test failures with the current master #3330

ev-br · 2014-02-14T14:56:14Z

>>> from scipy.sparse import test
>>> test()

gives a bunch of errors of this sort:

======================================================================
ERROR: test_base.TestLIL.test_sum(<type 'numpy.complex128'>, 4)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/br/virtualenvs/scipy-dev/local/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/br/virtualenvs/scipy-dev/local/lib/python2.7/site-packages/scipy/sparse/tests/test_base.py", line 654, in check
    assert_array_almost_equal(dat.sum(axis=-2), datsp.sum(axis=-2))
  File "/home/br/virtualenvs/scipy-dev/local/lib/python2.7/site-packages/numpy/matrixlib/defmatrix.py", line 435, in sum
    return N.ndarray.sum(self, axis, dtype, out)._align(axis)
  File "/home/br/virtualenvs/scipy-dev/`local/lib/python2.7/site-packages/numpy/matrixlib/defmatrix.py", line 376, in _align
    raise ValueError, "unsupported axis"
ValueError: unsupported axis

----------------------------------------------------------------------
Ran 8645 tests in 80.713s

FAILED (KNOWNFAIL=251, SKIP=997, errors=246)
<nose.result.TextTestResult run=8645 errors=246 failures=0>
>>> import scipy
>>> scipy.__version__
'0.14.0.dev-9812e4c'
>>> import numpy
>>> numpy.__version__
'1.6.2'

This all is with python 2.7.3 on 64 bit Ubuntu Precise with stock atlas.

The text was updated successfully, but these errors were encountered:

pv · 2014-02-14T15:11:25Z

Numpy bug, which seems to have been (accidentally) fixed in a43d255dbc in v1.7.0.
Should protect that with Numpy version check in Scipy tests.

ev-br · 2014-02-14T22:59:23Z

Indeed, with numpy 1.8 (different machine, similar setup) sparse tests all pass. There are several pages of DeprecationWarnings though:

/home/br/installs/scipy/build/testenv/lib/python2.7/site-packages/scipy/sparse/tests/test_base.py:1619: DeprecationWarning: Implicitly casting between incompatible kinds. In a future numpy release, this will raise an error. Use casting="unsafe" if this is intentional.
  y += b

pv · 2014-02-15T00:02:54Z

The Deprecationwarnings should go away with Numpy 1.9, where += ends up calling __numpy_ufunc__.

rgommers · 2014-02-16T15:59:30Z

There are more issues in sparse with numpy 1.5.1:

FAILED (KNOWNFAIL=179, SKIP=961, errors=194, failures=355)

A lot of these:

======================================================================
FAIL: Failure: AssertionError (
Not equal to tolerance rtol=1e-07, atol=0

(mismatch 100.0%)
 x: array([[ 1.,  0.,  0.,  2.],
       [ 3.,  0.,  1.,  0.],
       [ 0.,  2.,  0.,  0.]])
 y: array([[ 0.,  0.,  0.,  4.],
       [ 0.,  0.,  2.,  0.],
       [ 0., -4.,  0.,  0.]]))
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/rgommers/.local/lib/python2.7/site-packages/nose/loader.py", line 518, in makeTest
    return self._makeTest(obj, parent)
  File "/home/rgommers/.local/lib/python2.7/site-packages/nose/loader.py", line 577, in _makeTest
    return MethodTestCase(obj)
  File "/home/rgommers/.local/lib/python2.7/site-packages/nose/case.py", line 345, in __init__
    self.inst = self.cls()
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 144, in __init__
    self.datsp = self.spmatrix(self.dat)
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 3554, in spmatrix
    assert_allclose(M.A, NC.A)
  File "/home/rgommers/.local/lib/python2.7/site-packages/numpy/testing/utils.py", line 1130, in assert_allclose
    verbose=verbose, header=header)
  File "/home/rgommers/.local/lib/python2.7/site-packages/numpy/testing/utils.py", line 618, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
Not equal to tolerance rtol=1e-07, atol=0

(mismatch 100.0%)
 x: array([[ 1.,  0.,  0.,  2.],
       [ 3.,  0.,  1.,  0.],
       [ 0.,  2.,  0.,  0.]])
 y: array([[ 0.,  0.,  0.,  4.],
       [ 0.,  0.,  2.,  0.],
       [ 0., -4.,  0.,  0.]])

And some of each of these three types of errors:

======================================================================
FAIL: test_base.Test64Bit.test_resiliency_limit_10(<class 'test_base.TestCSR'>, 'test_set_slice')
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/rgommers/.local/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/rgommers/.local/lib/python2.7/site-packages/numpy/testing/decorators.py", line 146, in skipper_func
    return f(*args, **kwargs)
  File "<string>", line 2, in check
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 93, in deco
    return func(*a, **kw)
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 3741, in check
    getattr(instance, method_name)()
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 2065, in test_set_slice
    assert_array_equal(A.todense(), B, repr(a))
  File "/home/rgommers/.local/lib/python2.7/site-packages/numpy/testing/utils.py", line 686, in assert_array_equal
    verbose=verbose, header='Arrays are not equal')
  File "/home/rgommers/.local/lib/python2.7/site-packages/numpy/testing/utils.py", line 618, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
Arrays are not equal
slice(3, None, None)
(mismatch 100.0%)
 x: matrix([[ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
        [ 1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.],
        [ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],...
 y: matrix([[ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],
        [ 1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.,  1.],
        [ 0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.],...

======================================================================
ERROR: test_base.Test64Bit.test_resiliency_limit_10(<class 'test_base.TestCSR'>, 'test_getnnz_axis')
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/rgommers/.local/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/rgommers/.local/lib/python2.7/site-packages/numpy/testing/decorators.py", line 146, in skipper_func
    return f(*args, **kwargs)
  File "<string>", line 2, in check
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 93, in deco
    return func(*a, **kw)
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 3741, in check
    getattr(instance, method_name)()
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 2694, in test_getnnz_axis
    assert_array_equal(bool_dat.sum(axis=0), datsp.getnnz(axis=0))
  File "/home/rgommers/Code/scipy/scipy/sparse/compressed.py", line 105, in getnnz
    return np.bincount(self.indices, minlength=N)
TypeError: 'minlength' is an invalid keyword argument for this function

======================================================================
ERROR: test_base.Test64Bit.test_resiliency_limit_10(<class 'test_base.TestBSR'>, 'test_minmax_axis')
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/rgommers/.local/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/rgommers/.local/lib/python2.7/site-packages/numpy/testing/decorators.py", line 146, in skipper_func
    return f(*args, **kwargs)
  File "<string>", line 2, in check
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 93, in deco
    return func(*a, **kw)
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 3741, in check
    getattr(instance, method_name)()
  File "/home/rgommers/Code/scipy/scipy/sparse/tests/test_base.py", line 2647, in test_minmax_axis
    assert_array_equal(X.max(axis=axis).A, D.max(axis=axis).A)
  File "/home/rgommers/Code/scipy/scipy/sparse/data.py", line 161, in max
    return self._min_or_max(axis, np.maximum)
  File "/home/rgommers/Code/scipy/scipy/sparse/data.py", line 147, in _min_or_max
    return self._min_or_max_axis(axis, min_or_max)
  File "/home/rgommers/Code/scipy/scipy/sparse/data.py", line 115, in _min_or_max_axis
    major_index, value = mat._minor_reduce(min_or_max)
  File "/home/rgommers/Code/scipy/scipy/sparse/compressed.py", line 591, in _minor_reduce
    value = ufunc.reduceat(self.data, self.indptr[major_index])
TypeError: array cannot be safely cast to required type

rgommers · 2014-02-25T20:57:53Z

numpy < 1.7.0 doesn't support negative values for axis keyword to sum method. Will send a fix for that in a bit.

pv · 2014-02-25T21:03:53Z

_csparsetools is also broken, as apparently fused types in Cython don't work as you'd believe.

Negative axis arguments are not supported by numpy < 1.7.0. Fixes part of scipygh-3330.

rgommers · 2014-02-25T22:25:05Z

The test_set_slice failures look like a real bug:

A = csc_matrix((5,10))
A[slice(3, None, None)] = 9
print(A.todense())
B = np.matrix(np.zeros((5,10), float))
B[slice(3, None, None)] = 9
print(B)

gives:

[[ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]
 [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]
 [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]
 [ 9.  9.  9.  9.  0.  9.  9.  9.  9.  9.]     # note the spurious 0 in this row
 [ 9.  9.  9.  9.  9.  9.  9.  9.  9.  9.]]
[[ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]
 [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]
 [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]
 [ 9.  9.  9.  9.  9.  9.  9.  9.  9.  9.]
 [ 9.  9.  9.  9.  9.  9.  9.  9.  9.  9.]]

Shows up with numpy 1.5.1 and 1.6.0, not for 1.8.0. On 32-bit linux.

pv · 2014-02-25T22:56:21Z

There's then some behavior change in numpy that breaks _cs_matrix._insert_many (pure-Python).

pv · 2014-02-25T23:12:58Z

This is the difference:

>>> x = np.array([0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 3, 3,3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6,6, 6, 6, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9,9], dtype=np.int32)

Numpy 1.5.1:

>>> np.unique(x, return_index=True)[1]
array([ 0, 13, 19, 21, 32, 40, 42, 55, 62, 68])

Numpy 1.7.1:

>>> np.unique(x, return_index=True)[1]
array([ 0,  7, 14, 21, 28, 35, 42, 49, 56, 63])

This seems to be a Numpy fixed bug: numpy/numpy#2655
Quite recent, too, so the issue also exists in Numpy 1.6.

rgommers · 2014-02-26T07:23:41Z

This one (full traceback in my first comment above), or more precisely 351x this one, is still present after all current PRs are merged:

FAIL: Failure: AssertionError (
Not equal to tolerance rtol=1e-07, atol=0

(mismatch 100.0%)
 x: array([[ 1.,  0.,  0.,  2.],
       [ 3.,  0.,  1.,  0.],
       [ 0.,  2.,  0.,  0.]])
 y: array([[ 0.,  0.,  0.,  4.],
       [ 0.,  0.,  2.,  0.],
       [ 0., -4.,  0.,  0.]]))

I haven't investigated yet.

rgommers · 2014-02-26T23:15:25Z

OK it's the failure in the comment above plus the usage of minlength keyword for bincount which are the remaining issues with 1.5.1. 1.6.0 and 1.8.0 are OK.

rgommers · 2014-02-27T20:07:33Z

All fixed. Thanks @pv.

jnothman · 2014-02-28T01:25:11Z

This one (full traceback in my first comment above), or more precisely 351x this one, is still present after all current PRs are merged:

Either this is a problem in toarray with duplicate indices, or in test_base._same_sum_duplicate. The former is basically all SWIGed. The latter includes arr.repeat(...) and arr[1::2] *= -1 but doesn't get much more complicated. Maybe I've missed something.

rgommers · 2014-02-28T07:11:07Z

@jnothman it's fixed now. I think 0ef20f9 did it.

jnothman · 2014-03-01T22:04:44Z

Oh! Thanks ;)

On 28 February 2014 18:11, Ralf Gommers notifications@github.com wrote:

@jnothman https://github.com/jnothman it's fixed now. I think 0ef20f90ef20f998adcdid it.

Reply to this email directly or view it on GitHubhttps://github.com//issues/3330#issuecomment-36326935
.

ev-br added the scipy.sparse label Feb 14, 2014

rgommers added the defect label Feb 14, 2014

rgommers added this to the 0.14.0 milestone Feb 14, 2014

This was referenced Feb 23, 2014

MAINT: make travis build one target against Numpy 1.5 #3377

Merged

BUG: stable evaluation of expit #3386

Merged

rgommers added a commit to rgommers/scipy that referenced this issue Feb 25, 2014

TST: fix sparse test errors due to axis=-1,-2 usage in np.matrix.sum().

48d994f

Negative axis arguments are not supported by numpy < 1.7.0. Fixes part of scipygh-3330.

rgommers mentioned this issue Feb 25, 2014

TST: fix sparse test errors due to axis=-1,-2 usage in np.matrix.sum(). #3391

Merged

rgommers mentioned this issue Feb 26, 2014

BUG: sparse/lil: fix up Cython bugs in fused type lookup #3392

Closed

pv mentioned this issue Feb 27, 2014

MAINT: sparse: Numpy 1.5 compatibility fixes #3408

Merged

rgommers closed this as completed Feb 27, 2014

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

test failures with the current master #3330

test failures with the current master #3330

ev-br commented Feb 14, 2014

pv commented Feb 14, 2014

ev-br commented Feb 14, 2014

pv commented Feb 15, 2014

rgommers commented Feb 16, 2014

rgommers commented Feb 25, 2014

pv commented Feb 25, 2014

rgommers commented Feb 25, 2014

pv commented Feb 25, 2014

pv commented Feb 25, 2014

rgommers commented Feb 26, 2014

rgommers commented Feb 26, 2014

rgommers commented Feb 27, 2014

jnothman commented Feb 28, 2014

rgommers commented Feb 28, 2014

jnothman commented Mar 1, 2014

test failures with the current master #3330

test failures with the current master #3330

Comments

ev-br commented Feb 14, 2014

pv commented Feb 14, 2014

ev-br commented Feb 14, 2014

pv commented Feb 15, 2014

rgommers commented Feb 16, 2014

rgommers commented Feb 25, 2014

pv commented Feb 25, 2014

rgommers commented Feb 25, 2014

pv commented Feb 25, 2014

pv commented Feb 25, 2014

rgommers commented Feb 26, 2014

rgommers commented Feb 26, 2014

rgommers commented Feb 27, 2014

jnothman commented Feb 28, 2014

rgommers commented Feb 28, 2014

jnothman commented Mar 1, 2014