`fromarray` (and other array creation functions) don't like 0d arrays

[bgrant]

It gives a maximum recursion limit exceeded error.

[bgrant]

More explicitly:

>>> context.fromarray(np.array(42))

...

RuntimeError: maximum recursion depth exceeded in comparison

[cowlicks]

Relevant parts of the omitted traceback:

In [1]: import numpy as np

In [2]: import distarray.dist as d

In [3]: c = d.Context()

In [4]: c.fromarray(np.array(42))
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-4-b1fd8fb2d2ef> in <module>()
----> 1 c.fromarray(np.array(42))

/home/blake/distarray/distarray/dist/context.py in fromndarray(self, arr, distribution)
    494         """
    495         if distribution is None:
--> 496             distribution = Distribution.from_shape(self, arr.shape)
    497         out = self.empty(distribution, dtype=arr.dtype)
    498         for index, value in numpy.ndenumerate(arr):

/home/blake/distarray/distarray/dist/maps.py in from_shape(cls, context, shape, dist, grid_shape, targets)
    450         if grid_shape is None:  # Make a new grid_shape if not provided.
    451             self.grid_shape = make_grid_shape(self.shape, self.dist,
--> 452                                               len(self.targets))
    453         else:  # Otherwise normalize the one passed in.
    454             self.grid_shape = normalize_grid_shape(grid_shape, self.ndim)

/home/blake/distarray/distarray/metadata_utils.py in make_grid_shape(shape, dist, comm_size)
     84 
     85     else:  # Main case: comm_size > 1, ndistdim > 1.
---> 86         factors = utils.mult_partitions(comm_size, ndistdim)
     87         if not factors:  # Can't factorize appropriately.
     88             raise GridShapeError("Cannot distribute array over processors.")

/home/blake/distarray/distarray/utils.py in mult_partitions(n, s)
     58     [(2, 26), (4, 13)]
     59     """
---> 60     return [tuple(flatten(p)) for p in mult_partitions_recurs(n, s)]
     61 
     62 

/home/blake/distarray/distarray/utils.py in mult_partitions_recurs(n, s, pd)
     67     fs = []
     68     for d in divs:
---> 69         fs.extend([(d,f) for f in mult_partitions_recurs(n/d, s-1, pd)])
     70         pd = d
     71     return fs

/home/blake/distarray/distarray/utils.py in mult_partitions_recurs(n, s, pd)
     67     fs = []
     68     for d in divs:
---> 69         fs.extend([(d,f) for f in mult_partitions_recurs(n/d, s-1, pd)])
     70         pd = d
     71     return fs

/home/blake/distarray/distarray/utils.py in mult_partitions_recurs(n, s, pd)
     67     fs = []
     68     for d in divs:
---> 69         fs.extend([(d,f) for f in mult_partitions_recurs(n/d, s-1, pd)])
     70         pd = d
     71     return fs
...

/home/blake/distarray/distarray/utils.py in mult_partitions_recurs(n, s, pd)
     66     divs = divisors_minmax(n, pd, int(sqrt(n)))
     67     fs = []
---> 68     for d in divs:
     69         fs.extend([(d,f) for f in mult_partitions_recurs(n/d, s-1, pd)])
     70         pd = d

/home/blake/distarray/distarray/utils.py in divisors_minmax(n, dmin, dmax)
     30     """Find the divisors of n in the interval (dmin,dmax]."""
     31     i = dmin+1
---> 32     while i<=dmax:
     33         if n % i == 0:
     34             yield i

RuntimeError: maximum recursion depth exceeded in comparison

In [5]: 

[bgrant]

Update: The error has changed to the following:

In [12]: c.fromarray(np.array(42))
[0:execute]:
---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)<ipython-input-1-07b4635bce1a> in <module>()
----> 1 __distarray__6a539b6c4a14c2bc.__distarray__3e7a79c7acb47f41 = distarray.local.empty(distarray.local.maps.Distribution(comm=__distarray__99d1103efbe043b40, dim_data=__distarray__6a539b6c4a14c2bc.__distarray__9554b6c3a0073f6a[__distarray__99d1103efbe043b40.Get_rank()]), dtype=__distarray__6a539b6c4a14c2bc.__distarray__8f98612ea42f7df7)
IndexError: list index out of range

[1:execute]:
---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)<ipython-input-1-07b4635bce1a> in <module>()
----> 1 __distarray__6a539b6c4a14c2bc.__distarray__3e7a79c7acb47f41 = distarray.local.empty(distarray.local.maps.Distribution(comm=__distarray__99d1103efbe043b40, dim_data=__distarray__6a539b6c4a14c2bc.__distarray__9554b6c3a0073f6a[__distarray__99d1103efbe043b40.Get_rank()]), dtype=__distarray__6a539b6c4a14c2bc.__distarray__8f98612ea42f7df7)
IndexError: list index out of range

[2:execute]:
---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)<ipython-input-1-07b4635bce1a> in <module>()
----> 1 __distarray__6a539b6c4a14c2bc.__distarray__3e7a79c7acb47f41 = distarray.local.empty(distarray.local.maps.Distribution(comm=__distarray__99d1103efbe043b40, dim_data=__distarray__6a539b6c4a14c2bc.__distarray__9554b6c3a0073f6a[__distarray__99d1103efbe043b40.Get_rank()]), dtype=__distarray__6a539b6c4a14c2bc.__distarray__8f98612ea42f7df7)
IndexError: list index out of range

[3:execute]:
---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)<ipython-input-1-07b4635bce1a> in <module>()
----> 1 __distarray__6a539b6c4a14c2bc.__distarray__3e7a79c7acb47f41 = distarray.local.empty(distarray.local.maps.Distribution(comm=__distarray__99d1103efbe043b40, dim_data=__distarray__6a539b6c4a14c2bc.__distarray__9554b6c3a0073f6a[__distarray__99d1103efbe043b40.Get_rank()]), dtype=__distarray__6a539b6c4a14c2bc.__distarray__8f98612ea42f7df7)
IndexError: list index out of range

[bgrant]

And although I can make a 0d Distribution, I get the same error as above when trying to do a c.zeros(distribution) for such a distribution.

[kwmsmith]

@bgrant is this already fixed, and if not, will it be fixed with your current 0D array work?

[bgrant]

I believe this is fixed by #479 and #480. Closing.