Enable resizing n-dimensional arrays in addition one-dimensional ones. #312
Conversation
adamreichold
force-pushed
the
n-dimensional-resize
branch
3 times, most recently
from
bbd9c45 to
656eb2b
Compare
| /// Extends or truncates the dimensions of an array. | ||
| /// | ||
| /// This method works only on [contiguous][`Self::is_contiguous`] arrays. | ||
| /// Missing elements will be initialized as if calling [`zeros`][Self::zeros]. |
There was a problem hiding this comment.
Out of curiosity what does this do for object? Fill with null references?
There was a problem hiding this comment.
From the doc comments of PyArray::zeros:
For elements with DATA_TYPE == DataType::Object, this will fill the array with valid pointers to zero-valued Python integer objects.
So more specifically
>>> a = np.zeros(1, dtype=object)
>>> a[0]
0
>>> type(a[0])
<class 'int'>
(PyArray::new correspoding to np.empty will put zero pointers into object arrays (instead of uninitialized values for primitives like integers), so these arrays are always safe to drop, but not to use.)
((I noticed this still references Element::DATA_TYPE which does not actually exist any more. Will resolve this as part of the copy-editing done for #309.))
There was a problem hiding this comment.
(This behaviour is yet another reason why PyArray<PyObject, D> works, but PyArray<Py<T>, D> with arbitrary T is really problematic.)
adamreichold
force-pushed
the
n-dimensional-resize
branch
from
656eb2b to
a31367d
Compare
No description provided.