def _add_run_to_runs_table(
dataset: DataSetProtocol,
target_conn: ConnectionPlus,
target_exp_id: int,
create_run_table: bool = True,
) -> Optional[str]:
metadata = dataset.metadata
snapshot_raw = dataset._snapshot_raw
captured_run_id = dataset.captured_run_id
captured_counter = dataset.captured_counter
parent_dataset_links = links_to_str(dataset.parent_dataset_links)
_, target_run_id, target_table_name = create_run(
target_conn,
target_exp_id,
name=dataset.name,
guid=dataset.guid,
metadata=metadata,
captured_run_id=captured_run_id,
captured_counter=captured_counter,
parent_dataset_links=parent_dataset_links,
create_run_table=create_run_table,
snapshot_raw=snapshot_raw,
description=dataset.description,
)
mark_run_complete(target_conn, target_run_id)
_rewrite_timestamps(
target_conn,
target_run_id,
dataset.run_timestamp_raw,
dataset.completed_timestamp_raw,
)
return target_table_name
def _extract_single_dataset_into_db(dataset: DataSet,
target_conn: ConnectionPlus,
target_exp_id: int) -> None:
"""
NB: This function should only be called from within
:meth:extract_runs_into_db
Insert the given dataset into the specified database file as the latest
run.
Trying to insert a run already in the DB is a NOOP.
Args:
dataset: A dataset representing the run to be copied
target_conn: connection to the DB. Must be atomically guarded
target_exp_id: The exp_id of the (target DB) experiment in which to
insert the run
"""
if not dataset.completed:
raise ValueError('Dataset not completed. An incomplete dataset '
'can not be copied. The incomplete dataset has '
f'GUID: {dataset.guid} and run_id: {dataset.run_id}')
source_conn = dataset.conn
run_id = get_runid_from_guid(target_conn, dataset.guid)
if run_id != -1:
return
if dataset.parameters is not None:
param_names = dataset.parameters.split(',')
else:
param_names = []
parspecs_dict = {
p.name: p
for p in new_to_old(dataset._interdeps).paramspecs
}
parspecs = [parspecs_dict[p] for p in param_names]
metadata = dataset.metadata
snapshot_raw = dataset.snapshot_raw
_, target_run_id, target_table_name = create_run(target_conn,
target_exp_id,
name=dataset.name,
guid=dataset.guid,
parameters=parspecs,
metadata=metadata)
_populate_results_table(source_conn, target_conn, dataset.table_name,
target_table_name)
mark_run_complete(target_conn, target_run_id)
_rewrite_timestamps(target_conn, target_run_id, dataset.run_timestamp_raw,
dataset.completed_timestamp_raw)
if snapshot_raw is not None:
add_meta_data(target_conn, target_run_id, {'snapshot': snapshot_raw})
def _complete(self, value: bool) -> None:
with contextlib.closing(
conn_from_dbpath_or_conn(conn=None,
path_to_db=self._path_to_db)) as conn:
if value:
self._completed_timestamp_raw = time.time()
mark_run_complete(conn, self.run_id,
self._completed_timestamp_raw)
def test_mark_run_complete(dataset):
assert dataset.run_timestamp_raw is None
assert dataset.completed_timestamp_raw is None
time_now = time.time()
mut_queries.set_run_timestamp(dataset.conn, dataset.run_id)
time.sleep(1) # for slower test platforms
mut_queries.mark_run_complete(dataset.conn, dataset.run_id)
assert dataset.completed_timestamp_raw > time_now
def completed(self, value):
self._completed = value
if value:
mark_run_complete(self.conn, self.run_id)