Scalar array-like with asarray=False passed though #2823
Description
Using the new asarray=False functionality with dask.from_array on a scalar array-like (shape=()) results in the array-like itself being returned by compute():
>>> import numpy as np
>>> import dask.array as da
>>> class MyArray:
... shape = ()
... dtype = np.dtype('i4')
... def __getitem__(self, idx):
... # This is never called by Dask
... pass
...
>>> a = MyArray()
>>> d = da.from_array(a, 1, asarray=False)
>>> d.compute()
<__main__.MyArray instance at 0x7f36bc726998>
This is preumably because __getitem__ is never called by Dask for scalar arrays. This results in errors if you try to process the Dask object, for example reshaping it:
>>> d = d.reshape((1,))
>>> d.compute()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File ".../dask/base.py", line 99, in compute
(result,) = compute(self, traverse=False, **kwargs)
File ".../dask/base.py", line 206, in compute
results = get(dsk, keys, **kwargs)
File ".../dask/local.py", line 562, in get_sync
return get_async(apply_sync, 1, dsk, keys, **kwargs)
File ".../dask/local.py", line 529, in get_async
fire_task()
File ".../dask/local.py", line 504, in fire_task
callback=queue.put)
File ".../local.py", line 551, in apply_sync
res = func(*args, **kwds)
File ".../dask/local.py", line 295, in execute_task
result = pack_exception(e, dumps)
File ".../dask/local.py", line 290, in execute_task
result = _execute_task(task, data)
File ".../dask/local.py", line 270, in _execute_task
args2 = [_execute_task(a, cache) for a in args]
File ".../dask/local.py", line 271, in _execute_task
return func(*args2)
File ".../dask/utils.py", line 688, in __call__
return getattr(obj, self.method)(*args, **kwargs)
AttributeError: MyArray instance has no attribute 'reshape'
The asarray keyword was added as part of the work to allow Dask to return masked arrays, introduced in #2301 and released in v0.15.3, though I believe v0.15.4 is more stable. I don't think it was intended to ever return the original array-like (though I could be mistaken).