrc` and :file:`ipy_user_conf.py`
   are no longer supported. Users should migrate there configuration files to
   the new format described :ref:`here <config_overview>` and :ref:`here
   <configuring_ipython>`.
 
 * The old IPython extension API that relied on :func:`ipapi` has been
   completely removed. The new extension API is described :ref:`here
   <configuring_ipython>`.
 
 * Support for ``qt3`` has been dropped.  Users who need this should use
   previous versions of IPython.
 
 * Removed :mod:`shellglobals` as it was obsolete.
 
 * Removed all the threaded shells in :mod:`IPython.core.shell`. These are no
   longer needed because of the new capabilities in
   :mod:`IPython.lib.inputhook`.
 
 * New top-level sub-packages have been created: :mod:`IPython.core`, 
   :mod:`IPython.lib`, :mod:`IPython.utils`, :mod:`IPython.deathrow`,
   :mod:`IPython.quarantine`.  All existing top-level modules have been
   moved to appropriate sub-packages.  All internal import statements
   have been updated and tests have been added.  The build system (setup.py
   and friends) have been updated.  See :ref:`this section <module_reorg>` of the
   documentation for descriptions of these new sub-packages.
 
 * :mod:`IPython.ipapi` has been moved to :mod:`IPython.core.ipapi`.
   :mod:`IPython.Shell` and :mod:`IPython.iplib` have been split and removed as
   part of the refactor.
 
 * :mod:`Extensions` has been moved to :mod:`extensions` and all existing
   extensions have been moved to either :mod:`IPython.quarantine` or
   :mod:`IPython.deathrow`. :mod:`IPython.quarantine` contains modules that we
   plan on keeping but that need to be updated. :mod:`IPython.deathrow` contains
   modules that are either dead or that should be maintained as third party
   libraries. More details about this can be found :ref:`here <module_reorg>`.
 
 * Previous IPython GUIs in :mod:`IPython.frontend` and :mod:`IPython.gui` are
   likely broken, and have been removed to :mod:`IPython.deathrow` because of the
   refactoring in the core. With proper updates, these should still work.
 
 
 Known Regressions
 -----------------
 
 We do our best to improve IPython, but there are some known regressions in 0.11
 relative to 0.10.2.  First of all, there are features that have yet to be
 ported to the new APIs, and in order to ensure that all of the installed code
 runs for our users, we have moved them to two separate directories in the
 source distribution, `quarantine` and `deathrow`.  Finally, we have some other
 miscellaneous regressions that we hope to fix as soon as possible.  We now
 describe all of these in more detail.
 
 Quarantine
 ~~~~~~~~~~
 
 These are tools and extensions that we consider relatively easy to update to
 the new classes and APIs, but that we simply haven't had time for.  Any user
 who is interested in one of these is encouraged to help us by porting it and
 submitting a pull request on our `development site`_.
 
 .. _development site: http://github.com/ipython/ipython
 
 Currently, the quarantine directory contains::
 
     clearcmd.py            ipy_fsops.py            ipy_signals.py
     envpersist.py          ipy_gnuglobal.py        ipy_synchronize_with.py
     ext_rescapture.py      ipy_greedycompleter.py  ipy_system_conf.py
     InterpreterExec.py     ipy_jot.py              ipy_which.py
     ipy_app_completers.py  ipy_lookfor.py          ipy_winpdb.py
     ipy_autoreload.py      ipy_profile_doctest.py  ipy_workdir.py
     ipy_completers.py      ipy_pydb.py             jobctrl.py
     ipy_editors.py         ipy_rehashdir.py        ledit.py
     ipy_exportdb.py        ipy_render.py           pspersistence.py
     ipy_extutil.py         ipy_server.py           win32clip.py
 
 Deathrow
 ~~~~~~~~
 
 These packages may be harder to update or make most sense as third-party
 libraries.  Some of them are completely obsolete and have been already replaced
 by better functionality (we simply haven't had the time to carefully weed them
 out so they are kept here for now). Others simply require fixes to code that
 the current core team may not be familiar with.  If a tool you were used to is
 included here, we encourage you to contact the dev list and we can discuss
 whether it makes sense to keep it in IPython (if it can be maintained).
 
 Currently, the deathrow directory contains::
 
     astyle.py              ipy_defaults.py          ipy_vimserver.py
     dtutils.py             ipy_kitcfg.py            numeric_formats.py
     Gnuplot2.py            ipy_legacy.py            numutils.py
     GnuplotInteractive.py  ipy_p4.py                outputtrap.py
     GnuplotRuntime.py      ipy_profile_none.py      PhysicalQInput.py
     ibrowse.py             ipy_profile_numpy.py     PhysicalQInteractive.py
     igrid.py               ipy_profile_scipy.py     quitter.py*
     ipipe.py               ipy_profile_sh.py        scitedirector.py
     iplib.py               ipy_profile_zope.py      Shell.py
     ipy_constants.py       ipy_traits_completer.py  twshell.py
 
 
 Other regressions
 ~~~~~~~~~~~~~~~~~
 
 * The machinery that adds functionality to the 'sh' profile for using IPython
   as your system shell has not been updated to use the new APIs.  As a result,
   only the aesthetic (prompt) changes are still implemented. We intend to fix
   this by 0.12.  Tracked as issue 547_.
 
 .. _547: https://github.com/ipython/ipython/issues/547
 
 * The installation of scripts on Windows was broken without setuptools, so we
   now depend on setuptools on Windows.  We hope to fix setuptools-less
   installation, and then remove the setuptools dependency.  Issue 539_.
 
 .. _539: https://github.com/ipython/ipython/issues/539
 
 * The directory history `_dh` is not saved between sessions.  Issue 634_.
 
 .. _634: https://github.com/ipython/ipython/issues/634
 
 
 Removed Features
 ----------------
 
 As part of the updating of IPython, we have removed a few features for the
 purposes of cleaning up the codebase and interfaces.  These removals are
 permanent, but for any item listed below, equivalent functionality is
 available.
 
 * The magics Exit and Quit have been dropped as ways to exit IPython. Instead,
   the lowercase forms of both work either as a bare name (``exit``) or a
   function call (``exit()``).  You can assign these to other names using
   exec_lines in the config file.
 
 
 .. _credits_011:
 
 Credits
 -------
 
 Many users and developers contributed code, features, bug reports and ideas to
 this release.  Please do not hesitate in contacting us if we've failed to
 acknowledge your contribution here.  In particular, for this release we have
 contribution from the following people, a mix of new and regular names (in
 alphabetical order by first name):
 
 * Aenugu Sai Kiran Reddy <saikrn08-at-gmail.com>
 * andy wilson <wilson.andrew.j+github-at-gmail.com>
 * Antonio Cuni <antocuni>
 * Barry Wark <barrywark-at-gmail.com>
 * Beetoju Anuradha <anu.beethoju-at-gmail.com>
 * Benjamin Ragan-Kelley <minrk-at-Mercury.local>
 * Brad Reisfeld
 * Brian E. Granger <ellisonbg-at-gmail.com>
 * Christoph Gohlke <cgohlke-at-uci.edu>
 * Cody Precord
 * dan.milstein
 * Darren Dale <dsdale24-at-gmail.com>
 * Dav Clark <davclark-at-berkeley.edu>
 * David Warde-Farley <wardefar-at-iro.umontreal.ca>
 * epatters <ejpatters-at-gmail.com>
 * epatters <epatters-at-caltech.edu>
 * epatters <epatters-at-enthought.com>
 * Eric Firing <efiring-at-hawaii.edu>
 * Erik Tollerud <erik.tollerud-at-gmail.com>
 * Evan Patterson <epatters-at-enthought.com>
 * Fernando Perez <Fernando.Perez-at-berkeley.edu>
 * Gael Varoquaux <gael.varoquaux-at-normalesup.org>
 * Gerardo <muzgash-at-Muzpelheim>
 * Jason Grout <jason.grout-at-drake.edu>
 * John Hunter <jdh2358-at-gmail.com>
 * Jens Hedegaard Nielsen <jenshnielsen-at-gmail.com>
 * Johann Cohen-Tanugi <johann.cohentanugi-at-gmail.com>
 * Jörgen Stenarson <jorgen.stenarson-at-bostream.nu>
 * Justin Riley <justin.t.riley-at-gmail.com>
 * Kiorky
 * Laurent Dufrechou <laurent.dufrechou-at-gmail.com>
 * Luis Pedro Coelho <lpc-at-cmu.edu>
 * Mani chandra <mchandra-at-iitk.ac.in>
 * Mark E. Smith
 * Mark Voorhies <mark.voorhies-at-ucsf.edu>
 * Martin Spacek <git-at-mspacek.mm.st>
 * Michael Droettboom <mdroe-at-stsci.edu>
 * MinRK <benjaminrk-at-gmail.com>
 * muzuiget <muzuiget-at-gmail.com>
 * Nick Tarleton <nick-at-quixey.com>
 * Nicolas Rougier <Nicolas.rougier-at-inria.fr>
 * Omar Andres Zapata Mesa <andresete.chaos-at-gmail.com>
 * Paul Ivanov <pivanov314-at-gmail.com>
 * Pauli Virtanen <pauli.virtanen-at-iki.fi>
 * Prabhu Ramachandran
 * Ramana <sramana9-at-gmail.com>
 * Robert Kern <robert.kern-at-gmail.com>
 * Sathesh Chandra <satheshchandra88-at-gmail.com>
 * Satrajit Ghosh <satra-at-mit.edu>
 * Sebastian Busch
 * Skipper Seabold <jsseabold-at-gmail.com>
 * Stefan van der Walt <bzr-at-mentat.za.net>
 * Stephan Peijnik <debian-at-sp.or.at>
 * Steven Bethard
 * Thomas Kluyver <takowl-at-gmail.com>
 * Thomas Spura <tomspur-at-fedoraproject.org>
 * Tom Fetherston <tfetherston-at-aol.com>
 * Tom MacWright
 * tzanko
 * vankayala sowjanya <hai.sowjanya-at-gmail.com>
 * Vivian De Smedt <vds2212-at-VIVIAN>
 * Ville M. Vainio <vivainio-at-gmail.com>
 * Vishal Vatsa <vishal.vatsa-at-gmail.com>
 * Vishnu S G <sgvishnu777-at-gmail.com>
 * Walter Doerwald <walter-at-livinglogic.de>
 
 .. note::
 
     This list was generated with the output of
     ``git log dev-0.11 HEAD --format='* %aN <%aE>' | sed 's/@/\-at\-/' | sed 's/<>//' | sort -u``
     after some cleanup.  If you should be on this list, please add yourself.

docs/source/whatsnew/version0.13.rst

@@ -1,673 +1,673 b''
 =============
  0.13 Series
 =============
 
 Release 0.13
 ============
 
 IPython 0.13 contains several major new features, as well as a large amount of
 bug and regression fixes.  The previous version (0.12) was released on December
 19 2011, and in this development cycle we had:
 
 - ~6 months of work.
 - 373 pull requests merged.
 - 742 issues closed (non-pull requests).
 - contributions from 62 authors.
 - 1760 commits.
 - a diff of 114226 lines.
 
 The amount of work included in this release is so large, that we can only cover
 here the main highlights; please see our :ref:`detailed release statistics
 <issues_list_013>` for links to every issue and pull request closed on GitHub
 as well as a full list of individual contributors.
 
 
 Major Notebook improvements: new user interface and more
 --------------------------------------------------------
 
 The IPython Notebook, which has proven since its release to be wildly popular,
 has seen a massive amount of work in this release cycle, leading to a
 significantly improved user experience as well as many new features.
 
 The first user-visible change is a reorganization of the user interface; the
 left panel has been removed and was replaced by a real menu system and a
 toolbar with icons.  Both the toolbar and the header above the menu can be
 collapsed to leave an unobstructed working area:
 
 .. image:: ../_images/ipy_013_notebook_spectrogram.png
     :width: 460px
     :alt: New user interface for Notebook
     :align: center
     :target: ../_images/ipy_013_notebook_spectrogram.png
 
 The notebook handles very long outputs much better than before (this was a
 serious usability issue when running processes that generated massive amounts
 of output).  Now, in the presence of outputs longer than ~100 lines, the
 notebook will automatically collapse to a scrollable area and the entire left
 part of this area controls the display: one click in this area will expand the
 output region completely, and a double-click will hide it completely.  This
 figure shows both the scrolled and hidden modes:
 
 .. image:: ../_images/ipy_013_notebook_long_out.png
     :width: 460px
     :alt: Scrolling and hiding of long output in the notebook.
     :align: center
     :target: ../_images/ipy_013_notebook_long_out.png
 
 .. note::
 
    The auto-folding of long outputs is disabled in Firefox due to bugs in its
    scrolling behavior.  See :ghpull:`2047` for details.
 
 Uploading notebooks to the dashboard is now easier: in addition to drag and
 drop (which can be finicky sometimes), you can now click on the upload text and
 use a regular file dialog box to select notebooks to upload. Furthermore, the
 notebook dashboard now auto-refreshes its contents and offers buttons to shut
 down any running kernels (:ghpull:`1739`):
 
 .. image:: ../_images/ipy_013_dashboard.png
     :width: 460px
     :alt: Improved dashboard
     :align: center
     :target: ../_images/ipy_013_dashboard.png
 
 
 Cluster management
 ~~~~~~~~~~~~~~~~~~
 
 The notebook dashboard can now also start and stop clusters, thanks to a new
 tab in the dashboard user interface:
 
 .. image:: ../_images/ipy_013_dashboard_cluster.png
     :width: 460px
     :alt: Cluster management from the notebook dashboard
     :align: center
     :target: ../_images/ipy_013_dashboard_cluster.png
 
 This interface allows, for each profile you have configured, to start and stop
 a cluster (and optionally override the default number of engines corresponding
 to that configuration).  While this hides all error reporting, once you have a
 configuration that you know works smoothly, it is a very convenient interface
 for controlling your parallel resources.
 
 
 New notebook format
 ~~~~~~~~~~~~~~~~~~~
 
 The notebooks saved now use version 3 of our format, which supports heading
 levels as well as the concept of 'raw' text cells that are not rendered as
 Markdown.  These will be useful with converters_ we are developing, to pass raw
 markup (say LaTeX).  That conversion code is still under heavy development and
 not quite ready for prime time, but we welcome help on this front so that we
 can merge it for full production use as soon as possible.
 
 .. _converters: https://github.com/ipython/nbconvert
 
 .. note::
 
    v3 notebooks can *not* be read by older versions of IPython, but we provide
    a `simple script`_ that you can use in case you need to export a v3
    notebook to share with a v2 user.
 
 .. _simple script: https://gist.github.com/1935808
 
 
 JavaScript refactoring
 ~~~~~~~~~~~~~~~~~~~~~~
   
 All the client-side JavaScript has been decoupled to ease reuse of parts of the
 machinery without having to build a full-blown notebook. This will make it much
 easier to communicate with an IPython kernel from existing web pages and to
 integrate single cells into other sites, without loading the full notebook
 document-like UI. :ghpull:`1711`.
     
 This refactoring also enables the possibility of writing dynamic javascript
 widgets that are returned from Python code and that present an interactive view
 to the user, with callbacks in Javascript executing calls to the Kernel.  This
 will enable many interactive elements to be added by users in notebooks.
 
 An example of this capability has been provided as a proof of concept in
-:file:`docs/examples/widgets` that lets you directly communicate with one or more
+:file:`examples/widgets` that lets you directly communicate with one or more
 parallel engines, acting as a mini-console for parallel debugging and
 introspection.
 
     
 Improved tooltips
 ~~~~~~~~~~~~~~~~~
 
 The object tooltips have gained some new functionality. By pressing tab several
 times, you can expand them to see more of a docstring, keep them visible as you
 fill in a function's parameters, or transfer the information to the pager at the
 bottom of the screen. For the details, look at the example notebook
 :file:`01_notebook_introduction.ipynb`.
 
 .. figure:: ../_images/ipy_013_notebook_tooltip.png
     :width: 460px
     :alt: Improved tooltips in the notebook.
     :align: center
     :target: ../_images/ipy_013_notebook_tooltip.png
 
     The new notebook tooltips.
 
 Other improvements to the Notebook
 ----------------------------------
 
 These are some other notable small improvements to the notebook, in addition to
 many bug fixes and minor changes to add polish and robustness throughout:
 
 * The notebook pager (the area at the bottom) is now resizeable by dragging its
   divider handle, a feature that had been requested many times by just about
   anyone who had used the notebook system.  :ghpull:`1705`.
 
 * It is now possible to open notebooks directly from the command line; for
   example: ``ipython notebook path/`` will automatically set ``path/`` as the
   notebook directory, and ``ipython notebook path/foo.ipynb`` will further
   start with the ``foo.ipynb`` notebook opened.  :ghpull:`1686`.
   
 * If a notebook directory is specified with ``--notebook-dir`` (or with the
   corresponding configuration flag ``NotebookManager.notebook_dir``), all
   kernels start in this directory.
 
 * Fix codemirror clearing of cells with ``Ctrl-Z``; :ghpull:`1965`.
   
 * Text (markdown) cells now line wrap correctly in the notebook, making them
   much easier to edit :ghpull:`1330`.
   
 * PNG and JPEG figures returned from plots can be interactively resized in the
   notebook, by dragging them from their lower left corner. :ghpull:`1832`.
 
 * Clear ``In []`` prompt numbers on "Clear All Output".  For more
   version-control-friendly ``.ipynb`` files, we now strip all prompt numbers
   when doing a "Clear all output".  This reduces the amount of noise in
   commit-to-commit diffs that would otherwise show the (highly variable) prompt
   number changes. :ghpull:`1621`.
 
 * The notebook server now requires *two* consecutive ``Ctrl-C`` within 5
   seconds (or an interactive confirmation) to terminate operation.  This makes
   it less likely that you will accidentally kill a long-running server by
   typing ``Ctrl-C`` in the wrong terminal.  :ghpull:`1609`.
 
 * Using ``Ctrl-S`` (or ``Cmd-S`` on a Mac) actually saves the notebook rather
   than providing the fairly useless browser html save dialog.  :ghpull:`1334`.
   
 * Allow accessing local files from the notebook (in urls), by serving any local
   file as the url ``files/<relativepath>``.  This makes it possible to, for
   example, embed local images in a notebook.  :ghpull:`1211`.
 
       
 Cell magics
 -----------
 
 We have completely refactored the magic system, finally moving the magic
 objects to standalone, independent objects instead of being the mixin class
 we'd had since the beginning of IPython (:ghpull:`1732`).  Now, a separate base
 class is provided in :class:`IPython.core.magic.Magics` that users can subclass
 to create their own magics.  Decorators are also provided to create magics from
 simple functions without the need for object orientation.  Please see the
 :ref:`magic` docs for further details.
 
 All builtin magics now exist in a few subclasses that group together related
 functionality, and the new :mod:`IPython.core.magics` package has been created
 to organize this into smaller files.
     
 This cleanup was the last major piece of deep refactoring needed from the
 original 2001 codebase.
     
 We have also introduced a new type of magic function, prefixed with `%%`
 instead of `%`, which operates at the whole-cell level.  A cell magic receives
 two arguments: the line it is called on (like a line magic) and the body of the
 cell below it.
     
 Cell magics are most natural in the notebook, but they also work in the
 terminal and qt console, with the usual approach of using a blank line to
 signal cell termination.
     
 For example, to time the execution of several statements::
 
     %%timeit x = 0   # setup
     for i in range(100000):
         x += i**2
 
 This is particularly useful to integrate code in another language, and cell
 magics already exist for shell scripts, Cython, R and Octave. Using ``%%script
 /usr/bin/foo``, you can run a cell in any interpreter that accepts code via
 stdin.
 
 Another handy cell magic makes it easy to write short text files: ``%%file
 ~/save/to/here.txt``.
 
 The following cell magics are now included by default; all those that use
 special interpreters (Perl, Ruby, bash, etc.) assume you have the requisite
 interpreter installed:
 
 * ``%%!``: run cell body with the underlying OS shell; this is similar to
   prefixing every line in the cell with ``!``.
   
 * ``%%bash``: run cell body under bash.
   
 * ``%%capture``: capture the output of the code in the cell (and stderr as
   well).  Useful to run codes that produce too much output that you don't even
   want scrolled.
   
 * ``%%file``: save cell body as a file.
   
 * ``%%perl``: run cell body using Perl.
   
 * ``%%prun``: run cell body with profiler (cell extension of ``%prun``).
   
 * ``%%python3``: run cell body using Python 3.
   
 * ``%%ruby``: run cell body using Ruby.
   
 * ``%%script``: run cell body with the script specified in the first line.
   
 * ``%%sh``: run cell body using sh.
   
 * ``%%sx``: run cell with system shell and capture process output (cell
   extension of ``%sx``).
   
 * ``%%system``: run cell with system shell (``%%!`` is an alias to this).
   
 * ``%%timeit``: time the execution of the cell (extension of ``%timeit``).
 
 This is what some of the script-related magics look like in action:
 
 .. image:: ../_images/ipy_013_notebook_script_cells.png
     :width: 460px
     :alt: Cluster management from the notebook dashboard
     :align: center
     :target: ../_images/ipy_013_notebook_script_cells.png
   
 In addition, we have also a number of :ref:`extensions <extensions_overview>`
 that provide specialized magics.  These typically require additional software
 to run and must be manually loaded via ``%load_ext <extension name>``, but are
 extremely useful.  The following extensions are provided:
 
 **Cython magics** (extension :ref:`cythonmagic <extensions_cythonmagic>`)
     This extension provides magics to automatically build and compile Python
     extension modules using the Cython_ language. You must install Cython
     separately, as well as a C compiler, for this to work.  The examples
     directory in the source distribution ships with a full notebook
     demonstrating these capabilities:
 
 .. image:: ../_images/ipy_013_notebook_cythonmagic.png
     :width: 460px
     :alt: Cython magic
     :align: center
     :target: ../_images/ipy_013_notebook_cythonmagic.png
 
 .. _cython: http://cython.org
 
 **Octave magics** (extension :ref:`octavemagic <extensions_octavemagic>`)
     This extension provides several magics that support calling code written in
     the Octave_ language for numerical computing.  You can execute single-lines
     or whole blocks of Octave code, capture both output and figures inline
     (just like matplotlib plots), and have variables automatically converted
     between the two languages.  To use this extension, you must have Octave
     installed as well as the oct2py_ package.  The examples
     directory in the source distribution ships with a full notebook
     demonstrating these capabilities:
 
 .. image:: ../_images/ipy_013_notebook_octavemagic.png
     :width: 460px
     :alt: Octave magic
     :align: center
     :target: ../_images/ipy_013_notebook_octavemagic.png
 
 .. _octave: http://www.gnu.org/software/octave
 .. _oct2py: http://pypi.python.org/pypi/oct2py
 
 **R magics** (extension :ref:`rmagic <extensions_rmagic>`)
     This extension provides several magics that support calling code written in
     the R_ language for statistical data analysis.  You can execute
     single-lines or whole blocks of R code, capture both output and figures
     inline (just like matplotlib plots), and have variables automatically
     converted between the two languages.  To use this extension, you must have
     R installed as well as the rpy2_ package that bridges Python and R.  The
     examples directory in the source distribution ships with a full notebook
     demonstrating these capabilities:
 
 .. image:: ../_images/ipy_013_notebook_rmagic.png
     :width: 460px
     :alt: R magic
     :align: center
     :target: ../_images/ipy_013_notebook_rmagic.png
 
 .. _R: http://www.r-project.org
 .. _rpy2: http://rpy.sourceforge.net/rpy2.html
 
 
 Tab completer improvements
 --------------------------
 
 Useful tab-completion based on live inspection of objects is one of the most
 popular features of IPython. To make this process even more user-friendly, the
 completers of both the Qt console and the Notebook have been reworked.
 
 The Qt console comes with a new ncurses-like tab completer, activated by
 default, which lets you cycle through the available completions by pressing tab,
 or select a completion with the arrow keys (:ghpull:`1851`).
 
 .. figure:: ../_images/ipy_013_qtconsole_completer.png
     :width: 460px
     :alt: ncurses-like completer, with highlighted selection.
     :align: center
     :target: ../_images/ipy_013_qtconsole_completer.png
 
     The new improved Qt console's ncurses-like completer allows to easily
     navigate thought long list of completions.
 
 In the notebook, completions are now sourced both from object introspection and
 analysis of surrounding code, so limited completions can be offered for
 variables defined in the current cell, or while the kernel is busy 
 (:ghpull:`1711`).
 
 
 We have implemented a new configurable flag to control tab completion on
 modules that provide the ``__all__`` attribute::
 
   IPCompleter.limit_to__all__= Boolean
 
 This instructs the completer to honor ``__all__`` for the completion.
 Specifically, when completing on ``object.<tab>``, if True: only those names
 in ``obj.__all__`` will be included.  When False [default]: the ``__all__``
 attribute is ignored. :ghpull:`1529`.
 
 
 Improvements to the Qt console
 ------------------------------
 
 The Qt console continues to receive improvements and refinements, despite the
 fact that it is by now a fairly mature and robust component.  Lots of small
 polish has gone into it, here are a few highlights:
 
 * A number of changes were made to the underlying code for easier integration
   into other projects such as Spyder_ (:ghpull:`2007`, :ghpull:`2024`).
 
 * Improved menus with a new Magic menu that is organized by magic groups (this
   was made possible by the reorganization of the magic system
   internals). :ghpull:`1782`.
 
 * Allow for restarting kernels without clearing the qtconsole, while leaving a
   visible indication that the kernel has restarted. :ghpull:`1681`.
   
 * Allow the native display of jpeg images in the qtconsole. :ghpull:`1643`.
 
 .. _spyder: https://code.google.com/p/spyderlib
 
 
   
 Parallel
 --------
 
 The parallel tools have been improved and fine-tuned on multiple fronts.  Now,
 the creation of an :class:`IPython.parallel.Client` object automatically
 activates a line and cell magic function ``px`` that sends its code to all the
 engines. Further magics can be easily created with the :meth:`.Client.activate`
 method, to conveniently execute code on any subset of engines. :ghpull:`1893`.
 
 The ``%%px`` cell magic can also be given an optional targets argument, as well
 as a ``--out`` argument for storing its output.
 
 A new magic has also been added, ``%pxconfig``, that lets you configure various
 defaults of the parallel magics.  As usual, type  ``%pxconfig?`` for details.
 
 The exception reporting in parallel contexts has been improved to be easier to
 read.  Now, IPython directly reports the remote exceptions without showing any
 of the internal execution parts:
 
 .. image::  ../_images/ipy_013_par_tb.png
     :width: 460px
     :alt: Improved parallel exceptions.
     :align: center
     :target: ../_images/ipy_013_par_tb.png
 
 The parallel tools now default to using ``NoDB`` as the storage backend for
 intermediate results.  This means that the default usage case will have a
 significantly reduced memory footprint, though certain advanced features are
 not available with this backend.  For more details, see :ref:`parallel_db`.
 
 The parallel magics now display all output, so you can do parallel plotting or
 other actions with complex display.  The ``px`` magic has now both line and cell
 modes, and in cell mode finer control has been added about how to collate
 output from multiple engines. :ghpull:`1768`.
 
 There have also been incremental improvements to the SSH launchers:
     
 * add to_send/fetch steps for moving connection files around.
   
 * add SSHProxyEngineSetLauncher, for invoking to `ipcluster engines` on a
   remote host. This can be used to start a set of engines via PBS/SGE/MPI
   *remotely*.
     
 This makes the SSHLauncher usable on machines without shared filesystems.
 
 A number of 'sugar' methods/properties were added to AsyncResult that are
 quite useful (:ghpull:`1548`) for everday work:
     
     * ``ar.wall_time`` = received - submitted
     * ``ar.serial_time`` = sum of serial computation time
     * ``ar.elapsed`` = time since submission (wall_time if done)
     * ``ar.progress`` = (int) number of sub-tasks that have completed
     * ``len(ar)`` = # of tasks
     * ``ar.wait_interactive()``: prints progress
     
 Added :meth:`.Client.spin_thread` / :meth:`~.Client.stop_spin_thread` for
 running spin in a background thread, to keep zmq queue clear.  This can be used
 to ensure that timing information is as accurate as possible (at the cost of
 having a background thread active).
 
 Set TaskScheduler.hwm default to 1 instead of 0.  1 has more
 predictable/intuitive behavior, if often slower, and thus a more logical
 default.  Users whose workloads require maximum throughput and are largely
 homogeneous in time per task can make the optimization themselves, but now the
 behavior will be less surprising to new users. :ghpull:`1294`.
 
 
 Kernel/Engine unification
 -------------------------
 
 This is mostly work 'under the hood', but it is actually a *major* achievement
 for the project that has deep implications in the long term: at last, we have
 unified the main object that executes as the user's interactive shell (which we
 refer to as the *IPython kernel*) with the objects that run in all the worker
 nodes of the parallel computing facilities (the *IPython engines*).  Ever since
 the first implementation of IPython's parallel code back in 2006, we had wanted
 to have these two roles be played by the same machinery, but a number of
 technical reasons had prevented that from being true.
 
 In this release we have now merged them, and this has a number of important
 consequences:
 
 * It is now possible to connect any of our clients (qtconsole or terminal
   console) to any individual parallel engine, with the *exact* behavior of
   working at a 'regular' IPython console/qtconsole.  This makes debugging,
   plotting, etc. in parallel scenarios vastly easier.
 
 * Parallel engines can always execute arbitrary 'IPython code', that is, code
   that has magics, shell extensions, etc.  In combination with the ``%%px``
   magics, it is thus extremely natural for example to send to all engines a
   block of Cython or R code to be executed via the new Cython and R magics. For
   example, this snippet would send the R block to all active engines in a
   cluster::
 
     %%px
     %%R
     ... R code goes here
   
 * It is possible to embed not only an interactive shell with the
   :func:`IPython.embed` call as always, but now you can also embed a *kernel*
   with :func:`IPython.embed_kernel()`.  Embedding an IPython kernel in an
   application is useful when you want to use :func:`IPython.embed` but don't
   have a terminal attached on stdin and stdout.
 
 * The new :func:`IPython.parallel.bind_kernel` allows you to promote Engines to
   listening Kernels, and connect QtConsoles to an Engine and debug it
   directly.
 
 In addition, having a single core object through our entire architecture also
 makes the project conceptually cleaner, easier to maintain and more robust.
 This took a lot of work to get in place, but we are thrilled to have this major
 piece of architecture finally where we'd always wanted it to be.
   
 
 Official Public API
 -------------------
 
 We have begun organizing our API for easier public use, with an eye towards an
 official IPython 1.0 release which will firmly maintain this API compatible for
 its entire lifecycle.  There is now an :mod:`IPython.display` module that
 aggregates all display routines, and the :mod:`IPython.config` namespace has
 all public configuration tools.  We will continue improving our public API
 layout so that users only need to import names one level deeper than the main
 ``IPython`` package to access all public namespaces.
 
 
 IPython notebook file icons
 ---------------------------
 
 The directory ``docs/resources`` in the source distribution contains SVG and
 PNG versions of our file icons, as well as an ``Info.plist.example`` file with
 instructions to install them on Mac OSX.  This is a first draft of our icons,
 and we encourage contributions from users with graphic talent to improve them
 in the future:
 
 .. image:: ../../resources/ipynb_icon_128x128.png
    :alt:  IPython notebook file icon.
 
 	  
 New top-level `locate` command
 ------------------------------
 
 Add `locate` entry points; these would be useful for quickly locating IPython
 directories and profiles from other (non-Python) applications. :ghpull:`1762`.
     
 Examples::
     
     $> ipython locate
     /Users/me/.ipython
   
     $> ipython locate profile foo
     /Users/me/.ipython/profile_foo
   
     $> ipython locate profile
     /Users/me/.ipython/profile_default
   
     $> ipython locate profile dne
     [ProfileLocate] Profile u'dne' not found.
 
 	
 Other new features and improvements
 -----------------------------------
 
 * **%install_ext**: A new magic function to install an IPython extension from
   a URL. E.g. ``%install_ext
   https://bitbucket.org/birkenfeld/ipython-physics/raw/default/physics.py``.
 
 * The ``%loadpy`` magic is no longer restricted to Python files, and has been
   renamed ``%load``. The old name remains as an alias.
 
 * New command line arguments will help external programs find IPython folders:
   ``ipython locate`` finds the user's IPython directory, and ``ipython locate
   profile foo`` finds the folder for the 'foo' profile (if it exists).
 
 * The :envvar:`IPYTHON_DIR` environment variable, introduced in the Great
   Reorganization of 0.11 and existing only in versions 0.11-0.13, has been
   deprecated. As described in :ghpull:`1167`, the complexity and confusion of
   migrating to this variable is not worth the aesthetic improvement. Please use
   the historical :envvar:`IPYTHONDIR` environment variable instead.
 
 * The default value of *interactivity* passed from
   :meth:`~IPython.core.interactiveshell.InteractiveShell.run_cell` to
   :meth:`~IPython.core.interactiveshell.InteractiveShell.run_ast_nodes`
   is now configurable.
 
 * New ``%alias_magic`` function to conveniently create aliases of existing
   magics, if you prefer to have shorter names for personal use.
 
 * We ship unminified versions of the JavaScript libraries we use, to better
   comply with Debian's packaging policies.
 
 * Simplify the information presented by ``obj?/obj??`` to eliminate a few
   redundant fields when possible.  :ghpull:`2038`.
 
 * Improved continuous integration for IPython.  We now have automated test runs
   on `Shining Panda <https://jenkins.shiningpanda.com/ipython>`_ and `Travis-CI
   <http://travis-ci.org/#!/ipython/ipython>`_, as well as `Tox support
   <http://tox.testrun.org>`_.
 
 * The `vim-ipython`_ functionality (externally developed) has been updated to
   the latest version.
 
 .. _vim-ipython: https://github.com/ivanov/vim-ipython
 
 * The ``%save`` magic now has a ``-f`` flag to force overwriting, which makes
   it much more usable in the notebook where it is not possible to reply to
   interactive questions from the kernel. :ghpull:`1937`.
 
 * Use dvipng to format sympy.Matrix, enabling display of matrices in the Qt
   console with the sympy printing extension. :ghpull:`1861`.
 
 * Our messaging protocol now has a reasonable test suite, helping ensure that
   we don't accidentally deviate from the spec and possibly break third-party
   applications that may have been using it.  We encourage users to contribute
   more stringent tests to this part of the test suite.  :ghpull:`1627`.
 
 * Use LaTeX to display, on output, various built-in types with the SymPy
   printing extension. :ghpull:`1399`.
 
 * Add Gtk3 event loop integration and example. :ghpull:`1588`.
 
 * ``clear_output`` improvements, which allow things like progress bars and other
   simple animations to work well in the notebook (:ghpull:`1563`):
     
     * `clear_output()` clears the line, even in terminal IPython, the QtConsole
       and plain Python as well, by printing `\r` to streams.
     
     * `clear_output()` avoids the flicker in the notebook by adding a delay,
       and firing immediately upon the next actual display message.
     
     * `display_javascript` hides its `output_area` element, so using display to
       run a bunch of javascript doesn't result in ever-growing vertical space.
 
 * Add simple support for running inside a virtualenv.  While this doesn't
   supplant proper installation (as users should do), it helps ad-hoc calling of
   IPython from inside a virtualenv. :ghpull:`1388`.
 
   
 Major Bugs fixed
 ----------------
 
 In this cycle, we have :ref:`closed over 740 issues <issues_list_013>`, but a
 few major ones merit special mention:
 
 * The ``%pastebin`` magic has been updated to point to gist.github.com, since
   unfortunately http://paste.pocoo.org has closed down. We also added a -d flag
   for the user to provide a gist description string. :ghpull:`1670`.
 
 * Fix ``%paste`` that would reject certain valid inputs. :ghpull:`1258`.
 
 * Fix sending and receiving of Numpy structured arrays (those with composite
   dtypes, often used as recarrays). :ghpull:`2034`.
 
 * Reconnect when the websocket connection closes unexpectedly. :ghpull:`1577`.
 
 * Fix truncated representation of objects in the debugger by showing at least
   80 characters' worth of information.  :ghpull:`1793`.
 
 * Fix logger to be Unicode-aware: logging could crash ipython if there was
   unicode in the input. :ghpull:`1792`.
 
 * Fix images missing from XML/SVG export in the Qt console. :ghpull:`1449`.
 
 * Fix deepreload on Python 3. :ghpull:`1625`, as well as having a much cleaner
   and more robust implementation of deepreload in general. :ghpull:`1457`.
 
 
 Backwards incompatible changes
 ------------------------------
 
 * The exception :exc:`IPython.core.error.TryNext` previously accepted
   arguments and keyword arguments to be passed to the next implementation
   of the hook. This feature was removed as it made error message propagation
   difficult and violated the principle of loose coupling.