def __getattr__(self, key):
"""Summary
Args:
key (TYPE): Description
Returns:
TYPE: Description
Raises:
Exception: Description
"""
if key == 'values':
if self.predicates is None:
return self.df.values
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
)
raise Exception("Attr %s does not exist" % key)
def filter(self, predicates):
if isinstance(predicates, SeriesWeld):
predicates = predicates.expr
return SeriesWeld(
grizzly_impl.filter(
self.expr,
predicates,
self.weld_type
),
self.weld_type,
self.df,
self.column_name
)
def __getitem__(self, key):
"""Summary
Args:
key (TYPE): Description
Returns:
TYPE: Description
Raises:
Exception: Description
"""
if isinstance(key, str): # Single-key get
# First check if key corresponds to an un-materialized column
if key in self.unmaterialized_cols:
return self.unmaterialized_cols[key]
raw_column = self.df[key].values
dtype = str(raw_column.dtype)
# If column type is "object", then cast as "vec[char]" in Weld
if dtype == 'object':
raw_column = self.raw_columns[key]
weld_type = WeldVec(WeldChar())
else:
weld_type = grizzly_impl.numpy_to_weld_type_mapping[dtype]
if self.predicates is None:
return SeriesWeld(raw_column, weld_type, self, key)
return SeriesWeld(
grizzly_impl.filter(
raw_column,
self.predicates.expr,
weld_type
),
weld_type,
self,
key
)
elif isinstance(key, list):
# For multi-key get, return type is a dataframe
return DataFrameWeld(self.df[key], self.predicates)
elif isinstance(key, SeriesWeld):
# Can also apply predicate to a dataframe
return DataFrameWeld(self.df, key)
raise Exception("Invalid type in __getitem__")
def __getitem__(self, key):
if isinstance(self.grizzly_obj, SeriesWeld):
series = self.grizzly_obj
if isinstance(key, SeriesWeld):
if series.index_type is not None:
index_expr = grizzly_impl.get_field(series.expr, 0)
column_expr = grizzly_impl.get_field(series.expr, 1)
zip_expr = grizzly_impl.zip_columns(
[index_expr, column_expr])
predicate_expr = grizzly_impl.isin(index_expr, key.expr,
series.index_type)
filtered_expr = grizzly_impl.filter(
zip_expr, predicate_expr)
unzip_expr = grizzly_impl.unzip_columns(
filtered_expr, [series.index_type, series.weld_type])
return SeriesWeld(unzip_expr, series.weld_type, series.df,
series.column_name, series.index_type,
series.index_name)
# TODO : Need to implement for non-pivot tables
raise Exception("Cannot invoke getitem on non SeriesWeld object")
def __getitem__(self, key):
"""Summary
Args:
key (TYPE): Description
Returns:
TYPE: Description
Raises:
Exception: Description
"""
if isinstance(key, str): # Single-key get
# First check if key corresponds to an un-materialized column
if key in self.unmaterialized_cols:
return self.unmaterialized_cols[key]
raw_column = self.df[key].values
dtype = str(raw_column.dtype)
# If column type is "object", then cast as "vec[char]" in Weld
if dtype == 'object':
raw_column = self.raw_columns[key]
weld_type = WeldVec(WeldChar())
else:
weld_type = grizzly_impl.numpy_to_weld_type_mapping[dtype]
if self.predicates is None:
return SeriesWeld(raw_column, weld_type, self, key)
return SeriesWeld(
grizzly_impl.filter(raw_column, self.predicates.expr,
weld_type), weld_type, self, key)
elif isinstance(key, list):
# For multi-key get, return type is a dataframe
return DataFrameWeld(self.df[key], self.predicates)
elif isinstance(key, SeriesWeld):
# Can also apply predicate to a dataframe
if self.predicates is not None:
return DataFrameWeld(self.df,
key.per_element_and(self.predicates))
return DataFrameWeld(self.df, key)
raise Exception("Invalid type in __getitem__")
def __getattr__(self, key):
"""Summary
Args:
key (TYPE): Description
Returns:
TYPE: Description
Raises:
Exception: Description
"""
if key == 'values':
if self.predicates is None:
return self.df.values
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)
raise Exception("Attr %s does not exist" % key)
def filter(self, predicates):
if isinstance(predicates, SeriesWeld):
predicates = predicates.expr
return SeriesWeld(
grizzly_impl.filter(self.expr, predicates, self.weld_type),
self.weld_type, self.df, self.column_name)
