def load_by_guid(guid: str, conn: Optional[ConnectionPlus] = None) -> DataSet:
"""
Load a dataset by its GUID
If no connection is provided, lookup is performed in the database file that
is specified in the config.
Args:
guid: guid of the dataset
conn: connection to the database to load from
Returns:
dataset with the given guid
Raises:
NameError: if no run with the given GUID exists in the database
RuntimeError: if several runs with the given GUID are found
"""
conn = conn or connect(get_DB_location())
# this function raises a RuntimeError if more than one run matches the GUID
run_id = get_runid_from_guid(conn, guid)
if run_id == -1:
raise NameError(f'No run with GUID: {guid} found in database.')
return DataSet(run_id=run_id, conn=conn)
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 _dataset_is_in_runs_table(self,
path_to_db: Optional[Union[str, Path]] = None
) -> bool:
"""
Does this run exist in the given db
"""
if isinstance(path_to_db, Path):
path_to_db = str(path_to_db)
with contextlib.closing(
conn_from_dbpath_or_conn(conn=None,
path_to_db=path_to_db)) as conn:
run_id = get_runid_from_guid(conn, self.guid)
return run_id is not None
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 is not None:
return
target_table_name = _add_run_to_runs_table(dataset, target_conn, target_exp_id)
assert target_table_name is not None
_populate_results_table(
source_conn, target_conn, dataset.table_name, target_table_name
)
def run_id(self):
if not self._run_id:
self._run_id = get_runid_from_guid(self._conn, self._guid)
return self._run_id