0.11.0.dev-4b22372.win-amd64-py2.7: nosetests error

[bluefir]

ERROR: test_operators_timedelta64 (pandas.tests.test_series.TestSeries)

Traceback (most recent call last):
File "C:\Python27\lib\site-packages\pandas\tests\test_series.py", line 1752, i
n test_operators_timedelta64
result = resultb + df['A']
File "C:\Python27\lib\site-packages\pandas\core\series.py", line 144, in wrapp
er
lvalues = lvalues.values
AttributeError: 'numpy.ndarray' object has no attribute 'values'

---

Ran 3113 tests in 241.670s

FAILED (SKIP=94, errors=1)

[ghost]

I've run the full test suite several times today with no errros. can you
try it on the latest git master and tell us if there's still a problem?
I don't think any of the devs are on windows, we might be missing a problem.

[jreback]

I just ran on windows x64/python 2.7.3, with latest master, test_series/frame passes

maybe try a later binary (think they are generated 5pm?)

[bluefir]

I usually grab binaries on http://pandas.pydata.org/pandas-build/dev/, and the latest is dated March 14. That's the one I used. If you guys say that the latest code works fine, I'll just wait for the new binary.

[jreback]

yep...i think the box those are built on is in the process of moving.....check back periodically...

[stephenwlin]

hmm, i don't know what's up but i just got this exact error too (ubuntu, 32-bit...fresh installing on different machine than my normal one from master)...have not investigated yet

[jreback]

@bluefir looks like the binaries are updated....can you give a test?

[stephenwlin]

@jreback replying here instead

yep, same exact error as above, on two different machines

pip freeze output:

Cython==0.18
Pygments==1.6
SQLAlchemy==0.8.0
argparse==1.2.1
ipython==0.13.1
logilab-astng==0.24.2
logilab-common==0.59.0
matplotlib==1.2.0
nose==1.2.1
numexpr==2.0.1
numpy==1.7.0
openpyxl==1.6.1
pandas==0.10.1
pylint==0.27.0
python-dateutil==1.5
pytz==2013b
pyzmq==13.0.0
scipy==0.11.0
six==1.3.0
statsmodels==0.4.3
vbench==0.1
wsgiref==0.1.2
xlrd==0.9.0
xlwt==0.7.4
yolk==0.4.3

some of that might be relevant

[jreback]

@bluefir can you conform that you are on numpy 1.7.0?

[jreback]

@y-p can you make that new build (for slow tests, which uses 2.7, also use numpy 1.7?),
I believe we test 1.6.2 on the main 2.7 one

[ghost]

pushed.

[bluefir]

Yes, I am on numpy 1.7.0. Just ran the most recent binaries pandas-0.11.0.dev-eb07c5a.win-amd64-py2.7 and got this:

ERROR: test_operators_timedelta64 (pandas.tests.test_series.TestSeries)

Traceback (most recent call last):
File "C:\Python27\lib\site-packages\pandas\tests\test_series.py", line 1775, i
n test_operators_timedelta64
result = resultb + df['A']
File "C:\Python27\lib\site-packages\pandas\core\series.py", line 144, in wrapp
er
lvalues = lvalues.values
AttributeError: 'numpy.ndarray' object has no attribute 'values'

---

[jreback]

@stephenwlin ok....fixed up, will push in a few

basically numpy 1.7 does something even more weird with timedelta64 that p3k does, it coerces when you do np.array(s) when s has dtype timedelta64[ns] (timedelta64[ns] is an integer dtype), noting that neigher 1.6.2 or py3k do this (when on 1.7)...weird

[jreback]

@bluefir thanks.....these were built JUST before we realized the change was because of 1.7. thanks for the report!

[jreback]

@stephenwlin pls update to master and try ....should work