Closed AndreasAlbert closed 5 years ago
Thank you for the explicit example! I had heard rumors of a bug like this, affecting argmax
but not max
because max
is explicitly protected. I had been looking at the wrong part of the argmax
code, but now I've found it.
The fix is going through continuous integration. When it's done, it will be version 0.12.0rc2. Since that's a release candidate (you caught me in the middle of working on a semimajor version), you have to ask for it as
pip install "awkward>=0.12.0rc2"
(or similarly for conda) in order to get a release from the 0.12 series, since there are no release-releases in this series yet.
When the third-to-last is green, it's ready for pip/conda-install.
https://travis-ci.org/scikit-hep/awkward-array/builds/556511902
Thanks for being super efficient about this! Will test as soon as possible.
The release candidate is deployed: https://pypi.org/project/awkward/#history but remember that you have to explicitly request it from pip or conda. (Or just use GitHub master.)
You caught me at a good time. :)
Hi, I encountered a problem when trying to get the
argmax
of an array I apply some selection to. If the selection happens to kill all entries,argmax
will fail even thoughmax
still works. I first suspected this was a duplicate of #100, but sincemax
seems to work I am not so sure. Anyway, here's a snippet of what I mean:The error message is:
I would have expected to get
which would be consistent with
Is this expected?
Thanks