Issues with mod_wsgi?

[evandavey]

Hi,
I keep getting a 'module' object has no attribute 'core' when using with mod_wsgi. Please see http://stackoverflow.com/questions/7766400/pandas-django-mod-wsgi-virtualenv for the full error description.
Evan

[wesm]

What error do you get trying to do:

import pandas._tseries
import pandas._sparse

This sort of error happens when there was an import error suppressed somewhere so that pandas/init.py didn't fully execute.

I am not well-equipped to debug this myself. Not sure what to tell you.

[evandavey]

Ok, that produced a slightly different error. Looks like this line 'module compiled against API version 6 but this version of numpy is 4' is the problem but not sure what it means... Both numpy and pandas installed using pip inside the virtual environment as in "pip install numpy; pip install pandas"

[Mon Oct 24 15:05:35 2011] [info] Initial (No.1) HTTPS request received for child 0 (server openportfolio-staging.XX.com:443)
[Mon Oct 24 15:05:35 2011] [info] mod_wsgi (pid=12433): Create interpreter 'openportfolio-staging'.
[Mon Oct 24 15:05:35 2011] [info] [client XX.XX.XX.XX] mod_wsgi (pid=12433, process='openportfolio-staging', application='openportfolio-staging'): Loading WSGI script '/usr/local/web/django/www/staging/openportfolio/apache/staging.wsgi'.
[Mon Oct 24 15:05:35 2011] [error] RuntimeError: module compiled against API version 6 but this version of numpy is 4
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] mod_wsgi (pid=12433): Target WSGI script '/usr/local/web/django/www/staging/openportfolio/apache/staging.wsgi' cannot be loaded as Python module.
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] mod_wsgi (pid=12433): Exception occurred processing WSGI script '/usr/local/web/django/www/staging/openportfolio/apache/staging.wsgi'.
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] Traceback (most recent call last):
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] File "/usr/local/web/django/www/staging/openportfolio/apache/staging.wsgi", line 22, in
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] import pandas._tseries
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] File "/usr/local/web/django/www/staging/env/openportfolio/lib/python2.7/site-packages/pandas/init.py", line 12, in
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] from pandas.core.api import *
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] File "/usr/local/web/django/www/staging/env/openportfolio/lib/python2.7/site-packages/pandas/core/api.py", line 8, in
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] from pandas.core.common import isnull, notnull, set_printoptions
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] File "/usr/local/web/django/www/staging/env/openportfolio/lib/python2.7/site-packages/pandas/core/common.py", line 13, in
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] import pandas._tseries as lib
[Mon Oct 24 15:05:35 2011] [error] [client XX.XX.XX.XX] ImportError: numpy.core.multiarray failed to import

[wesm]

hm, do you have multiple numpy's or multiple python interpreters installed on this box? try rebuilding pandas against the right one

[wesm]

Also, FYI you should take some care about installing NumPy from source. You almost certainly won't get optimized linear algebra out of the box-- so if you're using np.dot anywhere or anything in numpy.linalg (or any libraries that do) you should take a closer look there.

[evandavey]

So it turns out this was a Python path order issue. By running sys.path.reverse() in my wsgi config file, the code now runs.

I double checked that numpy and pandas had been built against the right version - which they had (makes sense since it all worked outside mod_wsgi). However, due to the order of Python path, the built in OS X numpy library must have been imported first causing the issue.

Thanks for your help.

[evandavey]

(and thanks for the info re: efficiency - I only build from source because I am trying to create an easy to install django app and a pip requirements file seems the easiest way to achieve this).

[wesm]

Cool, glad you got it working =)