BUG, MAINT: concatenate of empty sequences, fixes #1586 #7450
Conversation
|
Hmm... I'm wondering if this isn't one of those functions that would look better in Cython? We already have an |
|
That would be a simple method of moving things to Cython, and I guess speed wise probably fine. I have difficulties to get an idea of how much this specific code would be simplified. One thing that I always think could be very useful, would be if there was nditer/NpyIter aware code generation. The new iterator is pretty convoluted to set up, but I guess for that to be really useful, we have to trust fused types and I seem to remember there were some limitations to them, though maybe they are cleared up and I have used them before. |
|
|
|
☔ The latest upstream changes (presumably #8584) made this pull request unmergeable. Please resolve the merge conflicts. |
Empty non-arrays no longer participate in determining the dtype of the result. Have also refactored the code quite a bit, unifying the flattened and multidim branches wherever possible, as well as the error checks. Also a handful of new tests, including both the bug fixed and existing untested functionality.
c950966 to
3b9f052
Compare
|
☔ The latest upstream changes (presumably #8894) made this pull request unmergeable. Please resolve the merge conflicts. |
|
Sorry to keep creating conflicts here! Also, the commit message prefix is missing an I |
|
What does this PR think about: np.concatenate([np.array([], float), np.array([1, 2], int)]).dtypeI think we want the answer to be |
|
Closing. The new tests pass except for Eric's comment about Please reopen if it should be |
(#1586)
Empty non-arrays no longer participate in determining the dtype of
the result.
Have also refactored the code quite a bit, unifying the flattened
and multidim branches wherever possible, as well as the error checks.
Also a handful of new tests, including both the bug fixed and
existing untested functionality.