split view fixes

Author: Ehsan Totoni

hpat/hiframes/boxing.py

@@ -234,6 +234,10 @@ def codegen(context, builder, sig, args):
 
         if data_typ == string_array_type:
             native_val = unbox_str_series(string_array_type, arr_obj, c)
+        elif data_typ == string_array_split_view_type:
+            # XXX dummy unboxing to avoid errors in _get_dataframe_data()
+            out_view = context.make_helper(builder, string_array_split_view_type)
+            native_val = NativeValue(out_view._getvalue())
         elif data_typ == list_string_array_type:
             native_val = _unbox_array_list_str(arr_obj, c)
         else:

hpat/hiframes/filter.py

@@ -9,6 +9,7 @@
 from hpat.distributed_analysis import Distribution
 from hpat.utils import debug_prints
 from hpat.str_arr_ext import string_array_type
+from hpat.hiframes.split_impl import string_array_split_view_type
 
 
 class Filter(ir.Stmt):
@@ -49,7 +50,8 @@ def filter_array_analysis(filter_node, equiv_set, typemap, array_analysis):
     for _, col_var in filter_node.df_in_vars.items():
         typ = typemap[col_var.name]
         # TODO handle list_string_array_type in other nodes
-        if typ in (string_array_type, list_string_array_type):
+        if typ in (string_array_type, list_string_array_type,
+                    string_array_split_view_type):
             continue
         col_shape = equiv_set.get_shape(col_var)
         all_shapes.append(col_shape[0])
@@ -63,7 +65,8 @@ def filter_array_analysis(filter_node, equiv_set, typemap, array_analysis):
     all_shapes = []
     for _, col_var in filter_node.df_out_vars.items():
         typ = typemap[col_var.name]
-        if typ in (string_array_type, list_string_array_type):
+        if typ in (string_array_type, list_string_array_type,
+                    string_array_split_view_type):
             continue
         (shape, c_post) = array_analysis._gen_shape_call(
             equiv_set, col_var, typ.ndim, None)

hpat/hiframes/hiframes_typed.py

@@ -2003,6 +2003,8 @@ def _handle_df_dropna(self, assign, lhs, rhs):
                                                     if is_str_arr_typ(t)]
         list_str_colnames = [in_names[i] for i, t in enumerate(in_typ.types)
                         if t == list_string_array_type]
+        split_view_colnames = [in_names[i] for i, t in enumerate(in_typ.types)
+                        if t == string_array_split_view_type]
         isna_calls = ['hpat.hiframes.api.isna({}, i)'.format(v) for v in in_names]
 
         func_text = "def _dropna_impl(arr_tup, inplace):\n"
@@ -2021,12 +2023,17 @@ def _handle_df_dropna(self, assign, lhs, rhs):
                 func_text += "  {} = hpat.str_arr_ext.pre_alloc_string_array(new_len, num_chars_{})\n".format(out, v)
             elif v in list_str_colnames:
                 func_text += "  {} = hpat.str_ext.alloc_list_list_str(new_len)\n".format(out)
+            elif v in split_view_colnames:
+                # TODO support dropna() for split view
+                func_text += "  {} = {}\n".format(out, v)
             else:
                 func_text += "  {} = np.empty(new_len, {}.dtype)\n".format(out, v)
         func_text += "  curr_ind = 0\n"
         func_text += "  for i in numba.parfor.internal_prange(old_len):\n"
         func_text += "    if not ({}):\n".format(' or '.join(isna_calls))
         for v, out in zip(in_names, out_names):
+            if v in split_view_colnames:
+                continue
             func_text += "      {}[curr_ind] = {}[i]\n".format(out, v)
         func_text += "      curr_ind += 1\n"
         func_text += "  return ({},)\n".format(", ".join(out_names))

hpat/hiframes/pd_series_ext.py

@@ -173,6 +173,9 @@ def arr_to_series_type(arr):
         series_type = SeriesType(string_type)
     elif arr == list_string_array_type:
         series_type = SeriesType(types.List(string_type))
+    elif arr == string_array_split_view_type:
+        series_type = SeriesType(types.List(string_type),
+                        string_array_split_view_type)
     return series_type
 
 
@@ -781,6 +784,7 @@ def generic(self, args, kws):
             else:
                 sig = GetItemStringArray.generic(self, (in_arr, in_idx), kws)
         elif in_arr == list_string_array_type:
+            # TODO: split view
             # mimic array indexing for list
             if (isinstance(in_idx, types.Array) and in_idx.ndim == 1
                     and isinstance(