0.12.21

New Features

• Add support for Python 3.14.

Bug Fixes

• Fix Variables._relabel() method. Previously when given a superset
  of labels that happen to correspond to valid indices the wrong
  relabelling would sometimes be applied. See also
  #1408.

0.12.20

Bug Fixes

• Fix inheritance in ComposedSampler and ComposedPolySampler so that
  composite attributes override sampler attributes.

  See #1402.

0.12.19

New Features

• Use dict rather than collections.OrderedDict to capture the
  inputs in TrackingComposite. See also #1395.
• Add dimod.Scoped abstract base class for components that allocate
  scope-bound resources.
• Implement dimod.Scoped interface on dimod.Sampler and
  dimod.Composite.
  See dwave-system#555.
• Add quadratic_multi_knapsack generator function to
  dimod.generators.

0.12.18

New Features

• Add SampleSet.wait_id() as a proxy to the underlying QPU result
  future, when sample set is constructed from such a future. See
  #1392.
• Add multi_knapsack() generator function to dimod.generators.
• Add quadratic_knapsack() generator function to dimod.generators.

Upgrade Notes

• Drop support for Python 3.8.

Bug Fixes

• Fix warnings caused by using deprecated type aliases.

0.12.17

New Features

• Allow BinaryQuadraticModel.contract_variables() to contract
  variables without an interaction. Previously it raised a
  ValueError.
• Add power_r() generator function to dimod.generators. The
  weights of the generated Ising problems follow a power-law
  distribution with respect to their magnitude.
• Add bin_packing generator function to dimod.generators.
• Add labels keyword argument to
  ConstrainedQuadraticModel.iter_constraint_data() and
  ConstrainedQuadraticModel.iter_violations() methods. The labels
  keyword argument allows a user to generate the data/violations for a
  subset of the constraints. See #1381.
• Add knapsack generator function to dimod.generators.
• Support Python 3.13.
• Add quadratic_assignment generator function to dimod.generators.

0.12.16

New Features

• Add dimod/typing.pdx. This allows users to cimport common C
  types into other namespaces.
• Support NumPy 2.0. Also, no longer require NumPy at build-time.
• Reduce the size of manylinux wheels.

Upgrade Notes

• Remove Cython fused types dimod.cyutilities.SignedInteger,
  dimod.cyutilities.UnsignedInteger, dimod.cyutilities.Integer,
  and dimod.cyutilities.ConstInteger, dimod.cyutilities.Numeric,
  and dimod.cyutilities.ConstNumeric. Use the types in
  dimod.typing instead.
• Remove Cython fused types
  dimod.discrete.cydiscrete_quadratic_model.Unsigned,
  dimod.discrete.cydiscrete_quadratic_model.Integral32plus,
  dimod.discrete.cydiscrete_quadratic_model.Numeric, and
  dimod.discrete.cydiscrete_quadratic_model.Numeric32plus. Use the
  types in dimod.typing instead.
• Don't include debug info in built .so files.

0.12.15

New Features

• Add C++ QuadraticModelBase::remove_interactions() method.
• Add dimod.generators.random_binary_paint_shop_problem() binary paint shop problem generator.
• Add an abstract method QuadraticViewsMixin.add_quadratic().
• Add a method QuadraticViewsMixin.add_quadratic_from() to add quadratic interactions. This method is intended to be used as follows: ConstrainedQuadraticModel.objective.add_quadratic_from(), similarly to ConstrainedQuadraticModel.objective.add_linear_from().
• Add check_overlaps keyword argument to ConstrainedQuadraticModel.add_discrete(), see #1364.

Upgrade Notes

• All sublcasses of QuadraticViewsMixin, including ObjectiveView, QuadraticModel and BinaryQuadraticModel, now require to have the method add_quadratic().

0.12.14

New Features

• Add C++ dimod::abc::QuadraticModelBase::remove_variables() method
  and accompanying overloads.
• Speed up C++ dimod::Expression::remove_variables() method.

0.12.13

New Features

• Build wheels for Python 3.12.
• Build C++ extensions with Cython 3.

0.12.12

New Features

• Add mimo() function for generating a multi-input multiple-output (MIMO) channel-decoding problem.
• Add coordinated_multipoint() function for generating a coordinated multi-point (CoMP) decoding problem.