def generate_lazy_op_list(arrays):
'''
Slightly hacky way to match the group operator syntax.
'''
ret = []
for a in arrays:
lazy_arr = LazyOpResult(a.weldobj, a._weld_type, 1)
ret.append(lazy_arr)
return ret
def values(self):
if self.predicates is None:
return self.df.values
else:
if isinstance(self.df.values, np.ndarray):
weld_type = grizzly_impl.numpy_to_weld_type_mapping[str(
self.df.values.dtype)]
dim = self.df.values.ndim
return LazyOpResult(
grizzly_impl.filter(self.df.values, self.predicates.expr,
weld_type), weld_type, dim)
def get_column(self, column_name, column_type, index, verbose=True):
"""Summary
Args:
column_name (TYPE): Description
column_type (TYPE): Description
index (TYPE): Description
Returns:
TYPE: Description
"""
return LazyOpResult(
grizzly_impl.get_column(self.expr, self.weld_type, index),
column_type, 1).evaluate(verbose=verbose)
def std(self):
"""Standard deviation
Note that is by default normalizd by n - 1
# TODO, what does pandas do for multiple grouping columns?
# Currently we are just going to use one grouping column
"""
std_expr = grizzly_impl.groupby_std([self.column], [self.column_type],
self.grouping_columns,
self.grouping_column_types)
unzipped_columns = grizzly_impl.unzip_columns(
std_expr,
self.grouping_column_types + [WeldDouble()],
)
index_expr = LazyOpResult(grizzly_impl.get_field(unzipped_columns, 0),
self.grouping_column_types[0], 1)
column_expr = LazyOpResult(grizzly_impl.get_field(unzipped_columns, 1),
self.grouping_column_types[0], 1)
group_expr = utils.group([index_expr, column_expr])
return SeriesWeld(group_expr.expr,
WeldDouble(),
index_type=self.grouping_column_types[0],
index_name=self.grouping_column_names[0])
def __sub__(self, other):
# TODO subtractionw without index variables
if self.index_type is not None:
index = grizzly_impl.get_field(self.expr, 0)
expr1 = grizzly_impl.get_field(self.expr, 1)
else:
expr1 = self.expr
if other.index_type is not None:
index2 = grizzly_impl.get_field(other.expr, 0)
expr2 = grizzly_impl.get_field(other.expr, 1)
else:
expr2 = other.expr
index_expr = LazyOpResult(index, self.index_type, 0)
sub_expr = SeriesWeld(
grizzly_impl.element_wise_op(expr1, expr2, "-", self.weld_type),
self.weld_type, self.df, self.column_name)
index_sub_expr = utils.group([index_expr, sub_expr])
return SeriesWeld(index_sub_expr.expr, self.weld_type, self.df,
self.column_name, self.index_type, self.index_name)
def generate_lazy_op_list(arrays):
ret = []
for a in arrays:
lazy_arr = LazyOpResult(a.weldobj, a._weld_type, 1)
ret.append(lazy_arr)
return ret