Loman

* Added ``get_original_inputs`` to see source inputs of entire computation or a given set of nodes
* Added ``get_outputs``, ``o`` attribute-style access to get list of nodes fed by a particular node
* Added ``get_final_outputs`` to get end nodes of a computation or a given set of nodes
* Added ``restrict`` method to remove nodes unnecessary to calculate a given set of outputs
* Added ``rename_node`` method to rename a node, while ensuring that nodes which use it as an input continue to do so
* Added ``repoint`` method allowing all nodes which use a given node as an input to use an alternative node instead
* Documented ``get_inputs`` and ``i`` attribute-style accessor

---------------------

* Added class-style definitions of computations

---------------------

* Added support for multithreading when calculating nodes
* Update to use networkx 2.0
* Added ``print_errors`` method
* Added ``force`` parameter to ``insert`` method to allow no recalculation if value is not updated
* BUGFIX: Fix behavior when calculation node overwritten with value node

---------------------

* Methods set_tag and clear_tag support lists or generators of tags. Method nodes_by_tag can retrieve a list of nodes with a specific tag.
* Remove set_tags and clear_tags.
* Add node computation timing data, accessible through tim attribute-style access or get_timing method.
* compute method can accept a list of nodes to compute.
* Loman now uses pydotplus for visualization. Internally, visualization has two steps: converting a Computation to a networkx visualization DAG, and then converting that to a pydotplus Dot object.
* Added view method - creates and opens a temporary pdf visualization.
* draw and view methods can show timing information with colors='timing' option

---------------------

* Add node function decorator
* Add ConstantValue (with alias C) to provide constant values to function parameters without creating a placeholder node for that constant
* BUGFIX: Visualizing computations was broken in v0.1.1!

---------------------

* Support for Python 3.4 and 3.5
* Method and attribute-style accessors support lists of nodes
* Added support for node-tagging
* Compute method can optionally throw exceptions, for easier interactive debugging
* ``get_inputs`` method and ``i`` attribute-style access to get list of inputs to a node
* ``add_node`` takes optional inspect parameter to avoid inspection for performance
* ``add_node`` takes optional group to render graph layout with subgraphs
* ``draw_graphviz`` renamed to ``draw``
* ``draw_nx`` removed
* ``get_df`` renamed to ``to_df``
* ``get_value_dict`` renamed to ``to_dict``
* BUGFIX: implementation of _get_calc_nodes used by compute fixed
* BUGFIX: args parameters do not create spurious nodes
* BUGFIX: default function parameters do not cause placeholder node to be created
* BUGFIX: node states correctly updated when calling add_node with value parameter