gnu/packages/patches/python-statsmodels-fix-tests.patch

   1 This patch fixes a couple of test failures introduced by changes to the pandas
   2 package.  It was extracted from this pull request:
   3
   4 https://github.com/statsmodels/statsmodels/pull/2675
   5
   6
   7 From c9ef60a7bc4407766ab9e9f12c8a6b89013046ee Mon Sep 17 00:00:00 2001
   8 From: Ralf Gommers <ralf.gommers@gmail.com>
   9 Date: Tue, 20 Oct 2015 07:34:11 +0200
  10 Subject: [PATCH 1/4] MAINT: fix use of old_behavior kw for numpy.correlate.
  11  Was removed in 1.10.0
  12
  13 Numpy PR that removed it: https://github.com/numpy/numpy/pull/5991
  14
  15 Closes gh-2667.
  16 ---
  17  statsmodels/tsa/ar_model.py | 6 ++----
  18  1 file changed, 2 insertions(+), 4 deletions(-)
  19
  20 diff --git a/statsmodels/tsa/ar_model.py b/statsmodels/tsa/ar_model.py
  21 index 087a9e0..02984bd 100644
  22 --- a/statsmodels/tsa/ar_model.py
  23 +++ b/statsmodels/tsa/ar_model.py
  24 @@ -261,10 +261,8 @@ def _presample_varcov(self, params):
  25
  26          Vpinv = np.zeros((p, p), dtype=params.dtype)
  27          for i in range(1, p1):
  28 -            Vpinv[i-1, i-1:] = np.correlate(params0, params0[:i],
  29 -                                            old_behavior=False)[:-1]
  30 -            Vpinv[i-1, i-1:] -= np.correlate(params0[-i:], params0,
  31 -                                             old_behavior=False)[:-1]
  32 +            Vpinv[i-1, i-1:] = np.correlate(params0, params0[:i],)[:-1]
  33 +            Vpinv[i-1, i-1:] -= np.correlate(params0[-i:], params0,)[:-1]
  34
  35          Vpinv = Vpinv + Vpinv.T - np.diag(Vpinv.diagonal())
  36          return Vpinv
  37
  38 From f1dc8979b09bc1736149993f895943b3158ee2db Mon Sep 17 00:00:00 2001
  39 From: Ralf Gommers <ralf.gommers@gmail.com>
  40 Date: Wed, 21 Oct 2015 22:05:52 +0200
  41 Subject: [PATCH 2/4] MAINT: fix graphics module for changes in recent pandas
  42  versions.
  43
  44 ---
  45  statsmodels/graphics/tests/test_mosaicplot.py | 2 +-
  46  statsmodels/graphics/tests/test_tsaplots.py   | 6 +++---
  47  statsmodels/graphics/tsaplots.py              | 2 +-
  48  3 files changed, 5 insertions(+), 5 deletions(-)
  49
  50 diff --git a/statsmodels/graphics/tests/test_mosaicplot.py b/statsmodels/graphics/tests/test_mosaicplot.py
  51 index cb9bbbe..e41020e 100644
  52 --- a/statsmodels/graphics/tests/test_mosaicplot.py
  53 +++ b/statsmodels/graphics/tests/test_mosaicplot.py
  54 @@ -113,7 +113,7 @@ def test_mosaic():
  55      # sort by the marriage quality and give meaningful name
  56      # [rate_marriage, age, yrs_married, children,
  57      # religious, educ, occupation, occupation_husb]
  58 -    datas = datas.sort(['rate_marriage', 'religious'])
  59 +    datas = datas.sort_values(by=['rate_marriage', 'religious'])
  60      num_to_desc = {1: 'awful', 2: 'bad', 3: 'intermediate',
  61                        4: 'good', 5: 'wonderful'}
  62      datas['rate_marriage'] = datas['rate_marriage'].map(num_to_desc)
  63 diff --git a/statsmodels/graphics/tests/test_tsaplots.py b/statsmodels/graphics/tests/test_tsaplots.py
  64 index 511f18f..365be82 100644
  65 --- a/statsmodels/graphics/tests/test_tsaplots.py
  66 +++ b/statsmodels/graphics/tests/test_tsaplots.py
  67 @@ -1,4 +1,4 @@
  68 -from statsmodels.compat.python import lmap, lzip, map
  69 +from statsmodels.compat.python import lmap, map
  70  import numpy as np
  71  import pandas as pd
  72  from numpy.testing import dec
  73 @@ -51,8 +51,8 @@ def test_plot_month():
  74      dta = sm.datasets.elnino.load_pandas().data
  75      dta['YEAR'] = dta.YEAR.astype(int).apply(str)
  76      dta = dta.set_index('YEAR').T.unstack()
  77 -    dates = lmap(lambda x : pd.datetools.parse('1 '+' '.join(x)),
  78 -                                            dta.index.values)
  79 +    dates = lmap(lambda x : pd.datetools.parse_time_string('1 '+' '.join(x))[0],
  80 +                                                           dta.index.values)
  81
  82      # test dates argument
  83      fig = month_plot(dta.values, dates=dates, ylabel='el nino')
  84 diff --git a/statsmodels/graphics/tsaplots.py b/statsmodels/graphics/tsaplots.py
  85 index 3d04692..94626c9 100644
  86 --- a/statsmodels/graphics/tsaplots.py
  87 +++ b/statsmodels/graphics/tsaplots.py
  88 @@ -200,7 +200,7 @@ def seasonal_plot(grouped_x, xticklabels, ylabel=None, ax=None):
  89      ticks = []
  90      for season, df in grouped_x:
  91          df = df.copy() # or sort balks for series. may be better way
  92 -        df.sort()
  93 +        df.sort_values(inplace=True)
  94          nobs = len(df)
  95          x_plot = np.arange(start, start + nobs)
  96          ticks.append(x_plot.mean())
  97
  98 From 4cfbef6af137629c6953f1f025d9cfc781874256 Mon Sep 17 00:00:00 2001
  99 From: Ralf Gommers <ralf.gommers@gmail.com>
 100 Date: Wed, 21 Oct 2015 22:15:25 +0200
 101 Subject: [PATCH 3/4] MAINT: work around pandas breaking backwards compat for
 102  pandas.version
 103
 104 ---
 105  setup.py                     | 5 ++++-
 106  statsmodels/tools/testing.py | 6 ++----
 107  2 files changed, 6 insertions(+), 5 deletions(-)
 108
 109 diff --git a/setup.py b/setup.py
 110 index 0002840..74aefb8 100644
 111 --- a/setup.py
 112 +++ b/setup.py
 113 @@ -134,7 +134,10 @@ def check_dependency_versions(min_versions):
 114                                (spversion, min_versions['scipy']))
 115
 116      try:
 117 -        from pandas.version import short_version as pversion
 118 +        import pandas
 119 +        #FIXME: this will break for pandas 1.0.0.  Needs elaborate parsing now,
 120 +        # due to pandas removing version.short_version
 121 +        pversion = pandas.__version__[:6]
 122      except ImportError:
 123          install_requires.append('pandas')
 124      else:
 125 diff --git a/statsmodels/tools/testing.py b/statsmodels/tools/testing.py
 126 index e207e44..643f79f 100644
 127 --- a/statsmodels/tools/testing.py
 128 +++ b/statsmodels/tools/testing.py
 129 @@ -16,10 +16,8 @@ def strip_rc(version):
 130
 131
 132  def is_pandas_min_version(min_version):
 133 -    '''check whether pandas is at least min_version
 134 -    '''
 135 -    from pandas.version import short_version as pversion
 136 -    return StrictVersion(strip_rc(pversion)) >= min_version
 137 +    '''check whether pandas is at least min_version '''
 138 +    return StrictVersion((pandas.__version__[:6])) >= min_version
 139
 140
 141  # local copies, all unchanged
 142
 143 From c894c3f4882d570efb517950069d83afa9794db8 Mon Sep 17 00:00:00 2001
 144 From: Ralf Gommers <ralf.gommers@gmail.com>
 145 Date: Mon, 26 Oct 2015 20:47:51 +0100
 146 Subject: [PATCH 4/4] BUG: fix use of Series.sort_values for older pandas.
 147
 148 Some failing tests in the previous commits because older ``pandas`` versions
 149 don't have ``Series.sort_values``.  That method was only added in pandas 0.17,
 150 in https://github.com/pydata/pandas/pull/10726
 151 ---
 152  statsmodels/graphics/tests/test_mosaicplot.py | 6 +++++-
 153  statsmodels/graphics/tsaplots.py              | 6 +++++-
 154  2 files changed, 10 insertions(+), 2 deletions(-)
 155
 156 diff --git a/statsmodels/graphics/tests/test_mosaicplot.py b/statsmodels/graphics/tests/test_mosaicplot.py
 157 index e41020e..2a873e7 100644
 158 --- a/statsmodels/graphics/tests/test_mosaicplot.py
 159 +++ b/statsmodels/graphics/tests/test_mosaicplot.py
 160 @@ -113,7 +113,11 @@ def test_mosaic():
 161      # sort by the marriage quality and give meaningful name
 162      # [rate_marriage, age, yrs_married, children,
 163      # religious, educ, occupation, occupation_husb]
 164 -    datas = datas.sort_values(by=['rate_marriage', 'religious'])
 165 +    if pandas.__version__ < '0.17.0':
 166 +        datas = datas.sort(['rate_marriage', 'religious'])
 167 +    else:
 168 +        datas = datas.sort_values(by=['rate_marriage', 'religious'])
 169 +
 170      num_to_desc = {1: 'awful', 2: 'bad', 3: 'intermediate',
 171                        4: 'good', 5: 'wonderful'}
 172      datas['rate_marriage'] = datas['rate_marriage'].map(num_to_desc)
 173 diff --git a/statsmodels/graphics/tsaplots.py b/statsmodels/graphics/tsaplots.py
 174 index 94626c9..217724f 100644
 175 --- a/statsmodels/graphics/tsaplots.py
 176 +++ b/statsmodels/graphics/tsaplots.py
 177 @@ -2,6 +2,7 @@
 178
 179
 180  import numpy as np
 181 +import pandas
 182
 183  from statsmodels.graphics import utils
 184  from statsmodels.tsa.stattools import acf, pacf
 185 @@ -200,7 +201,10 @@ def seasonal_plot(grouped_x, xticklabels, ylabel=None, ax=None):
 186      ticks = []
 187      for season, df in grouped_x:
 188          df = df.copy() # or sort balks for series. may be better way
 189 -        df.sort_values(inplace=True)
 190 +        if pandas.__version__ < '0.17.0':
 191 +            df.sort()
 192 +        else:
 193 +            df.sort_values(inplace=True)
 194          nobs = len(df)
 195          x_plot = np.arange(start, start + nobs)
 196          ticks.append(x_plot.mean())