FEAT: Adding pickle support to quaddtype

[SwayamInSync]

This PR adds the pickle support as per #231
Usage:

original = np.array([1.5, 2.5, 3.5, 4.5], dtype=QuadPrecDType(backend=backend))
# Pickle and unpickle
pickled = pickle.dumps(original)
unpickled = pickle.loads(pickled)

Using np.savez

arr1 = np.array([1.5, 2.5, 3.5], dtype=QuadPrecDType(backend=backend))
arr2 = np.array([[1.0, 2.0], [3.0, 4.0]], dtype=QuadPrecDType(backend=backend))

np.savez(fname, array1=arr1, array2=arr2)
loaded = np.load(fname, allow_pickle=True)	# allow_pickle must be True during loading
loaded_arr1 = loaded['array1']
loaded_arr2 = loaded['array2']

[SwayamInSync]

We might not need to use NPY_LIST_PICKLE as when NumPy pickles a QuadPrecDType array:

• It saves the dtype (using our reduce) (QuadPrecDType, ('sleef',))
• It saves the shape
• It saves the data, PyArray_ToString returns raw 128-bit bytes

StringDType needs NPY_LIST_PICKLE because:

• Its memory contains pointers to Python strings
• PyArray_ToString would just copy worthless pointer values
• So it uses _getlist_pkl to extract actual string objects

Our QuadPrecDType stores actual numeric values in memory, so PyArray_ToString works perfectly

[SwayamInSync]

hash implementation might require some careful handling https://github.com/python/cpython/blob/20b69aac0d19a5e5358362410d9710887762f0e7/Python/pyhash.c#L87

I'll do this in a different PR and then we can close the #231

[jorenham]

stub changes looks good to me

[juntyr]

Why is the backend provided as an integer? A string literal might be a lot clearer, or if it has to be an int maybe we could add an int enum?

[SwayamInSync]

Why is the backend provided as an integer? A string literal might be a lot clearer, or if it has to be an int maybe we could add an int enum?

Internally backend is defined as enum, we provide the friendly API for the user to add input as string. For the sake of static tools, IDE suggestions I think defining as enum might make sense maybe @jorenham can give some inputs here

[jorenham]

I suppose you could do Backend: typing.TypeAlias = typing.Literal[1, 2], or maybe even an enum.IntEnum

[SwayamInSync]

@ngoldbaum a gentle reminder here for review

[ngoldbaum]

@SwayamInSync I'd appreciate it if you could give me more than a couple of days - especially over a weekend - before pinging me on PR reviews. I had your PRs on my to-do list.

[ngoldbaum]

The C code and tests overall look good. I didn't review the tests in detail but I'm glad to see such extensive coverage of usages.

[ngoldbaum]

My only comment is that this test is a little too tightly coupled to the implementation to be useful.

You could, for example, decide to use __getnewargs__ instead of __reduce__ but then you'd have to update this test too. https://docs.python.org/3/library/pickle.html#object.__getnewargs__.

[SwayamInSync]

Right, I think I should just remove this test? as we already have tests which are just focussed on working of the pickle rather than implementation?

[ngoldbaum]

Yeah, I'm suggesting to delete the test.

[SwayamInSync]

@SwayamInSync I'd appreciate it if you could give me more than a couple of days - especially over a weekend - before pinging me on PR reviews. I had your PRs on my to-do list.

So sorry, I was working this weekend as well so slightly forgot the track of days :)

[SwayamInSync]

Merging this in