Merge pull request #3 from rapidsai/branch-0.9

New changes

CHANGELOG.md

@@ -76,6 +76,7 @@
 - PR #2406 Moved all existing `table` related files to a `legacy/` directory
 - PR #2350 Performance related changes to get_dummies
 - PR #2420 Remove `cudautils.astype` and replace with `typecast.apply_cast`
+- PR #2456 Small improvement to typecast utility
 - PR #2458 Fix handling of thirdparty packages in `isort` config
 
 ## Bug Fixes

python/cudf/cudf/bindings/copying.pyx

@@ -54,10 +54,9 @@ def apply_gather(in_cols, maps, out_cols=None):
     else:
         in_size = in_cols[0].data.size
 
-    import cudf.bindings.typecast as typecast
     from cudf.dataframe import columnops
-    col = typecast.apply_cast(columnops.as_column(maps), dtype=np.int32)
-    maps = col.data.mem
+    maps = columnops.as_column(maps).astype("int32")
+    maps = maps.data.mem
     # TODO: replace with libcudf pymod when available
     maps = modulo(maps, in_size)
 

python/cudf/cudf/bindings/sort.pyx

@@ -182,12 +182,8 @@ class SegmentedRadixSortPlan(object):
 
         # Note: .astype is required below because .copy_to_device
         #       is just a plain memcpy
-        import cudf.bindings.typecast as typecast
         from cudf.dataframe import columnops
-        col = typecast.apply_cast(
-            columnops.as_column(segments),
-            dtype=seg_dtype
-        )
+        col = columnops.as_column(segments).astype(seg_dtype)
         d_begins.copy_to_device(col.data.mem)
         d_ends[-1:].copy_to_device(np.require([self.nelem], dtype=seg_dtype))
 

python/cudf/cudf/bindings/typecast.pyx

@@ -12,43 +12,43 @@ from cudf.dataframe.column import Column
 from libc.stdlib cimport free
 
 import numpy as np
-
+import pandas as pd
 
 _time_unit = {
-    'none': TIME_UNIT_NONE,
+    None: TIME_UNIT_NONE,
     's': TIME_UNIT_s,
     'ms': TIME_UNIT_ms,
     'us': TIME_UNIT_us,
     'ns': TIME_UNIT_ns,
 }
 
 
-def apply_cast(incol, **kwargs):
+def apply_cast(incol, dtype="float64", time_unit=None):
     """
-      Cast from incol.dtype to outcol.dtype
+    Return a Column with values in `incol` casted to `dtype`.
+    Currently supports numeric and datetime dtypes.
     """
 
     check_gdf_compatibility(incol)
+    dtype = pd.api.types.pandas_dtype(dtype).type
 
     cdef gdf_column* c_incol = column_view_from_column(incol)
 
-    npdtype = kwargs.get("dtype", np.float64)
-    cdef gdf_dtype dtype = dtypes[npdtype]
-    cdef uintptr_t category
+    cdef gdf_dtype c_dtype = dtypes[dtype]
+    cdef uintptr_t c_category
 
     cdef gdf_dtype_extra_info info = gdf_dtype_extra_info(
         time_unit=TIME_UNIT_NONE,
-        category=<void*>category
+        category=<void*>c_category
     )
-    unit = kwargs.get("time_unit", 'none')
-    info.time_unit = _time_unit[unit]
+    info.time_unit = _time_unit[time_unit]
 
     cdef gdf_column result
 
     with nogil:
         result = cast(
             c_incol[0],
-            dtype,
+            c_dtype,
             info
         )
 

python/cudf/cudf/dataframe/buffer.py

@@ -126,14 +126,13 @@ def append(self, element):
         self.extend(np.asarray(element, dtype=self.dtype))
 
     def extend(self, array):
+        from cudf.dataframe import columnops
+
         needed = array.size
         self._sentry_capacity(needed)
-        import cudf.bindings.typecast as typecast
-        from cudf.dataframe import columnops
 
-        array = typecast.apply_cast(
-            columnops.as_column(array), dtype=self.dtype.type
-        ).data.mem
+        array = columnops.as_column(array).astype(self.dtype).data.mem
+
         self.mem[self.size : self.size + needed].copy_to_device(array)
         self.size += needed
 

python/cudf/cudf/dataframe/datetime.py

@@ -123,20 +123,16 @@ def as_numerical(self):
         return self.view(
             numerical.NumericalColumn,
             dtype="int64",
-            data=typecast.apply_cast(self, dtype=np.int64).data,
+            data=typecast.apply_cast(self, np.int64).data,
         )
 
     def as_datetime_column(self, dtype, **kwargs):
         import cudf.bindings.typecast as typecast
 
-        return typecast.apply_cast(self, dtype=np.dtype(dtype).type)
+        return typecast.apply_cast(self, dtype=dtype)
 
     def as_numerical_column(self, dtype, **kwargs):
-        import cudf.bindings.typecast as typecast
-
-        return typecast.apply_cast(
-            self.as_numerical, dtype=np.dtype(dtype).type
-        )
+        return self.as_numerical.astype(dtype)
 
     def as_string_column(self, dtype, **kwargs):
         from cudf.dataframe import string

python/cudf/cudf/dataframe/numerical.py

@@ -142,15 +142,14 @@ def as_datetime_column(self, dtype, **kwargs):
         return self.view(
             datetime.DatetimeColumn,
             dtype=dtype,
-            data=typecast.apply_cast(self, dtype=np.dtype(dtype).type).data,
+            data=typecast.apply_cast(self, dtype=dtype).data,
         )
 
     def as_numerical_column(self, dtype, **kwargs):
         import cudf.bindings.typecast as typecast
 
         return self.replace(
-            data=typecast.apply_cast(self, dtype=np.dtype(dtype).type).data,
-            dtype=np.dtype(dtype),
+            data=typecast.apply_cast(self, dtype).data, dtype=np.dtype(dtype)
         )
 
     def sort_by_values(self, ascending=True, na_position="last"):