Semigroup.py fails when trying to join array and tuple

[mikesol]

Hi!

I've noticed that sometimes, the function concatArray in Semigroup.py receives a list and a tuple to concat, which results in the following error:

TypeError: can only concatenate tuple (not "list") to tuple

In my local copy of the library, I've changed it to this just to get things working:

def concatArray(xs):
    def _1(ys):
        foo = tuple(xs)
        bar = tuple(ys)
        return bar if not foo else foo if not bar else foo + bar
    return _1

But that feels a bit off.

Is there a general policy for tuples vs lists in python-land? For example, in Data.Functor in the prelude, I see that tuple is used, but in Data.Array I see that list is used.

If tuples and lists can be used interchangeably, then I'd make sure that the concat operator is changed to something like the code above.

[thautwarm]

Hello, thanks a lot for this issue.

I think the problem is not from concatArray, the bug should be an array is transformed into a python list, but it should be a tuple.

Is there a general policy for tuples vs lists in python-land? For example, in Data.Functor in the prelude, I see that tuple is used, but in Data.Array I see that list is used.

Yes, there is a document: https://github.com/purescript-python/purescript-python/blob/master/Interops.md

I think we have incorrectly made the Python FFI interfaces for Data.Array.

@Hong-Xiang This should be assigned to you, but if you are busy you could tell me where the incorrect codes are, I can make a fix.

@mikesol I'm sorry that we haven't reviewed some ffi libraries made in last months, including "purescript-arrays.py".

[mikesol]

I would recommend keeping a conversion to tuple and issuing a warning if something is not a tuple. As people add more packages to the FFI, it may be difficult to enforce the use of tuples, and having it crash here makes the system more fragile.