COMPAT: avoid calling getsizeof() on PyPy

Author: mattip

pandas/core/base.py

@@ -23,6 +23,8 @@
 _indexops_doc_kwargs = dict(klass='IndexOpsMixin', inplace='',
                             unique='IndexOpsMixin', duplicated='IndexOpsMixin')
 
+import platform
+IS_PYPY = platform.python_implementation() == 'PyPy'
 
 class StringMixin(object):
     """implements string methods so long as object defines a `__unicode__`
@@ -1061,7 +1063,7 @@ def memory_usage(self, deep=False):
         Notes
         -----
         Memory usage does not include memory consumed by elements that
-        are not components of the array if deep=False
+        are not components of the array if deep=False or if used on PyPy
 
         See Also
         --------
@@ -1071,9 +1073,8 @@ def memory_usage(self, deep=False):
             return self.values.memory_usage(deep=deep)
 
         v = self.values.nbytes
-        if deep and is_object_dtype(self):
+        if deep and is_object_dtype(self) and not IS_PYPY:
             v += lib.memory_usage_of_objects(self.values)
-
         return v
 
     def factorize(self, sort=False, na_sentinel=-1):

pandas/core/indexes/multi.py

@@ -467,7 +467,8 @@ def _nbytes(self, deep=False):
         """
         level_nbytes = sum((i.memory_usage(deep=deep) for i in self.levels))
         label_nbytes = sum((i.nbytes for i in self.labels))
-        names_nbytes = sum((getsizeof(i) for i in self.names))
+        objsize = 24  # for inplementations with no meaningfule getsizeof (PyPy)
+        names_nbytes = sum((getsizeof(i, 24) for i in self.names))
         result = level_nbytes + label_nbytes + names_nbytes
 
         # include our engine hashtable

pandas/core/indexes/range.py

@@ -194,8 +194,11 @@ def _format_data(self):
 
     @cache_readonly
     def nbytes(self):
-        """ return the number of bytes in the underlying data """
-        return sum([getsizeof(getattr(self, v)) for v in
+        """ return the number of bytes in the underlying data
+            On implementations where this is problematic (PyPy)
+            assume 24 bytes for each value
+        """
+        return sum([getsizeof(getattr(self, v), 24) for v in
                     ['_start', '_stop', '_step']])
 
     def memory_usage(self, deep=False):

pandas/tests/frame/test_repr_info.py

@@ -332,13 +332,15 @@ def test_info_memory_usage(self):
         res = buf.getvalue().splitlines()
         assert re.match(r"memory usage: [^+]+$", res[-1])
 
-        assert (df_with_object_index.memory_usage(
-            index=True, deep=True).sum() > df_with_object_index.memory_usage(
-            index=True).sum())
+        if not tm.IS_PYPY:
+            assert (df_with_object_index.memory_usage(
+                index=True, 
+                deep=True).sum() > df_with_object_index.memory_usage(
+                index=True).sum())
 
-        df_object = pd.DataFrame({'a': ['a']})
-        assert (df_object.memory_usage(deep=True).sum() >
-                df_object.memory_usage().sum())
+            df_object = pd.DataFrame({'a': ['a']})
+            assert (df_object.memory_usage(deep=True).sum() >
+                    df_object.memory_usage().sum())
 
         # Test a DataFrame with duplicate columns
         dtypes = ['int64', 'int64', 'int64', 'float64']
@@ -377,10 +379,11 @@ def test_info_memory_usage(self):
         df.memory_usage(index=True)
         df.index.values.nbytes
 
-        # sys.getsizeof will call the .memory_usage with
-        # deep=True, and add on some GC overhead
-        diff = df.memory_usage(deep=True).sum() - sys.getsizeof(df)
-        assert abs(diff) < 100
+        if not tm.IS_PYPY:
+            # sys.getsizeof will call the .memory_usage with
+            # deep=True, and add on some GC overhead
+            diff = df.memory_usage(deep=True).sum() - sys.getsizeof(df)
+            assert abs(diff) < 100
 
     def test_info_memory_usage_qualified(self):
 

pandas/tests/test_base.py

@@ -144,6 +144,7 @@ def f():
 
         pytest.raises(TypeError, f)
 
+    @pytest.mark.skipif(tm.IS_PYPY, "not relevant for PyPy")
     def test_memory_usage(self):
         # Delegate does not implement memory_usage.
         # Check that we fall back to in-built `__sizeof__`
@@ -941,6 +942,7 @@ def test_fillna(self):
                 # check shallow_copied
                 assert o is not result
 
+    @pytest.mark.skipif(tm.IS_PYPY, "not relevant for PyPy")
     def test_memory_usage(self):
         for o in self.objs:
             res = o.memory_usage()

pandas/tests/test_categorical.py

@@ -1448,10 +1448,11 @@ def test_memory_usage(self):
         cat = pd.Categorical(['foo', 'foo', 'bar'])
         assert cat.memory_usage(deep=True) > cat.nbytes
 
-        # sys.getsizeof will call the .memory_usage with
-        # deep=True, and add on some GC overhead
-        diff = cat.memory_usage(deep=True) - sys.getsizeof(cat)
-        assert abs(diff) < 100
+        if not tm.IS_PYPY:
+            # sys.getsizeof will call the .memory_usage with
+            # deep=True, and add on some GC overhead
+            diff = cat.memory_usage(deep=True) - sys.getsizeof(cat)
+            assert abs(diff) < 100
 
     def test_searchsorted(self):
         # https://github.com/pandas-dev/pandas/issues/8420

pandas/util/testing.py

@@ -56,6 +56,8 @@
 K = 4
 _RAISE_NETWORK_ERROR_DEFAULT = False
 
+import platform
+IS_PYPY = platform.python_implementation() == 'PyPy'
 
 # set testing_mode
 _testing_mode_warnings = (DeprecationWarning, compat.ResourceWarning)