- Enable tests
- Update to 2.1.3:
  - renamed popitem in delete_by_index in coold
  - removed pop() from coold
  - substituted _PyObject_GC_(UN)TRACK with PyObject_GC_(Un)Track
  - readding needed internal cpython header files for version >=
    3.8
  - restored compilation of dictobject.c
  - cleaning up thing for more canonical compilation (testing on
    3.7 only for now)
  - added segfault debugger and pure py builder
  - add cibuildwheel
  - fixed regression for coold.set() in Python 3.6
  - added to __repr__ in pure py version the module name,
    consistently with C extension
  - added support for | operand for C extension and for pure py
    version 3.6 - 3.8
  - added type hinting support for C extension
  - restored Py_BUILD_CORE instead of Py_BUILD_CORE_MODULE
  - updated code for cpython 3.7
  - readded deprecated FrozenOrderedDict
  - added pure py version of coold
  - fixed repr for subclasses
  - support for python 3.10
  - inhibited __ior__ in py version
  - added 3.7 and 3.8 c ext ; fixed subclass bugs
  - added python 3.8 support
  - aligned immutability exceptions with stdlib
  - other minor changes to readme
  - now __init__ does not throw an exception is reinvoked, as
    frozenset. Notable speedup in creation.
  - returned to the dict API, __hash__() apart
  - change `change __sub__ to be consistent with __and__
  - change `isdisjoint` accordingly to new `__and__` behavior
  - added `by="items" to `sorted()` method
  - docstring for __and__
  - now __and__ returns the intersection between a frozendict and
    a dict-like only with the *items* in common. Order of the
    *second parameter* is preserved.
  - get_deep() implemented
  - implemented isdisjoint()
  - improved bitwise end operator unit test
  - implemented method sorted()
  - added & operator (intersection)
  - implemented the CPython frozenset hash method. More slow but
    more memory-friendly
  - fixed deepcopy; various micro-optimizations; other minors
  - reverted calculation of hash using frozenset
  - improving `hash` speed and memory consumption
  - copy() and frozendict(another_frozendict) does return the
    same instance
  - fixed repr()
  - __sub__ now supports also text-like and generators-like
    iterables
  - now values of frozendict can be unhashable, as tuple.
    hash(myfrozendict) will raise an exception in that case; now
    empty frozendict is a singleton, like tuple and frozenset
  - speed improvements to constructor
  - removed useless attempt of micro-optimization
  - speedup to creating a copy of a frozendict; added testcase
    and benchmarks

• Files Changed
• Browse Source