* Deprecate ``bidict.OVERWRITE`` and ``bidict.IGNORE``.
A :class:`UserWarning` will now be emitted if these are used.
:attr:`bidict.DROP_OLD` and :attr:`bidict.DROP_NEW` should be used instead.
* Rename ``DuplicationPolicy`` to :class:`~bidict.OnDupAction`
(and implement it via an :class:`~enum.Enum`).
A :class:`~bidict.OnDupAction` may be one of
:attr:`~bidict.RAISE`,
:attr:`~bidict.DROP_OLD`, or
:attr:`~bidict.DROP_NEW`.
* Expose the new :class:`~bidict.OnDup` class,
a named (*key*, *val*, *kv*) tuple of :class:`~bidict.OnDupAction`\s
that should be taken upon encountering
the 3 kinds of duplication that can occur.
* Provide the
:attr:`~bidict.ON_DUP_DEFAULT`,
:attr:`~bidict.ON_DUP_RAISE`, and
:attr:`~bidict.ON_DUP_DROP_OLD`
:class:`~bidict.OnDup` convenience instances.
* Deprecate the
``on_dup_key``, ``on_dup_val``, and ``on_dup_kv`` arguments
of :meth:`~bidict.bidict.put` and :meth:`~bidict.bidict.putall`.
A :class:`UserWarning` will now be emitted if these are used.
They have been subsumed by the new *on_dup* argument,
which takes an :class:`~bidict.OnDup` instance.
Use it like this: ``bi.put(1, 2, OnDup(key=DROP_NEW))``.
Or better yet, pass one of the
``ON_DUP_*`` convenience instances
instead if possible.
See the updated :ref:`basic-usage:Values Must Be Unique` docs for more info.
* Deprecate the
``on_dup_key``, ``on_dup_val``, and ``on_dup_kv``
bidict class attributes.
A :class:`UserWarning` will now be emitted if these are used.
They have been subsumed by the new
:attr:`~bidict.bidict.on_dup` class attribute,
which takes an :class:`~bidict.OnDup` instance.
See the updated :doc:`extending` docs for example usage.
* Move
:meth:`bidict.BidictBase.values` to
:meth:`bidict.BidirectionalMapping.values`,
since the implementation is generic.
* No longer use ``__all__`` in ``bidict/__init__.py``.
* Cap max_size rather than disabling health checks and deadline as a
less heavyhanded way to improve hypothesis test reliability on Travis.
- Initialize from iterables instead of lists of items in more places.
- Generate arbitrarily many items (rather than max 10).
- Don't include floats and text in HASHABLES. Slows example generation
without actually finding more falsifying examples.
- Use simpler and more concise `st.tuples(...).map(...).filter(...)`
constructions rather than much more complicated `st.composite`-based
strategies.
- Simplify `test_iteritems_args_kw`.
- Split out more namedbidict tests so each one tests one thing.
- Use latest Travis pypy and pypy3.
- Only use a single Hypothesis profile ("custom"). Allow any of its
settings that should be tunable to be tuned via dedicated env vars
(currently just "HYPOTHESIS_MAX_EXAMPLES").
- Always turn off "too_slow" health check, not just for Travis cron builds.
Never want a build to fail only because tests took too long.
- Set --hypothesis-verbosity="verbose" for Travis cron builds.
- Set --hypothesis-show-statistics for Travis cron builds.
Joshua Bronson