wheerd / multiset Goto Github PK

Firstly, thanks for the module. There's 2 minor issues, sorry to cram them into one heap.

1. The documentation mentions .values() as a synonym for .multiplicities(), however this does not work. I would prefer .values() to work as documented (rather than say, deleting it from the docs). This would be consistent with the multiset working with other standard dict methods.

2. When dealing with large multisets, I am forced to do print(mset.__repr__()) because to the default .__str__() produces unwieldy output

>>> mset = Multiset({'Pumpkin': 1, 'Eggs': 3, 'Milk': 4, 'Sugar': 30})      
>>> print(mset.__repr__())
Multiset({'Pumpkin': 1, 'Eggs': 3, 'Milk': 4, 'Sugar': 30})
>>> print(mset)          
{Pumpkin, Eggs, Eggs, Eggs, Milk, Milk, Milk, Milk, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar, Sugar}

Multisets are often notated in a concise form, as referenced here

the multiset {a, a, a, a, a, b, b, b, c, d, d} ,which may be written in compact form as the map {a: 5, b: 3, c: 1, d: 2}, where for each distinct member we specify the multiplicity.

So I think the .__repr__() is perfectly acceptable stringification under all circumstances. If I need the entire list, I can always list(mset).

__hash__ in FrozenMultiset is implemented as return hash(tuple(sorted(self._elements.items()))). This is simple and correctly results in a hash value that does not depend on the order of items, but it is presumably quite slow and will fail if the elements are not orderable.
I would have to test to be sure, but I believe that a solution of simply summing the hashes, like Java's AbstractMap, would be faster and work for all hashable entries.

just started using multiset in my project and got these mypy messages:

Mypy found 1 error, 1 note in 1 file
verify_pgx_json_call.py : 1 error, 1 note
Skipping analyzing "multiset": module is installed, but missing library stubs or py.typed marker [import] (5:0)
See https://mypy.readthedocs.io/en/stable/running_mypy.html#missing-imports (5:0)

Any plan to add full type support to the library?

I have an AttributeError when I try to load a FrozenMultiset with pickle.load.
Namely, this code raises an AttributeError on line 118 of multiset.py when trying to access the _elements attribute.

import pickle as pk
import multiset as ms

fms = ms.FrozenMultiset("aabcd")
    
with open("my_file","w") as my_file:
    pk.dump(fms, my_file)

with open("my_file", "r") as my_file:
    fms_copy = pk.load(my_file)

I tried versions 2.1.1 and 2.0.3: both are missing the popitem() and keys() methods (but they are described in the doc). And apparently nothing replaced it.
Is this a feature or a bug?

Furthermore, the button "docs" links to doc version 2.0.4, for all the versions available.

• Multiset.__delitem__ is untested and appears to be broken: it doesn't update self._total. Maybe it should be implemented as self[element] = 0 to reuse the __setitem__ code?

• The docstring for BaseMultiset.union says "For a variant of the operation which modifies the multiset in place see :meth:`union`.". But... that's what I'm looking at :-)

• Your package metadata says you support both py2 and py3, but the wheel on pypi is marked as being py3-only. Maybe you need to set the universal=1 option in setup.cfg?

• Not really a bug, but I found the semantics of remove and discard surprising: I would have expected them to be the inverse of add, rather than removing/discarding all copies by default. Maybe a less surprising design would be to have remove remove 1 item, discard discard 0 or 1 items, and then have remove_all and discard_all methods for when you want to do that? I don't know if it's worth breaking backwards compatibility over though.

Look at these samples:

Input:
{
FrozenMultiset([frozenset({'b'}) , frozenset({'a'})])
} == {
FrozenMultiset([frozenset({'a'}) , frozenset({'b'})])
}
Output: False

Input:
{
frozenset([frozenset('b') , frozenset('a')])
}=={
frozenset([frozenset('a') , frozenset('b')])
}
Output: True

It seems Multiset can not perfectly inherit attributes of set ???

Dependabot couldn't authenticate with https://pypi.python.org/simple/.

You can provide authentication details in your Dependabot dashboard by clicking into the account menu (in the top right) and selecting 'Config variables'.

View the update logs.

I'd like to be able to write type hints with generic typing in the form FrozenMultiset[T] but the interpreter complaints, Could you extend Generic[T] from the typing package and then write typeing for the methods?