TST: failing builds on master #14808

jreback · 2016-12-06T11:42:59Z

3.4 slow https://travis-ci.org/pandas-dev/pandas/jobs/181534281

======================================================================
FAIL: test_bar_log_no_subplots (pandas.tests.plotting.test_frame.TestDataFramePlots)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/build/pandas-dev/pandas/pandas/tests/plotting/test_frame.py", line 1130, in test_bar_log_no_subplots
    tm.assert_numpy_array_equal(ax.yaxis.get_ticklocs(), expected)
  File "/home/travis/build/pandas-dev/pandas/pandas/util/testing.py", line 1079, in assert_numpy_array_equal
    _raise(left, right, err_msg)
  File "/home/travis/build/pandas-dev/pandas/pandas/util/testing.py", line 1062, in _raise
    .format(obj), left.shape, right.shape)
  File "/home/travis/build/pandas-dev/pandas/pandas/util/testing.py", line 1018, in raise_assert_detail
    raise AssertionError(msg)
AssertionError: numpy array are different
numpy array shapes are different
[left]:  (5,)
[right]: (4,)
======================================================================
FAIL: test_bar_log_subplots (pandas.tests.plotting.test_frame.TestDataFramePlots)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/build/pandas-dev/pandas/pandas/tests/plotting/test_frame.py", line 1141, in test_bar_log_subplots
    tm.assert_numpy_array_equal(ax[0].yaxis.get_ticklocs(), expected)
  File "/home/travis/build/pandas-dev/pandas/pandas/util/testing.py", line 1079, in assert_numpy_array_equal
    _raise(left, right, err_msg)
  File "/home/travis/build/pandas-dev/pandas/pandas/util/testing.py", line 1062, in _raise
    .format(obj), left.shape, right.shape)
  File "/home/travis/build/pandas-dev/pandas/pandas/util/testing.py", line 1018, in raise_assert_detail
    raise AssertionError(msg)
AssertionError: numpy array are different
numpy array shapes are different
[left]:  (7,)
[right]: (6,)
======================================================================
FAIL: test_bar_log (pandas.tests.plotting.test_series.TestSeriesPlots)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/build/pandas-dev/pandas/pandas/tests/plotting/test_series.py", line 207, in test_bar_log
    tm.assert_numpy_array_equal(ax.yaxis.get_ticklocs(), expected)
  File "/home/travis/build/pandas-dev/pandas/pandas/util/testing.py", line 1079, in assert_numpy_array_equal
    _raise(left, right, err_msg)
  File "/home/travis/build/pandas-dev/pandas/pandas/util/testing.py", line 1062, in _raise
    .format(obj), left.shape, right.shape)
  File "/home/travis/build/pandas-dev/pandas/pandas/util/testing.py", line 1018, in raise_assert_detail
    raise AssertionError(msg)
AssertionError: numpy array are different
numpy array shapes are different
[left]:  (7,)
[right]: (6,)
----------------------------------------------------------------------
Ran 252 tests in 1103.335s

numpy-dev build: https://travis-ci.org/pandas-dev/pandas/jobs/181534284, closed by COMPAT: numpy compat with 1-ndim object array compat and broadcasting #14809

======================================================================
ERROR: test_scalar_na_cmp_corners (pandas.tests.series.test_operators.TestSeriesOperators)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/build/pandas-dev/pandas/pandas/core/ops.py", line 1170, in na_op
    raise_on_error=True, **eval_kwargs)
  File "/home/travis/build/pandas-dev/pandas/pandas/computation/expressions.py", line 210, in evaluate
    **eval_kwargs)
  File "/home/travis/build/pandas-dev/pandas/pandas/computation/expressions.py", line 63, in _evaluate_standard
    return op(a, b)
TypeError: unsupported operand type(s) for &: 'float' and 'float'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
  File "/home/travis/build/pandas-dev/pandas/pandas/core/internals.py", line 1160, in eval
    result = get_result(other)
  File "/home/travis/build/pandas-dev/pandas/pandas/core/internals.py", line 1129, in get_result
    result = func(values, other)
  File "/home/travis/build/pandas-dev/pandas/pandas/core/ops.py", line 1179, in na_op
    yrav = yrav[mask]
IndexError: boolean index did not match indexed array along dimension 0; dimension is 1 but corresponding boolean dimension is 9
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
  File "/home/travis/build/pandas-dev/pandas/pandas/tests/series/test_operators.py", line 1417, in test_scalar_na_cmp_corners
    self.assertRaises(ValueError, lambda: d.__and__(s, axis='columns'))
  File "/home/travis/build/pandas-dev/pandas/pandas/util/testing.py", line 2380, in assertRaises
    _callable(*args, **kwargs)
  File "/home/travis/build/pandas-dev/pandas/pandas/tests/series/test_operators.py", line 1417, in <lambda>
    self.assertRaises(ValueError, lambda: d.__and__(s, axis='columns'))
  File "/home/travis/build/pandas-dev/pandas/pandas/core/ops.py", line 1223, in f
    return self._combine_series(other, na_op, fill_value, axis, level)
  File "/home/travis/build/pandas-dev/pandas/pandas/core/frame.py", line 3539, in _combine_series
    fill_value=fill_value)
  File "/home/travis/build/pandas-dev/pandas/pandas/core/frame.py", line 3573, in _combine_match_columns
    axes=[left.columns, self.index])
  File "/home/travis/build/pandas-dev/pandas/pandas/core/internals.py", line 3163, in eval
    return self.apply('eval', **kwargs)
  File "/home/travis/build/pandas-dev/pandas/pandas/core/internals.py", line 3057, in apply
    applied = getattr(b, f)(**kwargs)
  File "/home/travis/build/pandas-dev/pandas/pandas/core/internals.py", line 1167, in eval
    result = handle_error()
  File "/home/travis/build/pandas-dev/pandas/pandas/core/internals.py", line 1150, in handle_error
    (repr(other), str(detail)))  # noqa
TypeError: Could not operate array([nan], dtype=object) with block values boolean index did not match indexed array along dimension 0; dimension is 1 but corresponding boolean dimension is 9
----------------------------------------------------------------------

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

xref pandas-dev#14808

jorisvandenbossche · 2016-12-06T12:57:33Z

The first ones (plotting failures) started due to PR #14747, although the PR looks unrelated (and only started failing on the master branch, not in the PR)

…#14809) xref #14808

jorisvandenbossche · 2016-12-06T13:02:12Z

It's probably matplotlib 2.0.0b4 -> 2.0.0rc1 that is the culprit (that's a difference between the failing and passing builds)

jorisvandenbossche · 2016-12-06T13:52:18Z

@tacaswell

matplotlib 1.5.2:

In [2]: matplotlib.__version__
Out[2]: '1.5.2'

In [3]: df = pd.DataFrame({'A': [3] * 5, 'B': range(1, 6)}, index=range(5))

In [6]: ax = df.plot(kind='bar', grid=True, log=True)

In [7]: ax.yaxis.get_ticklocs()
Out[7]: array([   0.1,    1. ,   10. ,  100. ])

In [8]: [t.get_text() for t in ax.yaxis.get_ticklabels()]
Out[8]: ['', '$\\mathdefault{10^{0}}$', '$\\mathdefault{10^{1}}$', '']

matplotlib 2.0.0b4:

In [3]: matplotlib.__version__
Out[3]: '2.0.0b4'

In [4]: df = pd.DataFrame({'A': [3] * 5, 'B': range(1, 6)}, index=range(5))

In [5]: ax = df.plot(kind='bar', grid=True, log=True)

In [6]: ax.yaxis.get_ticklocs()
Out[6]: array([   0.1,    1. ,   10. ,  100. ])

In [7]: [t.get_text() for t in ax.yaxis.get_ticklabels()]
Out[7]: ['', '${10^{0}}$', '', '']

With matplotlib 2.0.0rc1

In [3]: matplotlib.__version__
Out[3]: '2.0.0rc1'

In [4]: df = pd.DataFrame({'A': [3] * 5, 'B': range(1, 6)}, index=range(5))

In [5]: ax = df.plot(kind='bar', grid=True, log=True)

In [6]: ax.yaxis.get_ticklocs()
Out[6]: 
array([  1.00000000e-02,   1.00000000e-01,   1.00000000e+00,
         1.00000000e+01,   1.00000000e+02])

In [7]: [t.get_text() for t in ax.yaxis.get_ticklabels()]
Out[7]: ['', '', '${10^{0}}$', '', '']

I suppose the change in the labels is intentional? ("vastly improved log scale ticks" in your rc1 announcement email. By the way, although you put this in your email as one of the highlights, it is almost not mentioned in the "What's new" / "Changes to the default style" documents)
But, is the fact that the ticklocs now are one element longer (and include two values for which there is no label at the start) also intentional? If so, we will just adjust our tests, that's no problem.

jorisvandenbossche · 2016-12-15T08:54:50Z

@tacaswell Can you give a quick confirmation that this change is intentional? (or not) Then I just adapt our tests.

…#14809) xref #14808 (cherry picked from commit 0412732)

tacaswell · 2016-12-17T22:01:22Z

Sorry it took me so long to respond, this got lost in the wall of gh notifications I get 😞. I got behind about a month ago and am slowly digging my self back out. In the future, if you need my attention for pandas related plotting issues and I am not responding on GH feel free to escalate communication channels.

The extra tick labels are intentional, the lack of write up in the what's new is because our doc-writing is behind our code writing.

The space at the bottom is not

jorisvandenbossche · 2016-12-17T23:19:58Z

@tacaswell No problem!
Just to check: the space at the bottom was a bug that is solved since rc1? The extra minor ticklabels are indeed intentional, but was the extra major tickloc at 0.01 (without label) a consequence of the erroneous space? Or was this intentional as well.

Sorry for the questions, I currently don't have a master version of matplotlib, only the tagged rc, installed.

tacaswell · 2016-12-18T04:24:26Z

In [1]: import pandas as pd

In [2]: import matplotlib.pyplot as plt
plt.i
In [3]: plt.ion()

In [4]: df = pd.DataFrame({'A': [3] * 5, 'B': range(1, 6)}, index=range(5))

In [5]: ax = df.plot(kind='bar', grid=True, log=True)

In [6]: ax.yaxis.get_ticklocs()
Out[6]: array([   0.1,    1. ,   10. ,  100. ])

In [7]: [t.get_text() for t in ax.yaxis.get_ticklabels()]
Out[7]: ['', '${10^{0}}$', '', '']

In [8]: plt.matplotlib.__version__
Out[8]: '2.0.0rc1.post77+gf99ae8b'

I think the not-visible ticks are something we consider a private detail (or maybe I just wish they were a private detail 😉 ).

attn @efiring

tacaswell · 2016-12-18T04:25:14Z

Assuming my internet comes back overnight, there should be an rc2 tomorrow.

efiring · 2016-12-18T08:11:14Z

The problem here is that when we call the LogLocator, it returns a much larger range than needed. That range has probably increased--I suspect for no good reason, but I'm not positive. The axis.get_ticklocs() just passes on what it gets from the locator, so this is hardly what a user would expect from a public method with a name like that. The trimming to fit the view limits doesn't happen until draw time, in _update_ticks(). (Actually, that method is called more than once, which seems inefficient.)
I suspect there are many opportunities for rationalization and optimization in this whole sequence, but it's a big job.
Apart from the change in the amount of range overshoot, which is partly a result of the change in default margin strategy, it looks like all of this has been largely unchanged for years.

I think the upshot for now, with respect to this Pandas test, is that the tick locations returned by get_ticklocs have changed, and may change again. If they are to be used in a test, then the test itself needs to trim off any that are outside the view limits.

jreback · 2016-12-21T13:44:08Z

@tacaswell looks like 2.0.0rc2 did something: https://travis-ci.org/pandas-dev/pandas/jobs/185737365 (didn't fix though)

jorisvandenbossche · 2016-12-22T14:01:21Z

Tried with rc2, and I get the same figure as @tacaswell showed above for rc1.post78.

Regarding the new error, I suppose this is due to a fix in rc2 (possibly matplotlib/matplotlib#7598), as the result of ax = pd.Series([0.1, 0.01, 0.001]).plot(log=True, kind='bar') with rc2 is:

looks much better than with 1.5.3:

jorisvandenbossche · 2016-12-22T14:07:17Z

PR to fix our test: #14957

efiring · 2016-12-22T18:17:46Z

Do you really like that missing third bar with 1.5.3? And the bottom-to-top first bar? I don't.

jorisvandenbossche · 2016-12-22T23:41:35Z

@efiring maybe you misinterpreted my wording, but I certainly like the new rc2 plot! :)
That is also the reason that I made a PR that just changes our tests for the new behaviour of rc2

efiring · 2016-12-23T05:03:50Z

@jorisvandenbossche, you're right, I completely misread your earlier comment. I'm not sure how I managed that--it was perfectly clear.

jreback added the Testing pandas testing functions or related to the test suite label Dec 6, 2016

jreback added this to the 0.19.2 milestone Dec 6, 2016

jreback added a commit to jreback/pandas that referenced this issue Dec 6, 2016

COMPAT: numpy compat with 1-ndim object array compat and broadcasting

845a6d9

xref pandas-dev#14808

jreback added a commit to jreback/pandas that referenced this issue Dec 6, 2016

COMPAT: numpy compat with 1-ndim object array compat and broadcasting

cd7233d

xref pandas-dev#14808

jreback mentioned this issue Dec 6, 2016

COMPAT: numpy compat with 1-ndim object array compat and broadcasting #14809

Merged

jreback added a commit that referenced this issue Dec 6, 2016

COMPAT: numpy compat with 1-ndim object array compat and broadcasting (…

0412732

…#14809) xref #14808

jorisvandenbossche pushed a commit that referenced this issue Dec 15, 2016

COMPAT: numpy compat with 1-ndim object array compat and broadcasting (…

13f28f5

…#14809) xref #14808 (cherry picked from commit 0412732)

jreback modified the milestones: 0.20.0, 0.19.2 Dec 20, 2016

jorisvandenbossche mentioned this issue Dec 22, 2016

TST: matplotlib 2.0 fix in log limits for barplot (GH14808) #14957

Merged

jorisvandenbossche closed this as completed in #14957 Dec 24, 2016

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

TST: failing builds on master #14808

TST: failing builds on master #14808

jreback commented Dec 6, 2016 •

edited

Loading

jorisvandenbossche commented Dec 6, 2016 •

edited

Loading

jorisvandenbossche commented Dec 6, 2016

jorisvandenbossche commented Dec 6, 2016 •

edited

Loading

jorisvandenbossche commented Dec 15, 2016

tacaswell commented Dec 17, 2016

jorisvandenbossche commented Dec 17, 2016

tacaswell commented Dec 18, 2016

tacaswell commented Dec 18, 2016

efiring commented Dec 18, 2016

jreback commented Dec 21, 2016

jorisvandenbossche commented Dec 22, 2016 •

edited

Loading

jorisvandenbossche commented Dec 22, 2016

efiring commented Dec 22, 2016

jorisvandenbossche commented Dec 22, 2016

efiring commented Dec 23, 2016

TST: failing builds on master #14808

TST: failing builds on master #14808

Comments

jreback commented Dec 6, 2016 • edited Loading

jorisvandenbossche commented Dec 6, 2016 • edited Loading

jorisvandenbossche commented Dec 6, 2016

jorisvandenbossche commented Dec 6, 2016 • edited Loading

jorisvandenbossche commented Dec 15, 2016

tacaswell commented Dec 17, 2016

jorisvandenbossche commented Dec 17, 2016

tacaswell commented Dec 18, 2016

tacaswell commented Dec 18, 2016

efiring commented Dec 18, 2016

jreback commented Dec 21, 2016

jorisvandenbossche commented Dec 22, 2016 • edited Loading

jorisvandenbossche commented Dec 22, 2016

efiring commented Dec 22, 2016

jorisvandenbossche commented Dec 22, 2016

efiring commented Dec 23, 2016

jreback commented Dec 6, 2016 •

edited

Loading

jorisvandenbossche commented Dec 6, 2016 •

edited

Loading

jorisvandenbossche commented Dec 6, 2016 •

edited

Loading

jorisvandenbossche commented Dec 22, 2016 •

edited

Loading