Dimod

Fixes
-----
* Fix `num_occurrences` not being aggregated correctly in `SampleSet.aggregate`

New Features
------------
* `TrackingComposite`, a new composite that records inputs and outputs for testing
* `NullSampler`, a new sampler that always returns an empty sample set
* `child_structure_dfs` function that searches child samplers for a structured
* `HigherOrderComposite` now supports the `initial_state` keyword argument
* `BinaryQuadraticModel.is_writeable` flag that toggles whether the BQM is treated as immutable or not
* `SampleSet.is_writeable` flag that toggles whether the sample set is treated as immutable or not

New Features
-----------
* `SampleSet.resolve` method that resolves any futures the sample set was constructed from.
* `SampleSet` is now pickleable in python2

Fixes
-----
* `ran_r` now creates vectors of the correct size.
* `BQM.from_numpy_vectors` now checks the size of the input vectors for correctness.
* `SampleSet`s with unresolved futures now block before pickling.

Fixes
-----
* `SampleSet.from_serializable` now correctly deserializes tuple variables
* `SampleSet.aggregate` now correctly preserves order
* `Variables.relabel` now ignores unknown labels
* `BinaryQuadraticModel.from_serializable` now correctly deserializes `BinaryQuadraticModel.info`

New Features
------------
* `ran_r` function to generate RANr binary quadratic models
* `SampleSet.slice` method to create a new sample set with rows sliced according to the standard Python slicing syntax
* `as_vartype` function to cast permitted inputs to a valid `Vartype` enum

Fixes
-----
* `SampleSet.from_serializable` now correctly preserves `SampleSet.info`

New Features
------------
* `SampleSet.lowest` method which returns a new sample set containing the lowest energy samples

Fixes
-----
* `SampleSet.samples` now returns an iterator. This is a backwards-compatibility fix and is a deprecated feature.