DataFrameWrapper masquerades a dataframe as an ndarray just well enough that we can run our test suite

gdementen · gdementen · commit fa73cbb72f35 · 2015-01-30T16:56:56.000+01:00
not very interesting yet because the DF is still assumed to be dense and we pass it through asarray before indexing it
diff --git a/larray/core.py b/larray/core.py
@@ -1365,7 +1365,7 @@ def opmethod(self, other):
             elif not np.isscalar(other):
                 raise TypeError("unsupported operand type(s) for %s: '%s' "
                                 "and '%s'" % (opname, type(self), type(other)))
-            return LArray(super_method(self.data, other), self.axes)
+            return LArray(super_method(np.asarray(self), other), self.axes)
         opmethod.__name__ = fullname
         return opmethod
 
@@ -1411,7 +1411,7 @@ def _unaryop(opname):
         super_method = getattr(np.ndarray, fullname)
 
         def opmethod(self):
-            return LArray(super_method(self.data), self.axes)
+            return LArray(super_method(np.asarray(self)), self.axes)
         opmethod.__name__ = fullname
         return opmethod
 
@@ -1541,6 +1541,15 @@ def to_clipboard(self, *args, **kwargs):
     def plot(self, *args, **kwargs):
         self.df.plot(*args, **kwargs)
 
+    #XXX: one less indirection as we have all the info at this level?
+    # @property
+    # def shape(self):
+    #     return tuple(len(a) for a in self.axes)
+    #
+    # @property
+    # def ndim(self):
+    #     return len(self.axes)
+
     @property
     def shape(self):
         return self.data.shape
@@ -1565,7 +1574,7 @@ def __len__(self):
         return len(self.data)
 
     def __array__(self, dtype=None):
-        return self.data
+        return np.asarray(self.data)
 
     __array_priority__ = 100
 
@@ -1658,6 +1667,31 @@ def __getitem__(self, key):
     def __getattr__(self, key):
         return getattr(self.df, key)
 
+    @property
+    def dtype(self):
+        # assumes df is homogeneous !
+        return self.df.dtypes[0]
+
+    @property
+    def ndim(self):
+        return self.df.index.nlevels + 1
+
+    @property
+    def shape(self):
+        shape = [len(level) for level in self.df.index.levels]
+        shape.append(len(self.df.columns))
+        return tuple(shape)
+
+    def copy(self):
+        return DataFrameWrapper(self.df.copy())
+
+    # not caught by __getattr__?
+    def __len__(self):
+        return self.shape[0]
+
+    def __array__(self, dtype=None):
+        return self.df.__array__(dtype).reshape(self.shape)
+
 
 #TODO: implement sort_columns
 def df_aslarray2(df, sort_rows=True, sort_columns=True, **kwargs):
diff --git a/larray/tests/test_la.py b/larray/tests/test_la.py
@@ -10,7 +10,7 @@
 import larray
 from larray import (LArray, Axis, ValueGroup, union, to_ticks, to_key,
                     srange, larray_equal, read_csv, read_hdf, df_aslarray,
-                    zeros, zeros_like, AxisCollection)
+                    zeros, zeros_like, AxisCollection, DataFrameWrapper)
 from larray.utils import array_equal, array_nan_equal
 
 
@@ -509,11 +509,23 @@ def setUp(self):
 
         self.array = np.arange(116 * 44 * 2 * 15).reshape(116, 44, 2, 15) \
                                                  .astype(float)
-        self.larray = LArray(self.array,
-                             axes=(self.age, self.geo, self.sex, self.lipro))
+        idx = pd.MultiIndex.from_product([self.age.labels, self.geo.labels,
+                                          self.sex.labels])
+        dfarray = self.array.reshape(116 * 44 * 2, 15)
+        df = pd.DataFrame(dfarray, idx, columns=self.lipro.labels)
+        wrapped = DataFrameWrapper(df)
+        self.larray = LArray(wrapped, (self.age, self.geo, self.sex,
+                                       self.lipro))
+        # self.larray = LArray(self.array,
+        #                      axes=(self.age, self.geo, self.sex, self.lipro))
+        # self.larray = read_hdf('c:/tmp/y.h5', 'y', sort_rows=False)
 
         self.small_data = np.arange(30).reshape(2, 15)
-        self.small = LArray(self.small_data, axes=(self.sex, self.lipro))
+        df = pd.DataFrame(self.small_data, self.sex.labels,
+                          columns=self.lipro.labels)
+        self.small = LArray(DataFrameWrapper(df), (self.sex, self.lipro))
+        # self.small = LArray(self.small_data, axes=(self.sex, self.lipro))
+        # self.small = read_hdf('c:/tmp/x.h5', 'x', sort_rows=False)
 
     def test_zeros(self):
         la = zeros((self.geo, self.age))
