def __propagate_column(dataset, parent_dataset):
"""Propagate columns in `parent_dataset` to `dataset`.
When a new calculation is added to a dataset this will propagate the
new column to all child (merged) datasets.
:param dataset: THe child dataet.
:param parent_dataset: The dataset to propagate.
"""
# delete the rows in this dataset from the parent
dataset.remove_parent_observations(parent_dataset.dataset_id)
# get this dataset without the out-of-date parent rows
dframe = dataset.dframe(keep_parent_ids=True)
# create new dframe from the upated parent and add parent id
parent_dframe = add_parent_column(parent_dataset.dframe(),
parent_dataset.dataset_id)
# merge this new dframe with the existing dframe
updated_dframe = concat([dframe, parent_dframe])
# save new dframe (updates schema)
dataset.replace_observations(updated_dframe)
dataset.clear_summary_stats()
# recur into merged dataset
[__propagate_column(x, dataset) for x in dataset.merged_datasets]
def calculate_updates(dataset, new_data, new_dframe_raw=None,
parent_dataset_id=None, update_id=None):
"""Update dataset with `new_data`.
This can result in race-conditions when:
- deleting ``controllers.Datasets.DELETE``
- updating ``controllers.Datasets.POST([dataset_id])``
Therefore, perform these actions asychronously.
:param new_data: Data to update this dataset with.
:param new_dframe_raw: DataFrame to update this dataset with.
:param parent_dataset_id: If passed add ID as parent ID to column,
default is None.
"""
__ensure_ready(dataset, update_id)
if new_dframe_raw is None:
new_dframe_raw = dframe_from_update(dataset, new_data)
new_dframe = recognize_dates(new_dframe_raw, dataset.schema)
new_dframe = __add_calculations(dataset, new_dframe)
# set parent id if provided
if parent_dataset_id:
new_dframe = add_parent_column(new_dframe, parent_dataset_id)
dataset.append_observations(new_dframe)
dataset.clear_summary_stats()
propagate(dataset, new_dframe=new_dframe, update={'add': new_dframe_raw})
dataset.update_complete(update_id)
def __merge_datasets(datasets, mapping):
"""Merge two or more datasets."""
dframes = []
if not mapping:
mapping = {}
for dataset in datasets:
dframe = dataset.dframe()
column_map = mapping.get(dataset.dataset_id)
if column_map:
dframe = dframe.rename(columns=column_map)
dframe = add_parent_column(dframe, dataset.dataset_id)
dframes.append(dframe)
return concat(dframes, ignore_index=True)
def __merge_datasets(datasets, mapping):
"""Merge two or more datasets."""
dframes = []
if not mapping:
mapping = {}
for dataset in datasets:
dframe = dataset.dframe()
column_map = mapping.get(dataset.dataset_id)
if column_map:
dframe = dframe.rename(columns=column_map)
dframe = add_parent_column(dframe, dataset.dataset_id)
dframes.append(dframe)
return concat(dframes, ignore_index=True)
def save(self, dataset):
"""Save this aggregation.
If an aggregated dataset for this aggregations group already exists
store in this dataset, if not create a new aggregated dataset and store
the aggregation in this new aggregated dataset.
"""
new_dframe = self.aggregation.eval(self.columns)
new_dframe = add_parent_column(new_dframe, dataset.dataset_id)
a_dataset = dataset.aggregated_dataset(self.groups)
if a_dataset is None:
a_dataset = aggregated_dataset(dataset, new_dframe, self.groups)
else:
a_dframe = a_dataset.dframe()
new_dframe = group_join(self.groups, a_dframe, new_dframe)
a_dataset.replace_observations(new_dframe)
self.new_dframe = new_dframe
def update(self, dataset, child_dataset, formula, reducible):
"""Attempt to reduce an update and store."""
parent_dataset_id = dataset.dataset_id
# get dframe only including rows from this parent
dframe = rows_for_parent_id(child_dataset.dframe(
keep_parent_ids=True, reload_=True), parent_dataset_id)
# remove rows in child from parent
child_dataset.remove_parent_observations(parent_dataset_id)
if reducible and self.__is_reducible():
dframe = self.aggregation.reduce(dframe, self.columns)
else:
dframe = self.updated_dframe(dataset, formula, dframe)
new_a_dframe = concat([child_dataset.dframe(), dframe])
new_a_dframe = add_parent_column(new_a_dframe, parent_dataset_id)
child_dataset.replace_observations(new_a_dframe)
return child_dataset.dframe()
def save(self, dataset):
"""Save this aggregation.
If an aggregated dataset for this aggregations group already exists
store in this dataset, if not create a new aggregated dataset and store
the aggregation in this new aggregated dataset.
"""
new_dframe = self.aggregation.eval(self.columns)
new_dframe = add_parent_column(new_dframe, dataset.dataset_id)
a_dataset = dataset.aggregated_dataset(self.groups)
if a_dataset is None:
a_dataset = aggregated_dataset(dataset, new_dframe, self.groups)
else:
a_dframe = a_dataset.dframe()
new_dframe = group_join(self.groups, a_dframe, new_dframe)
a_dataset.replace_observations(new_dframe)
self.new_dframe = new_dframe
def update(self, dataset, child_dataset, formula, reducible):
"""Attempt to reduce an update and store."""
parent_dataset_id = dataset.dataset_id
# get dframe only including rows from this parent
dframe = rows_for_parent_id(
child_dataset.dframe(keep_parent_ids=True, reload_=True),
parent_dataset_id)
# remove rows in child from parent
child_dataset.remove_parent_observations(parent_dataset_id)
if reducible and self.__is_reducible():
dframe = self.aggregation.reduce(dframe, self.columns)
else:
dframe = self.updated_dframe(dataset, formula, dframe)
new_a_dframe = concat([child_dataset.dframe(), dframe])
new_a_dframe = add_parent_column(new_a_dframe, parent_dataset_id)
child_dataset.replace_observations(new_a_dframe)
return child_dataset.dframe()
def calculate_updates(dataset, new_data, new_dframe_raw=None,
parent_dataset_id=None, update_id=None):
"""Update dataset with `new_data`.
This can result in race-conditions when:
- deleting ``controllers.Datasets.DELETE``
- updating ``controllers.Datasets.POST([dataset_id])``
Therefore, perform these actions asychronously.
:param new_data: Data to update this dataset with.
:param new_dframe_raw: DataFrame to update this dataset with.
:param parent_dataset_id: If passed add ID as parent ID to column,
default is None.
"""
if not __update_is_valid(dataset, new_dframe_raw):
dataset.remove_pending_update(update_id)
return
__ensure_ready(dataset, update_id)
if new_dframe_raw is None:
new_dframe_raw = dframe_from_update(dataset, new_data)
new_dframe = recognize_dates(new_dframe_raw, dataset.schema)
new_dframe = __add_calculations(dataset, new_dframe)
# set parent id if provided
if parent_dataset_id:
new_dframe = add_parent_column(new_dframe, parent_dataset_id)
dataset.append_observations(new_dframe)
dataset.clear_summary_stats()
propagate(dataset, new_dframe=new_dframe, update={'add': new_dframe_raw})
dataset.update_complete(update_id)
