Closed
Description
A structured ndarray can be indexed with multiple fields:
In [14]: X = numpy.empty(shape=1, dtype="(2)3S,(2)3U")
In [15]: X[["f0","f1"]]
Out[15]:
array([([b'', b''], ['', ''])],
dtype=[('f0', 'S3', (2,)), ('f1', '<U3', (2,))])
However, the corresponding numpy.void type cannot:
In [16]: X[0][["f0","f1"]]
---------------------------------------------------------------------------
IndexError Traceback (most recent call last)
<ipython-input-16-4f230bb69b0d> in <module>()
----> 1 X[0][["f0","f1"]]
IndexError: invalid index
The same is true for masked arrays.
For consistently, numpy.void with a structured dtype should support indexing with multiple fields exactly like numpy.ndarray does.
Metadata
Metadata
Assignees
Labels
No labels