def transform(self, data, **kwargs):
# TODO: make this a function and remove copies
levels = extract_level_names_dict(data)
cols_to_check = filter_levels(
levels,
self._columns
)
# if data has a multiindex
if data.columns.nlevels > 1 \
and not isinstance(cols_to_check[0], tuple):
if not isinstance(cols_to_check, dict):
levels[0] = cols_to_check
cols_to_check = levels
# check the tuple combinations of selected levels
cols_to_check = product(*cols_to_check.values())
i_to_keep = set()
n_index = np.arange(data.shape[0])
for col in cols_to_check:
valcount = data[col].value_counts()
vals_to_keep = valcount[valcount >= self._values].index
i_to_keep.update(n_index[data[col].isin(vals_to_keep)])
return data.copy().iloc[[*i_to_keep]]
def transform(self, data, **kwargs):
levels = extract_level_names_dict(data)
cols_to_check = filter_levels(
levels,
self._columns
)
# if data has a multiindex
if data.columns.nlevels > 1 \
and not isinstance(cols_to_check[0], tuple):
if not isinstance(cols_to_check, dict):
levels[0] = cols_to_check
cols_to_check = levels
# check the tuple combinations of selected levels
cols_to_check = product(*cols_to_check.values())
for col in cols_to_check:
# keep those values that are not (~) in self._values
data = data[~data[col].isin(self._values)]
return data
def transform(self, data, **kwargs):
"""Selects the specified columns or returns data as is if no column
was specified.
Returns:
Data of the same format as before but only only containing the
specified columns.
"""
cols = filter_levels(
extract_level_names_dict(data),
self._columns
)
return extract_cols(data, cols)
def transform(self, data, **kwargs):
inter_df = extract_cols(data,
filter_levels(
extract_level_names_dict(data),
self._columns
))
return data.copy()[
~inter_df.apply(
self._row_condition_builder(),
axis=1
)
]
def transform(self, data, **kwargs):
"""Apply custom transformation and insert back as specified
This applies the transformation in three main steps:
1. Extract specified columns
2. Apply modification
3. Insert columns if needed or return modified dataframe
These steps have further details for dealing with levels.
Raises:
RuntimeError: if transformed data is to be reintegrated but has a
different shape than data being reintegrated on the dataframe.
"""
cols = filter_levels(extract_level_names_dict(data), self._columns)
# inter_df should be a copied subset of the data
inter_df = extract_cols(data.copy(), cols)
orig_shape = inter_df.shape
inter_df = self._gen_cols(inter_df)
if isinstance(self._new_cols, str):
inter_df.columns = add_suffix(inter_df.columns, cols,
self._new_cols)
elif self._new_cols is not None:
inter_df.columns = self._new_cols
# No reintegration
if not self._reintegrate:
return inter_df
# Reintegration by joining
if self._new_cols is not None:
if self._drop:
data = drop_cols(data, cols)
# Join new columns into the data frame
return mi_join(data, inter_df)
# Reintegration by replacement
if inter_df.shape != orig_shape:
raise RuntimeError("Existing columns cannot be \
reintegrated if transformation changes data shape")
# Insert new columns on top of old ones.
# Copy is needed as original data is not intended to be changed.
return insert_cols(data.copy(), inter_df, cols)
def transform(self, data, **kwargs):
all_cols = extract_level_names_dict(data)
drop_cols = filter_levels(
all_cols,
self._columns
)
if not isinstance(drop_cols, dict):
drop_cols = {0: drop_cols}
cols = {level: list(set(ac).difference(dc))
for (level, ac), (level, dc)
in zip(all_cols.items(), drop_cols.items())}
return extract_cols(data, cols)
def transform(self, data, **kwargs):
all_cols = extract_level_names_dict(data)
for col, pos in self._map_dict.items():
all_cols[self._level].remove(col)
# this works for edge cases like last cols because of
# the remove
all_cols[self._level] = all_cols[self._level][:pos] \
+ [col] \
+ all_cols[self._level][pos:]
if data.columns.nlevels > 0:
all_cols = pd.MultiIndex.from_tuples(
[*product(*all_cols.values())]
)
else:
all_cols = pd.Index(all_cols[0])
inter_df = data.copy()
inter_df.columns = all_cols
return inter_df