Merge pull request #96 from njsmith/pandas-is-categorical-dtype-moved

Adapt to changes in pandas 0.19.0

Author: njsmith

.travis.yml

@@ -12,6 +12,12 @@ matrix:
       env: PANDAS_VERSION_STR="=0.14.0 libgfortran=1.0"
     - python: 2.7
       env: PANDAS_VERSION_STR="=0.14.0 libgfortran=1.0"
+    # 0.18.0 has is_categorical_dtype in a different place than 0.19.0+
+    - python: 3.4
+      env: PANDAS_VERSION_STR="=0.18.0"
+    - python: 2.7
+      env: PANDAS_VERSION_STR="=0.18.0"
+
 # This disables sudo, but makes builds start much faster
 # See http://blog.travis-ci.com/2014-12-17-faster-builds-with-container-based-infrastructure/
 sudo: false

patsy/categorical.py

@@ -134,7 +134,7 @@ def guess_categorical(data):
 
 def test_guess_categorical():
     if have_pandas_categorical:
-        c = pandas.Categorical.from_array([1, 2, 3])
+        c = pandas.Categorical([1, 2, 3])
         assert guess_categorical(c)
         if have_pandas_categorical_dtype:
             assert guess_categorical(pandas.Series(c))
@@ -242,15 +242,15 @@ def t(NA_types, datas, exp_finish_fast, exp_levels, exp_contrast=None):
             preps += [pandas.Series,
                       lambda x: C(pandas.Series(x))]
         for prep in preps:
-            t([], [prep(pandas.Categorical.from_array([1, 2, None]))],
+            t([], [prep(pandas.Categorical([1, 2, None]))],
               True, (1, 2))
             # check order preservation
             t([], [prep(pandas_Categorical_from_codes([1, 0], ["a", "b"]))],
               True, ("a", "b"))
             t([], [prep(pandas_Categorical_from_codes([1, 0], ["b", "a"]))],
               True, ("b", "a"))
             # check that if someone sticks a .contrast field onto our object
-            obj = prep(pandas.Categorical.from_array(["a", "b"]))
+            obj = prep(pandas.Categorical(["a", "b"]))
             obj.contrast = "CONTRAST"
             t([], [obj], True, ("a", "b"), "CONTRAST")
 

patsy/test_build.py

@@ -429,7 +429,7 @@ def __getitem__(self, key):
                                       return_type="dataframe")
     assert isinstance(int_df, pandas.DataFrame)
     assert np.array_equal(int_df, [[1], [1], [1]])
-    assert int_df.index.equals([10, 20, 30])
+    assert int_df.index.equals(pandas.Index([10, 20, 30]))
 
     import patsy.build
     had_pandas = patsy.build.have_pandas
@@ -448,7 +448,7 @@ def __getitem__(self, key):
                                                    dtype=object)},
                                   NA_action="drop",
                                   return_type="dataframe")
-    assert x_df.index.equals([2])
+    assert x_df.index.equals(pandas.Index([2]))
 
 def test_data_mismatch():
     test_cases_twoway = [
@@ -580,7 +580,7 @@ def test_categorical():
     data_categ = {"a": C(["a2", "a1", "a2"])}
     datas = [data_strings, data_categ]
     if have_pandas_categorical:
-        data_pandas = {"a": pandas.Categorical.from_array(["a1", "a2", "a2"])}
+        data_pandas = {"a": pandas.Categorical(["a1", "a2", "a2"])}
         datas.append(data_pandas)
     def t(data1, data2):
         def iter_maker():

patsy/test_highlevel.py

@@ -672,7 +672,7 @@ def test_dmatrix_NA_action():
         assert np.array_equal(mat, [[1, 2, 20],
                                     [1, 3, 30]])
         if return_type == "dataframe":
-            assert mat.index.equals([1, 2])
+            assert mat.index.equals(pandas.Index([1, 2]))
         assert_raises(PatsyError, dmatrix, "x + y", data=data,
                       return_type=return_type,
                       NA_action="raise")
@@ -681,8 +681,8 @@ def test_dmatrix_NA_action():
         assert np.array_equal(lmat, [[20], [30]])
         assert np.array_equal(rmat, [[1, 2], [1, 3]])
         if return_type == "dataframe":
-            assert lmat.index.equals([1, 2])
-            assert rmat.index.equals([1, 2])
+            assert lmat.index.equals(pandas.Index([1, 2]))
+            assert rmat.index.equals(pandas.Index([1, 2]))
         assert_raises(PatsyError,
                       dmatrices, "y ~ x", data=data, return_type=return_type,
                       NA_action="raise")
@@ -693,8 +693,8 @@ def test_dmatrix_NA_action():
         assert np.array_equal(lmat, [[20], [30], [40]])
         assert np.array_equal(rmat, [[1], [1], [1]])
         if return_type == "dataframe":
-            assert lmat.index.equals([1, 2, 3])
-            assert rmat.index.equals([1, 2, 3])
+            assert lmat.index.equals(pandas.Index([1, 2, 3]))
+            assert rmat.index.equals(pandas.Index([1, 2, 3]))
         assert_raises(PatsyError,
                       dmatrices, "y ~ 1", data=data, return_type=return_type,
                       NA_action="raise")

patsy/util.py

@@ -40,9 +40,20 @@
 # Can drop this guard whenever we drop support for such older versions of
 # pandas.
 have_pandas_categorical = (have_pandas and hasattr(pandas, "Categorical"))
-have_pandas_categorical_dtype = (have_pandas
-                                  and hasattr(pandas.core.common,
-                                              "is_categorical_dtype"))
+if not have_pandas:
+    have_pandas_categorical_dtype = False
+    _pandas_is_categorical_dtype = None
+else:
+    if hasattr(pandas, "api"):
+        # This is available starting in pandas v0.19.0
+        have_pandas_categorical_dtype = True
+        _pandas_is_categorical_dtype = pandas.api.types.is_categorical_dtype
+    else:
+        # This is needed for pandas v0.18.0 and earlier
+        _pandas_is_categorical_dtype = getattr(pandas.core.common,
+                                               "is_categorical_dtype", None)
+        have_pandas_categorical_dtype = (_pandas_is_categorical_dtype
+                                         is not None)
 
 # Passes through Series and DataFrames, call np.asarray() on everything else
 def asarray_or_pandas(a, copy=False, dtype=None, subok=False):
@@ -637,10 +648,7 @@ def test_pandas_Categorical_accessors():
 def safe_is_pandas_categorical_dtype(dt):
     if not have_pandas_categorical_dtype:
         return False
-    # WTF this incredibly crucial function is not even publically exported.
-    # Also if you read its source it uses a bare except: block which is broken
-    # by definition, but oh well there is not much I can do about this.
-    return pandas.core.common.is_categorical_dtype(dt)
+    return _pandas_is_categorical_dtype(dt)
 
 # Needed to support pandas >= 0.15 (!)
 def safe_is_pandas_categorical(data):
@@ -656,7 +664,7 @@ def test_safe_is_pandas_categorical():
     assert not safe_is_pandas_categorical(np.arange(10))
 
     if have_pandas_categorical:
-        c_obj = pandas.Categorical.from_array(["a", "b"])
+        c_obj = pandas.Categorical(["a", "b"])
         assert safe_is_pandas_categorical(c_obj)
 
     if have_pandas_categorical_dtype: