BUG: Preserve categorical dtypes when melting (pandas-dev#15853)

Add support for tile and not simply repeat.

Author: dsm054

pandas/compat/numpy/function.py

@@ -211,6 +211,10 @@ def validate_cum_func_with_skipna(skipna, args, kwargs, name):
 validate_repeat = CompatValidator(REPEAT_DEFAULTS, fname='repeat',
                                   method='both', max_fname_arg_count=1)
 
+TILE_DEFAULTS = dict(axis=None)
+validate_tile = CompatValidator(TILE_DEFAULTS, fname='tile',
+                                method='both', max_fname_arg_count=1)
+
 ROUND_DEFAULTS = dict(out=None)
 validate_round = CompatValidator(ROUND_DEFAULTS, fname='round',
                                  method='both', max_fname_arg_count=1)

pandas/core/arrays/categorical.py

@@ -2325,12 +2325,25 @@ def repeat(self, repeats, *args, **kwargs):
         See also
         --------
         numpy.ndarray.repeat
-
+        Categorical.tile
         """
         nv.validate_repeat(args, kwargs)
         codes = self._codes.repeat(repeats)
         return self._constructor(values=codes, dtype=self.dtype, fastpath=True)
 
+    def tile(self, reps, *args, **kwargs):
+        """
+        Tile elements of a Categorical.
+
+        See also
+        --------
+        numpy.tile
+        Categorical.repeat
+        """
+        nv.validate_tile(args, kwargs)
+        codes = np.tile(self._codes, reps)
+        return self._constructor(values=codes, dtype=self.dtype, fastpath=True)
+
     # Implement the ExtensionArray interface
     @property
     def _can_hold_na(self):

pandas/core/indexes/base.py

@@ -546,6 +546,7 @@ def _simple_new(cls, values, name=None, dtype=None, **kwargs):
     def _shallow_copy(self, values=None, **kwargs):
         if values is None:
             values = self.values
+
         attributes = self._get_attributes_dict()
         attributes.update(kwargs)
         if not len(values) and 'dtype' not in kwargs:
@@ -557,7 +558,6 @@ def _shallow_copy(self, values=None, **kwargs):
             # `self.values` returns `self` for tz-aware, so we need to unwrap
             #  more specifically
             values = values.asi8
-
         return self._simple_new(values, **attributes)
 
     def _shallow_copy_with_infer(self, values, **kwargs):
@@ -822,6 +822,7 @@ def repeat(self, repeats, *args, **kwargs):
         --------
         Series.repeat : Equivalent function for Series
         numpy.repeat : Underlying implementation
+        Series.tile : repeat the entire index as a group, not by element
 
         Examples
         --------
@@ -836,6 +837,47 @@ def repeat(self, repeats, *args, **kwargs):
         nv.validate_repeat(args, kwargs)
         return self._shallow_copy(self._values.repeat(repeats))
 
+    def tile(self, reps, *args, **kwargs):
+        """
+        Tile elements of an Index.
+
+        Returns a new index constructed by repeating the current index
+        the number of times given by reps.
+
+        .. versionadded:: 0.24.0
+
+        Parameters
+        ----------
+        reps : int
+            The number of repetitions of the element groups.
+        **kwargs
+            Additional keywords have no effect but might be accepted for
+            compatibility with numpy.
+
+        Returns
+        -------
+        pandas.Index
+            Newly created Index with tiled elements.
+
+        See Also
+        --------
+        Series.tile : Equivalent function for Series
+        numpy.tile : Underlying implementation
+        Series.repeat : repeat the index element by element, not as a group
+
+        Examples
+        --------
+        >>> idx = pd.Index([1, 2, 3])
+        >>> idx
+        Int64Index([1, 2, 3], dtype='int64')
+        >>> idx.tile(2)
+        Int64Index([1, 2, 3, 1, 2, 3], dtype='int64')
+        >>> idx.tile(3)
+        Int64Index([1, 2, 3, 1, 2, 3, 1, 2, 3], dtype='int64')
+        """
+        nv.validate_tile(args, kwargs)
+        return self._shallow_copy(np.tile(self._values[:], reps))
+
     _index_shared_docs['where'] = """
         .. versionadded:: 0.19.0
 

pandas/core/indexes/datetimelike.py

@@ -638,6 +638,18 @@ def repeat(self, repeats, *args, **kwargs):
         return self._shallow_copy(self.asi8.repeat(repeats),
                                   freq=freq)
 
+    def tile(self, reps, *args, **kwargs):
+        """
+        Analogous to numpy.tile
+        """
+        nv.validate_tile(args, kwargs)
+        if is_period_dtype(self):
+            freq = self.freq
+        else:
+            freq = None
+        return self._shallow_copy(np.tile(self.asi8, reps),
+                                  freq=freq)
+
     @Appender(_index_shared_docs['where'] % _index_doc_kwargs)
     def where(self, cond, other=None):
         other = _ensure_datetimelike_to_i8(other, to_utc=True)

pandas/core/series.py

@@ -989,11 +989,12 @@ def _set_values(self, key, value):
 
     def repeat(self, repeats, *args, **kwargs):
         """
-        Repeat elements of an Series. Refer to `numpy.ndarray.repeat`
+        Repeat elements of a Series. Refer to `numpy.ndarray.repeat`
         for more information about the `repeats` argument.
 
         See also
         --------
+        pd.Series.tile
         numpy.ndarray.repeat
         """
         nv.validate_repeat(args, kwargs)
@@ -1002,6 +1003,26 @@ def repeat(self, repeats, *args, **kwargs):
         return self._constructor(new_values,
                                  index=new_index).__finalize__(self)
 
+    def tile(self, reps, *args, **kwargs):
+        """
+        Tile elements of a Series. Refer to `numpy.tile`
+        for more information about the `reps` argument.
+
+        See also
+        --------
+        pd.Series.repeat
+        numpy.tile
+        """
+        nv.validate_tile(args, kwargs)
+        new_index = self.index.tile(reps)
+        if is_categorical_dtype(self.dtype):
+            return Categorical.from_codes(np.tile(self.cat.codes, reps),
+                                          categories=self.cat.categories,
+                                          ordered=self.cat.ordered)
+        new_values = np.tile(self._values, reps)
+        return self._constructor(new_values,
+                                 index=new_index).__finalize__(self)
+
     def get_value(self, label, takeable=False):
         """Quickly retrieve single value at passed index label