Explicitly set dtype of np.lexsort in group_rank #19679
Conversation
| @@ -531,7 +531,7 @@ def group_rank_{{name}}(ndarray[float64_t, ndim=2] out, | |||
| # each label corresponds to a different group value, | |||
| # the mask helps you differentiate missing values before | |||
| # performing sort on the actual values | |||
| _as = np.lexsort(order) | |||
| _as = np.lexsort(order).view(dtype=np.int64) | |||
There was a problem hiding this comment.
yeah this prob returns a platform int. this is ok
jreback
added
Dtype Conversions
|
ping on green. |
Codecov Report
@@ Coverage Diff @@
## master #19679 +/- ##
=======================================
Coverage 91.58% 91.58%
=======================================
Files 150 150
Lines 48864 48864
=======================================
Hits 44752 44752
Misses 4112 4112
Continue to review full report at Codecov.
|
|
@jreback should be good to go. Couple failures on the optional Travis tests that appear unrelated |
|
yeah these are actually NOT optional. they seem to be failing very recently. @TomAugspurger any idea what is happening? |
|
This is segaulting: https://travis-ci.org/MacPython/pandas-wheels/jobs/341233134 can you install a 32-bit version of miniconda and see if you can track the problem? |
|
https://travis-ci.org/pandas-dev/pandas/jobs/341072093 for the failure about I don't know why https://travis-ci.org/pandas-dev/pandas/jobs/341072091 timed out. Something in our build step seems to sometimes take a way longer time. |
|
I installed a 32 bit version of conda on my system, but can't seem to import pandas to debug. When trying to build I get the same error that Tom is getting above, and on import I get the below - is there an easy way to configure the development environment for a 32 bit installation to match what the Travis builds are doing? Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "pandas/__init__.py", line 19, in <module>
"Missing required dependencies {0}".format(missing_dependencies))
ImportError: Missing required dependencies ['pytz']# packages in environment at /Users/williamayd/miniconda3_32/envs/py27_32:
#
alabaster 0.7.3 py27_0
babel 2.0 py27_0
cython 0.23.1 py27_0
docutils 0.12 py27_0
jinja2 2.8 py27_0
markupsafe 0.23 py27_0
numpy 1.9.2 py27_0
openssl 1.0.1k 1
ortools-examples 6.4.4495 <pip>
pip 7.1.2 py27_0
py 1.5.2 py_0
pygments 2.0.2 py27_0
pytest 3.2.2 py_0
python 2.7.10 0
python-dateutil 2.4.2 py27_0
pytz 2018.3 py_0
readline 6.2 2
setuptools 18.1 py27_0
six 1.9.0 py27_0
snowballstemmer 1.2.0 py27_0
sphinx 1.3.1 py27_0
sphinx-rtd-theme 0.1.7 <pip>
sphinx_rtd_theme 0.1.7 py27_0
sqlite 3.8.4.1 1
tk 8.5.18 0
wheel 0.24.0 py27_0
zlib 1.2.8 0 |
|
Alright I was able to get an environment configured using |
|
right but can alwyas pip install dateutil |
|
Good to hear. I think Anaconda dropped official support for osx-32 a while
back.
…On Wed, Feb 14, 2018 at 1:19 PM, William Ayd ***@***.***> wrote:
Alright I was able to get an environment configured using pip instead of
conda. FWIW I don't think the latter is a possibility to use. At the very
least, the latest osx32 python-dateutil package on anaconda is 2.4.2 but
pandas requires at least 2.5.0
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#19679 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ABQHIgB2fBZacH5NVRLjtd22KvQpXeRSks5tUzHDgaJpZM4SEMAL>
.
|
xref #19481
I didn't see any other instance in
group_rankwhere the dtype was open to interpretation, so I'm wondering ifnp.lexsortis returning a plaininton 32 bit systems. Hoping that explicitly getting a view of thatnp.lexsortresult to match the dtype of_aswill resolve the test issues