def new_data_set(name,
exp_id: Optional[int] = None,
specs: SPECS = None,
values=None,
metadata=None,
conn=None) -> DataSet:
""" Create a new dataset.
If exp_id is not specified the last experiment will be loaded by default.
Args:
name: the name of the new dataset
exp_id: the id of the experiments this dataset belongs to
defaults to the last experiment
specs: list of parameters to create this data_set with
values: the values to associate with the parameters
metadata: the values to associate with the dataset
"""
path_to_db = get_DB_location()
d = DataSet(path_to_db, conn=conn)
if exp_id is None:
if len(get_experiments(d.conn)) > 0:
exp_id = get_last_experiment(d.conn)
else:
raise ValueError("No experiments found."
"You can start a new one with:"
" new_experiment(name, sample_name)")
d._new(name, exp_id, specs, values, metadata)
return d
def load_last_experiment() -> Experiment:
"""
Load last experiment
Returns:
last experiment
"""
conn = connect(get_DB_location())
return Experiment(exp_id=get_last_experiment(conn))
def load_last_experiment() -> Experiment:
"""
Load last experiment
Returns:
last experiment
"""
e = Experiment(get_DB_location())
e.exp_id = get_last_experiment(e.conn)
return e
def load_last_experiment() -> Experiment:
"""
Load last experiment
Returns:
last experiment
"""
e = Experiment(DB)
e.exp_id = get_last_experiment(e.conn)
return e
def load_last_experiment() -> Experiment:
"""
Load last experiment (from database file from config)
Returns:
last experiment
"""
last_exp_id = get_last_experiment(connect(get_DB_location()))
if last_exp_id is None:
raise ValueError('There are no experiments in the database file')
return Experiment(exp_id=last_exp_id)
def new_data_set(name,
exp_id: Optional[int] = None,
specs: SPECS = None,
values=None,
metadata=None,
conn=None) -> DataSet:
""" Create a new dataset.
If exp_id is not specified the last experiment will be loaded by default.
Args:
name: the name of the new dataset
exp_id: the id of the experiments this dataset belongs to
defaults to the last experiment
specs: list of parameters to create this data_set with
values: the values to associate with the parameters
metadata: the values to associate with the dataset
"""
path_to_db = get_DB_location()
if conn is None:
tempcon = True
conn = connect(get_DB_location())
else:
tempcon = False
if exp_id is None:
if len(get_experiments(conn)) > 0:
exp_id = get_last_experiment(conn)
else:
raise ValueError("No experiments found."
"You can start a new one with:"
" new_experiment(name, sample_name)")
# This is admittedly a bit weird. We create a dataset, link it to some
# run in the DB and then (using _new) change what it's linked to
if tempcon:
conn.close()
conn = None
d = DataSet(path_to_db, run_id=None, conn=conn)
d._new(name, exp_id, specs, values, metadata)
return d
def test_get_last_experiment_no_experiments(empty_temp_db):
conn = mut.connect(get_DB_location())
assert None is mut.get_last_experiment(conn)
def test_get_last_experiment(experiment):
assert experiment.exp_id == mut.get_last_experiment(experiment.conn)
def __init__(self,
path_to_db: str = None,
run_id: Optional[int] = None,
conn: Optional[ConnectionPlus] = None,
exp_id=None,
name: str = None,
specs: Optional[SpecsOrInterDeps] = None,
values=None,
metadata=None) -> None:
"""
Create a new DataSet object. The object can either hold a new run or
an already existing run. If a run_id is provided, then an old run is
looked up, else a new run is created.
Args:
path_to_db: path to the sqlite file on disk. If not provided, the
path will be read from the config.
run_id: provide this when loading an existing run, leave it
as None when creating a new run
conn: connection to the DB; if provided and `path_to_db` is
provided as well, then a ValueError is raised (this is to
prevent the possibility of providing a connection to a DB
file that is different from `path_to_db`)
exp_id: the id of the experiment in which to create a new run.
Ignored if run_id is provided.
name: the name of the dataset. Ignored if run_id is provided.
specs: paramspecs belonging to the dataset. Ignored if run_id is
provided.
values: values to insert into the dataset. Ignored if run_id is
provided.
metadata: metadata to insert into the dataset. Ignored if run_id
is provided.
"""
if path_to_db is not None and conn is not None:
raise ValueError("Both `path_to_db` and `conn` arguments have "
"been passed together with non-None values. "
"This is not allowed.")
self._path_to_db = path_to_db or get_DB_location()
self.conn = make_connection_plus_from(conn) if conn is not None else \
connect(self.path_to_db)
self._run_id = run_id
self._debug = False
self.subscribers: Dict[str, _Subscriber] = {}
self._interdeps: InterDependencies_
if run_id is not None:
if not run_exists(self.conn, run_id):
raise ValueError(f"Run with run_id {run_id} does not exist in "
f"the database")
self._completed = completed(self.conn, self.run_id)
run_desc = self._get_run_description_from_db()
if run_desc._old_style_deps:
# TODO: what if the old run had invalid interdep.s?
old_idps: InterDependencies = cast(InterDependencies,
run_desc.interdeps)
self._interdeps = old_to_new(old_idps)
else:
new_idps: InterDependencies_ = cast(InterDependencies_,
run_desc.interdeps)
self._interdeps = new_idps
self._metadata = get_metadata_from_run_id(self.conn, run_id)
self._started = self.run_timestamp_raw is not None
else:
# Actually perform all the side effects needed for the creation
# of a new dataset. Note that a dataset is created (in the DB)
# with no parameters; they are written to disk when the dataset
# is marked as started
if exp_id is None:
if len(get_experiments(self.conn)) > 0:
exp_id = get_last_experiment(self.conn)
else:
raise ValueError("No experiments found."
"You can start a new one with:"
" new_experiment(name, sample_name)")
name = name or "dataset"
_, run_id, __ = create_run(self.conn,
exp_id,
name,
generate_guid(),
parameters=None,
values=values,
metadata=metadata)
# this is really the UUID (an ever increasing count in the db)
self._run_id = run_id
self._completed = False
self._started = False
if isinstance(specs, InterDependencies_):
self._interdeps = specs
elif specs is not None:
self._interdeps = old_to_new(InterDependencies(*specs))
else:
self._interdeps = InterDependencies_()
self._metadata = get_metadata_from_run_id(self.conn, self.run_id)
def __init__(self, path_to_db: str=None,
run_id: Optional[int]=None,
conn=None,
exp_id=None,
name: str=None,
specs: SPECS=None,
values=None,
metadata=None) -> None:
"""
Create a new DataSet object. The object can either hold a new run or
an already existing run. If a run_id is provided, then an old run is
looked up, else a new run is created.
Args:
path_to_db: path to the sqlite file on disk. If not provided, the
path will be read from the config.
run_id: provide this when loading an existing run, leave it
as None when creating a new run
conn: connection to the DB
exp_id: the id of the experiment in which to create a new run.
Ignored if run_id is provided.
name: the name of the dataset. Ignored if run_id is provided.
specs: paramspecs belonging to the dataset. Ignored if run_id is
provided.
values: values to insert into the dataset. Ignored if run_id is
provided.
metadata: metadata to insert into the dataset. Ignored if run_id
is provided.
"""
# TODO: handle fail here by defaulting to
# a standard db
self.path_to_db = path_to_db or get_DB_location()
if conn is None:
self.conn = connect(self.path_to_db)
else:
self.conn = conn
self.run_id = run_id
self._debug = False
self.subscribers: Dict[str, _Subscriber] = {}
if run_id:
if not run_exists(self.conn, run_id):
raise ValueError(f"Run with run_id {run_id} does not exist in "
f"the database")
self._completed = completed(self.conn, self.run_id)
else:
if exp_id is None:
if len(get_experiments(self.conn)) > 0:
exp_id = get_last_experiment(self.conn)
else:
raise ValueError("No experiments found."
"You can start a new one with:"
" new_experiment(name, sample_name)")
# Actually perform all the side effects needed for
# the creation of a new dataset.
name = name or "dataset"
_, run_id, __ = create_run(self.conn, exp_id, name,
generate_guid(),
specs, values, metadata)
# this is really the UUID (an ever increasing count in the db)
self.run_id = run_id
self._completed = False
