Make weave tests behave

[rgommers]

Most importantly:

• make them not leave behind any files in the dir that the tests are run from
• do not print anything by default

Furthermore a lot of random cleanups and deletions of unused or commented out code.

[rgommers]

I have verified that the number of tests run and total test time is unchanged.

[coveralls]

Coverage remained the same when pulling 696492c on rgommers:weave-tests into 16fc0af on scipy:master.

[WarrenWeckesser]

First, a big "thank you" for doing this!

I just tried the PR, and got three weave errors:

Running unit tests for scipy
NumPy version 1.8.0
NumPy is installed in /home/warren/local_scipy/lib/python2.7/site-packages/numpy
SciPy version 0.14.0.dev-696492c
SciPy is installed in /home/warren/local_scipy/lib/python2.7/site-packages/scipy
Python version 2.7.6 |Anaconda 1.8.0 (64-bit)| (default, Nov 11 2013, 10:47:18) [GCC 4.1.2 20080704 (Red Hat 4.1.2-54)]
nose version 1.3.0

<snip>

======================================================================
FAIL: test_add_function_ordered (test_catalog.TestCatalog)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/warren/local_scipy/lib/python2.7/site-packages/scipy/weave/tests/test_catalog.py", line 521, in test_add_function_ordered
    assert_(funcs1[:2] == [string.lower,string.upper]),repr(funcs1)
  File "/home/warren/local_scipy/lib/python2.7/site-packages/numpy/testing/utils.py", line 44, in assert_
    raise AssertionError(msg)
AssertionError

======================================================================
FAIL: test_add_function_persistent1 (test_catalog.TestCatalog)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/warren/local_scipy/lib/python2.7/site-packages/scipy/weave/tests/test_catalog.py", line 473, in test_add_function_persistent1
    assert_(i in pfuncs)
  File "/home/warren/local_scipy/lib/python2.7/site-packages/numpy/testing/utils.py", line 44, in assert_
    raise AssertionError(msg)
AssertionError

======================================================================
FAIL: test_get_existing_files2 (test_catalog.TestCatalog)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/warren/local_scipy/lib/python2.7/site-packages/scipy/weave/tests/test_catalog.py", line 401, in test_get_existing_files2
    assert_(len(files) == 1)
  File "/home/warren/local_scipy/lib/python2.7/site-packages/numpy/testing/utils.py", line 44, in assert_
    raise AssertionError(msg)
AssertionError

----------------------------------------------------------------------
Ran 11334 tests in 709.137s

FAILED (KNOWNFAIL=171, SKIP=451, failures=4)

But there were no files left behind 😄

[argriffing]

works for me on a ubuntu derivative

runtests.py default mode

Running unit tests for scipy.weave
NumPy version 1.7.1
SciPy version 0.14.0.dev-696492c
Python version 2.7.5+ (default, Sep 19 2013, 13:48:49) [GCC 4.8.1]
Ran 9944 tests in 95.298s
OK (KNOWNFAIL=118, SKIP=447)

runtests.py --mode=full

Ran 11334 tests in 684.561s
OK (KNOWNFAIL=171, SKIP=462)

[rgommers]

@WarrenWeckesser I can't reproduce this. Neither can TravisCI or Alex. So could you investigate? I only changed the use of globals for backup_dir in these tests, in f0e4e5a.

[WarrenWeckesser]

So far all I know is that some catalog methods are returning empty lists when they shouldn't:

• In test_get_existing_files2, q.get_existing_files() returns an
  empty list.
• In test_add_function_persistent1, q.get_cataloged_functions('code')
  returns an empty list.
• In test_add_function_ordered, each of the calls t.get_functions('f'),
  t.get_functions('ff'), and t.get_functions('fff') returns an empty list.

I'll try to take another look this weekend.

FWIW, I'm using Anaconda's python in Ubuntu 12.04 (64 bit).

[rgommers]

I think it's a bug in add_function that somehow is triggered for you. If you add back import os and import string within the tests, are the failures gone?

[WarrenWeckesser]

I haven't debugged the errors, but I have some more information. On my system, import dbhash (in catalog.py before this PR) fails. If I put import dbhash back to force the use of _dumb_shelve.py for the shelve module in catalog.py, the tests pass.

[rgommers]

@WarrenWeckesser do you want me to just put import dbhash back (or you send me a PR to that effect)? Or do you want to debug this?

[WarrenWeckesser]

Apparently the Anaconda python distribution from Continuum is missing the library _bsddb.so (at least in Linux; see https://groups.google.com/a/continuum.io/forum/#!topic/anaconda/mCQL6fVx55A). That's why import dbhash fails. When this import fails, the code in catalog.py uses the module _dumb_shelve.py instead of python's standard shelve module. When this module is used, the tests pass.

With this PR, import dbhash has been removed, and import shelve succeeds. I don't know why using the standard shelve module triggers the three test failures. It might be related to the missing _bsddb.so, or perhaps there is something else wrong with shelve in Anaconda.

TL;DR: This might be a problem with Anaconda's build of the shelve module, and not a scipy problem. (Pinging @ilanschnell, @teoliphant, @asmeurer in case any of them can take a look.)

[rgommers]

OK now that rings a bell. bsddb is broken in various ways on various Pythons/platforms, not present at all in Python 3.x and is used in shelve under the hood. So I will revert removing what looked like an unused import and comment it properly. Then everything should be good to go; no need to investigate.

[coveralls]

Coverage remained the same when pulling 8f68821 on rgommers:weave-tests into 9c7017e on scipy:master.

[rgommers]

@WarrenWeckesser done, see last commit. Can you test again?

[WarrenWeckesser]

This function (test_function) looks like a test. It assigns values to a bunch of local variables that are never used. Nothing calls it. It should be removed or moved to a file in the tests directory. But that doesn't have to happen in this PR, if you want to restrict the changes in this PR to the more obvious bits of code clean up.

[WarrenWeckesser]

The tests now pass for me.

I made a few small suggestions as line notes. Other than those, I think the PR is ready to go.

[asmeurer]

Just now seeing this. Is there still something you want me to look at in the Anaconda package? I looked at our scipy recipe and there's nothing too out of the ordinary going on. We do have a small patch to skip a failing test, and we do some stuff when we link against MKL, but other than that it's just setup.py install.

[rgommers]

@asmeurer it's not your scipy but your Python install - that's lacking bsddb (stdlib module that sometimes doesn't get built, I don't remember exactly why). If you did that on purpose, no need to do anything. But you may want to fix that for Python 2.7 is you care about BSDDB support.

[asmeurer]

Yep, I just checked, and it's not there on Mac OS X or Linux (but it is on Windows). We do have some code to support it, but only on armv6l. I'm not sure what the store is there. @ilanschnell would have to comment.

But I also noticed that bsddb is deprecated, and was removed from Python 3 (http://docs.python.org/2/library/bsddb.html#module-bsddb).

[coveralls]

Coverage remained the same when pulling a78fbe5 on rgommers:weave-tests into df81077 on scipy:master.

[coveralls]

Coverage remained the same when pulling a78fbe5 on rgommers:weave-tests into df81077 on scipy:master.

[rgommers]

OK this all works now and contains added commits from Warren and Julian that needed to go in, so merging.