def test_updated_file(session_db, tmpdir, caplog):
session, db = session_db
# Make tmpdir a concrete path otherwise filesystem ops won't work
tmpdir = Path(tmpdir)
ncfile = "test1.nc"
ncpath = Path("test/data/indexing/longnames/output000/") / ncfile
shutil.copy(str(ncpath), str(tmpdir / ncfile))
indexed = database.build_index(str(tmpdir), session)
assert indexed == 1
# Should not reindex
reindexed = database.build_index(str(tmpdir), session)
assert reindexed == 0
# Should reindex as file is updated
time.sleep(1)
(tmpdir / ncfile).touch()
reindexed = database.build_index(str(tmpdir), session)
assert reindexed == 1
# Should not reindex as flagging as missing will not remove
# file from the database, so will not be reindexed
time.sleep(1)
(tmpdir / ncfile).touch()
with caplog.at_level(logging.WARNING):
reindexed = database.build_index(str(tmpdir), session, prune="flag")
assert reindexed == 0
assert "Set prune to 'delete' to reindex updated files" in caplog.text
def test_delete_experiment(session_db):
"""Test that we can completely delete an experiment
and its associated data.
"""
session, db = session_db
database.build_index("test/data/indexing/longnames", session)
# make sure we actually did index something
expt = (session.query(database.NCExperiment).filter(
database.NCExperiment.experiment == "longnames").one_or_none())
assert expt is not None
database.delete_experiment("longnames", session)
expt = (session.query(database.NCExperiment).filter(
database.NCExperiment.experiment == "longnames").one_or_none())
assert expt is None
# check that all files are removed
files = session.query(sa.func.count(database.NCFile.id)).scalar()
assert files == 0
# make sure all ncvars are removed
vars = session.query(sa.func.count(database.NCVar.id)).scalar()
assert vars == 0
def test_multiple_experiments(session_db):
session, db = session_db
# index multiple experiments, which have duplicate data and therefore push
# against some unique constraints
database.build_index(['test/data/indexing/multiple/experiment_a', 'test/data/indexing/multiple/experiment_b'], session)
q = session.query(database.NCExperiment)
assert(q.count() == 2)
def test_reindex_noupdate(session_db):
session, db = session_db
database.build_index('test/data/indexing/broken_file', session)
assert(db.check())
# re-run the index, make sure we don't re-index anything
reindexed = database.build_index('test/data/indexing/broken_file', session)
assert(reindexed == 0)
def test_distributed(client, session_db):
session, db = session_db
database.build_index("test/data/indexing/broken_file", session, client)
assert db.check()
q = session.query(database.NCExperiment)
r = q.all()
assert len(r) == 1
def test_reindex_force(session_db):
session, db = session_db
database.build_index("test/data/indexing/broken_file", session)
assert db.check()
# re-run the index, make sure re-index
reindexed = database.build_index("test/data/indexing/broken_file",
session,
force=True)
assert reindexed == 1
def test_update_nonew(session_db):
session, db = session_db
database.build_index("test/data/indexing/broken_file", session)
assert db.check()
# re-run the index, make sure we don't re-index anything
reindexed = database.build_index("test/data/indexing/broken_file",
session,
prune="flag")
assert reindexed == 0
def test_update_newfile(session_db, tmpdir):
session, db = session_db
shutil.copy('test/data/indexing/longnames/output000/test1.nc',
str(tmpdir / 'test1.nc'))
database.build_index(str(tmpdir), session)
# add another file
shutil.copy('test/data/indexing/longnames/output000/test2.nc',
str(tmpdir / 'test2.nc'))
database.build_index(str(tmpdir), session, update=True)
def test_update_newfile(session_db, tmpdir):
session, db = session_db
shutil.copy("test/data/indexing/longnames/output000/test1.nc",
str(tmpdir / "test1.nc"))
database.build_index(str(tmpdir), session)
# add another file
shutil.copy("test/data/indexing/longnames/output000/test2.nc",
str(tmpdir / "test2.nc"))
reindexed = database.build_index(str(tmpdir), session)
assert reindexed == 1
def test_getvar_with_metadata(session_db):
session, db = session_db
database.build_index("test/data/indexing/metadata", session)
with querying.getvar("metadata", "test", session, decode_times=False) as v:
assert v.attrs["long_name"] == "Test Variable"
assert v.attrs["contact"] == "The ACCESS Oracle"
assert v.attrs["email"] == "*****@*****.**"
assert v.attrs["created"] == "2018-01-01"
assert "description" in v.attrs
def test_single_broken(session_db):
session, db = session_db
database.build_index("test/data/indexing/single_broken_file", session)
# query ncfiles table -- should have two entries
q = session.query(func.count(database.NCFile.id))
assert q.scalar() == 2
# query ncvars table -- should have a single entry
q = session.query(func.count(database.NCVar.id))
assert q.scalar() == 1
def test_metadata(session_db):
session, db = session_db
database.build_index('test/data/indexing/metadata', session)
# query metadata
q = session.query(database.NCExperiment.contact,
database.NCExperiment.created,
database.NCExperiment.description)
r = q.one()
assert(r[0] == 'The ACCESS Oracle')
assert(r[1] == datetime(2018, 1, 1))
assert(len(r[2]) > 0)
def test_longnames(session_db):
session, db = session_db
database.build_index("test/data/indexing/longnames", session)
# query ncvars table -- should have two entries
q = session.query(func.count(database.NCVar.id))
assert q.scalar() == 2
# query generic table -- should only be a single variable
q = session.query(database.CFVariable)
r = q.all()
assert len(r) == 1
assert r[0].long_name == "Test Variable"
def test_missing_time_bounds(session_db):
session, db = session_db
database.build_index("test/data/indexing/time_bounds", session)
# Should have one experiment
q = session.query(database.NCExperiment)
assert q.count() == 1
# And one correctly indexed (present) file
q = session.query(database.NCFile)
r = q.all()
assert len(r) == 1
assert r[0].present
def test_same_expt_name(session_db):
session, db = session_db
# index multiple experiments with different root directories, but the same
# final path component (experiment name)
database.build_index(['test/data/indexing/multiple/experiment_a', 'test/data/indexing/alternate/experiment_a'], session)
# the indexing shouldn't fail, and we should have two distinct experiments
# with the same name
q = (session
.query(database.NCExperiment)
.filter(database.NCExperiment.experiment == 'experiment_a'))
r = q.all()
assert(len(r) == 2)
assert(r[0].root_dir != r[1].root_dir)
def test_metadata(session_db):
"""Test that metadata.yaml is read for an experiment during indexing"""
session, db = session_db
database.build_index("test/data/indexing/metadata", session)
# query metadata
q = session.query(
database.NCExperiment.contact,
database.NCExperiment.created,
database.NCExperiment.description,
)
r = q.one()
assert r[0] == "The ACCESS Oracle"
assert r[1] == "2018-01-01"
assert len(r[2]) > 0
def test_following_symlinks(session_db):
session, db = session_db
# Indexing symlinked experiment should fail with default arguments
database.build_index("test/data/indexing/symlinked/experiment_a", session)
q = session.query(database.NCExperiment)
assert q.count() == 0
# Now specify to follow symlinks
database.build_index("test/data/indexing/symlinked/experiment_a",
session,
followsymlinks=True)
q = session.query(database.NCExperiment)
assert q.count() == 1
def test_prune_missing_experiment(session_db):
session, db = session_db
database.build_index("test/data/indexing/broken_file", session)
assert db.check()
# check that we have one file
q = session.query(database.NCFile)
r = q.all()
assert len(r) == 1
# prune experiment
experiment = "incorrect_experiment"
with pytest.raises(RuntimeError,
match="No such experiment: ".format(experiment)):
database.prune_experiment(experiment, session)
def test_prune_broken(session_db):
session, db = session_db
database.build_index("test/data/indexing/broken_file", session)
assert db.check()
# check that we have one file
q = session.query(database.NCFile)
r = q.all()
assert len(r) == 1
# prune experiment
database.prune_experiment("broken_file", session)
# now the database should be empty
q = session.query(database.NCFile)
r = q.all()
assert len(r) == 0
def test_time_dimension(session_db):
session, db = session_db
database.build_index("test/data/indexing/time", session)
q = session.query(database.NCFile.time_start, database.NCFile.time_end)
assert q.count() == 5 # should pick up 5 files
q = q.filter((database.NCFile.time_start is None)
| (database.NCFile.time_end is None))
assert q.count() == 0 # but all of them should have times populated
# there should be 5 separate time variables
q = session.query(database.CFVariable)
assert q.count() == 5
# each file should have exactly one time dimension
q = (session.query(func.count(database.NCFile.ncvars)).join(
database.NCFile.ncvars).group_by(database.NCFile.id))
for r in q.all():
assert r[0] == 1
def test_index_with_prune_delete(session_db, tmpdir):
session, db = session_db
expt_dir = tmpdir / "expt"
expt_dir.mkdir()
# copy the file to a new experiment directory and index
shutil.copy("test/data/indexing/longnames/output000/test1.nc",
str(expt_dir / "test1.nc"))
database.build_index(str(expt_dir), session)
# check that we have a valid file
q = session.query(database.NCFile).filter(database.NCFile.present)
r = q.all()
assert len(r) == 1
# remove the file and build with pruning
os.remove(expt_dir / "test1.nc")
database.build_index(str(expt_dir), session, prune="delete")
# now we should still have no files
q = session.query(database.NCFile)
r = q.one_or_none()
assert r is None
def test_index_attributes(session_db):
session, db = session_db
database.build_index("test/data/querying", session)
inspector = inspect(session.get_bind())
assert inspector.get_indexes("ncattributes")[0] == {
"name": "ix_ncattributes_ncvar_id",
"column_names": ["ncvar_id"],
"unique": 0,
}
ncfile = "output000/ocean.nc"
# check that we have the right attributes for a file (just use a subset)
f = session.query(
database.NCFile).filter(database.NCFile.ncfile == ncfile).one()
file_attrs = {
"filename": "ocean.nc",
"title": "MOM5",
"grid_type": "mosaic",
"grid_tile": "1",
}
for attr, attr_val in file_attrs.items():
assert attr in f.attrs and f.attrs[attr] == attr_val
# and check a particular variable
v = (session.query(database.NCVar).join(
database.NCFile).filter(database.NCFile.ncfile == ncfile).filter(
database.NCVar.varname == "temp").one())
var_attrs = {
"long_name": "Potential temperature",
"cell_methods": "time: mean",
"coordinates": "geolon_t geolat_t",
}
for attr, attr_val in var_attrs.items():
assert attr in v.attrs and v.attrs[attr] == attr_val
def test_prune_nodelete(session_db, tmpdir):
session, db = session_db
expt_dir = tmpdir / "expt"
expt_dir.mkdir()
# copy the file to a new experiment directory and index
shutil.copy("test/data/indexing/longnames/output000/test1.nc",
str(expt_dir / "test1.nc"))
database.build_index(str(expt_dir), session)
# check that we have a valid file
q = session.query(database.NCFile).filter(database.NCFile.present)
r = q.all()
assert len(r) == 1
# remove the file and prune
os.remove(expt_dir / "test1.nc")
database.prune_experiment("expt", session, delete=False)
# now we should still have one file, but now not present
q = session.query(database.NCFile)
r = q.one_or_none()
assert r is not None
assert not r.present
def test_empty_file(session_db):
session, db = session_db
indexed = database.build_index("test/data/indexing/empty_file", session)
# as with test_broken, we should have seen a single file,
# but it should be marked as empty
assert db.check()
assert indexed == 1
q = session.query(database.NCFile)
r = q.all()
assert len(r) == 1
assert not r[0].present
# but there should be a valid variable
q = session.query(func.count(
database.NCVar.id)).filter(database.NCVar.varname == "ty_trans_rho")
assert q.scalar() == 1
def test_broken(session_db):
session, db = session_db
indexed = database.build_index("test/data/indexing/broken_file", session)
# make sure the database was created
assert db.check()
# we indexed a single file
assert indexed == 1
# query ncfiles table -- should have a single file, marked as empty
q = session.query(database.NCFile)
r = q.all()
assert len(r) == 1
assert not r[0].present
# query ncvars table -- should be empty
q = session.query(func.count(database.NCVar.id))
assert q.scalar() == 0
def test_unreadable(session_db, unreadable_dir):
session, db = session_db
with pytest.warns(UserWarning,
match="Some files or directories could not be read"):
indexed = database.build_index(str(unreadable_dir), session)
def test_get_experiments_metadata(session_db):
"""Test that get_experiments returns metadata correctly"""
session, db = session_db
database.build_index("test/data/indexing/metadata", session)
r = querying.get_experiments(session, contact=True)
df = pd.DataFrame.from_dict({
"experiment": ["metadata"],
"contact": ["The ACCESS Oracle"],
"ncfiles": [1]
})
assert_frame_equal(r, df)
r = querying.get_experiments(session, email=True)
df = pd.DataFrame.from_dict({
"experiment": ["metadata"],
"email": ["*****@*****.**"],
"ncfiles": [1]
})
assert_frame_equal(r, df)
r = querying.get_experiments(session, url=True)
df = pd.DataFrame.from_dict({
"experiment": ["metadata"],
"url": ["https://github.com/COSIMA/oracle"],
"ncfiles": [1],
})
assert_frame_equal(r, df)
r = querying.get_experiments(session, description=True)
df = pd.DataFrame.from_dict({
"experiment": ["metadata"],
"description":
[("Attempted spinup, using salt flux fix "
"https://arccss.slack.com/archives/C6PP0GU9Y/p1515460656000124 "
"and https://github.com/mom-ocean/MOM5/pull/208/commits/9f4ee6f8b72b76c96a25bf26f3f6cdf773b424d2 "
"from the start. Used mushy ice from July year 1 onwards to avoid vertical thermo error in cice "
"https://arccss.slack.com/archives/C6PP0GU9Y/p1515842016000079")],
"ncfiles": [1],
})
assert_frame_equal(r, df)
r = querying.get_experiments(session, notes=True)
df = pd.DataFrame.from_dict({
"experiment": ["metadata"],
"notes":
[("Stripy salt restoring: "
"https://github.com/OceansAus/access-om2/issues/74 tripole seam bug: "
"https://github.com/OceansAus/access-om2/issues/86 requires dt=300s "
"in May, dt=240s in Aug to maintain CFL in CICE near tripoles (storms "
"in those months in 8485RYF); all other months work with dt=400s")],
"ncfiles": [1],
})
assert_frame_equal(r, df)
r = querying.get_experiments(session, created=True)
df = pd.DataFrame.from_dict({
"experiment": ["metadata"],
"created": ["2018-01-01"],
"ncfiles": [1]
})
assert_frame_equal(r, df)
r = querying.get_experiments(session, root_dir=True)
# Won't try and match a path that can change on different platforms
# assert_frame_equal(r, df)
assert r.shape == (1, 3)
r = querying.get_experiments(session, all=True)
# Won't try and match everything, just check dimensions are correct
assert r.shape == (1, 9)
# Test turning off returning experiment (bit dumb, but hey ...)
r = querying.get_experiments(session, experiment=False)
df = pd.DataFrame.from_dict({"ncfiles": [1]})
assert_frame_equal(r, df)
def test_broken_metadata(session_db):
session, db = session_db
indexed = database.build_index("test/data/indexing/broken_metadata",
session)
assert indexed == 1
def test_get_experiments_with_keywords(session_db):
"""Test retrieval of experiments with keyword filtering"""
session, db = session_db
database.build_index("test/data/metadata/keywords", session)
database.build_index("test/data/metadata/keywords2", session)
# Test keyword common to both experiments
r = querying.get_experiments(session, keywords="cosima")
df = pd.DataFrame.from_dict({
"experiment": ["keywords", "keywords2"],
"ncfiles": [1, 1]
})
assert_frame_equal(r, df)
# Test keyword common to both experiments using wildcard
r = querying.get_experiments(session, keywords="cos%")
df = pd.DataFrame.from_dict({
"experiment": ["keywords", "keywords2"],
"ncfiles": [1, 1]
})
assert_frame_equal(r, df)
r = querying.get_experiments(session, keywords="%-%")
df = pd.DataFrame.from_dict({
"experiment": ["keywords", "keywords2"],
"ncfiles": [1, 1]
})
assert_frame_equal(r, df)
r = querying.get_experiments(session, keywords="access-om2%")
df = pd.DataFrame.from_dict({"experiment": ["keywords"], "ncfiles": [1]})
assert_frame_equal(r, df)
# Test keyword in only one experiment
r = querying.get_experiments(session, keywords="another-keyword")
df = pd.DataFrame.from_dict({"experiment": ["keywords2"], "ncfiles": [1]})
assert_frame_equal(r, df)
r = querying.get_experiments(session, keywords="ryf9091")
df = pd.DataFrame.from_dict({"experiment": ["keywords"], "ncfiles": [1]})
assert_frame_equal(r, df)
# Test passing an array of keywords that match only one experiment
r = querying.get_experiments(session,
keywords=["cosima", "another-keyword"])
df = pd.DataFrame.from_dict({"experiment": ["keywords2"], "ncfiles": [1]})
assert_frame_equal(r, df)
# Test passing an array of keywords that will not match any one experiment
r = querying.get_experiments(session,
keywords=["another-keyword", "ryf9091"])
df = pd.DataFrame(columns=["experiment", "ncfiles"])
assert_frame_equal(r, df)
# Test passing a non-existent keyword along with one present. Should return
# nothing as no experiment contains it
r = querying.get_experiments(session,
keywords=["ryf9091", "not-a-keyword"])
df = pd.DataFrame(columns=["experiment", "ncfiles"])
assert_frame_equal(r, df)
# Test passing only a non-existent keyword
r = querying.get_experiments(session, keywords=["not-a-keyword"])
df = pd.DataFrame(columns=["experiment", "ncfiles"])
assert_frame_equal(r, df)
# Test passing only a non-existent wildcard keyword
r = querying.get_experiments(session, keywords=["z%"])
df = pd.DataFrame(columns=["experiment", "ncfiles"])
assert_frame_equal(r, df)
