Pandas Datetime Fixes and Dropping Python 3.5 Support

This release removes support for Python 3.5.

Modified pandas datetime/timedelta handling to be represented as a string with proper format strings in swagger. While pandas by default represents these as integers, we'll use strings for better swagger handling. An important note however, this only works for a subset of possible Pandas DataFrame orients ('records', 'index', 'columns', 'table'), due to current limitations with conforming to OpenApi version 2. v3 support will come later, at which point this handling should work for all orients.

Pandas Column Label Type Fix

A fix to handle Pandas columns with non-string labels

Relax Wrapt Version Binding

Patch release. Added a wider range of support for the Wrapt dependency to remove some possible version conflict issues on install.

Enhanced Pandas Support

Added support for providing Pandas orients for the PandasParameterType. This allows users to define the orient they want used when InferenceSchema converts a Pandas DataFrame to JSON and back.

Added more detailed support for nested handling.

Adding Support for Nested Parameter Handling

Changes have been made to allow for nested parameter handling. This means that the different parameter types can be embedded within each other and will be parsed as they are found, for example a numpy array as a value of a dictionary will now have the expected type definition within the produced swagger and type conversion on input.

Patch Release for Init Fix

Introduced a regression to the parameter types init file. By design we don't require users to have all dependent packages installed, only the ones for the specific types they want to interact with. By adding the parameter types to the init file we broke that handling. This release reverts that change.

Version 1 Public Release

Publishing InferenceSchema publicly and releasing version 1