1.2.1

Minor release, as there are no new API features and no API changes.

However, this is quite a major update.

Fixes:

• Issue with cached randomization results when relying on member variables that are not themselves random variables.
• Performance issue for class based variables.

Improvements:

• Debugging experience is vastly improved with overhauled exceptions and reporting.
• Performance graphs for assessing the performance impact of changes.
• Docmentation updates.
• Benchmark now uses unit testing framework, so code is shared. Benchmark test cases now much easier to create.

1.2.0

• Add support for using class instance methods as constraints
  • Now supports inheritance and method overriding, better allowing constrainedrandom to be used in an object-oriented manner.
• Add documentation for using class features
• Add testing for using class features
• Expose sparsities to user via set_solver_mode()
• Fix obscure issue around random-length lists

Full Changelog: 1.1.2...1.2.0

https://pypi.org/project/constrainedrandom/1.2.0/

1.1.2

• Fix overflow bug for large variables in multi-var problems

https://pypi.org/project/constrainedrandom/1.1.2/

1.1.1

Bug fixes:

• Fix strange user bug when using non-pure function as a constraint.
• Add unit test to catch this bug.

Misc improvements:

• Change input sanity parsing to use exceptions, not asserts.

https://pypi.org/project/constrainedrandom/1.1.1/

1.1.0

1.1.0

New features:

• Add random-length lists by specifying rand_length in add_rand_var

Documentation:

• Documentation for random-length lists

Testing:

• Many new test cases for lists

Bug fixes:

• Very rare corner case for random lists using CSP found and fixed.

https://pypi.org/project/constrainedrandom/1.1.0/

1.0.3

• Better testing and bug fixes for globally-seeded objects.
• Fixed issue where list objects weren't deterministic when resetting the global seed with random.seed(x).
• Fixed issue that caused an unhandled exception when calling copy.deepcopy on a RandObj instance using global seeding.
• Split random list solver into naive and fast, add user-facing option to disable naive.
  • Previously this was optimized manually, but in a stateful way that impacted determinism.
  • Therefore this is considered a bug fix rather than a feature addition.
  • Change benchmarking for this accordingly.

1.0.2

• Fix issues when trying to call copy.deepcopy on a RandObj instance.
• Fix determinism issues between copied RandObj instances.
• Minor performance improvement for MultiVarProblem.

1.0.1

Bug fixes:

• Fix issues with list randomization quality of results.
• Fix bug with caching non-hashable types in sparse solver.

1.0.0

API changes:

• Constructor no longer mandates an instance of random.Random.
  • Determinism can be controlled globally via random.seed().
  • See documentation for details.
• add_multi_var_constraint -> add_constraint
  • Now supports adding extra single-var constraints too.
• set_naive_solve -> set_solver_mode
  • See documentation.
• Add with_constraints and with_values arguments to RandObj.randomize()
  • See documentation.
• add_variable now supports a length argument for random lists.
  • See documentation.

Other changes:

• Debugging information when constraints are unsolvable.
• Performance improvements.
• Determinism bugfixes around PYTHONHASHSEED.
• Improved "How to" guide.
• Much better unit tests.

0.0.1

Initial public Pypi package

Package:
https://pypi.org/project/constrainedrandom/0.0.1/

Documentation:
https://constrainedrandom.readthedocs.io/en/0.0.1/