def _rechunk_bgen(shared_datadir: Path, tmp_path: Path,
**kwargs: Any) -> Tuple[xr.Dataset, xr.Dataset, str]:
path = shared_datadir / "example.bgen"
ds = read_bgen(path, chunks=(10, -1, -1))
store = tmp_path / "example.zarr"
dsr = rechunk_bgen(ds, store, **kwargs)
return ds, dsr, str(store)
def test_read_bgen__with_sample_file(shared_datadir):
# The example file was generated using
# qctool -g sgkit_bgen/tests/data/example.bgen -og sgkit_bgen/tests/data/example-separate-samples.bgen -os sgkit_bgen/tests/data/example-separate-samples.sample -incl-samples sgkit_bgen/tests/data/samples
# Then editing example-separate-samples.sample to change the sample IDs
path = shared_datadir / "example-separate-samples.bgen"
ds = read_bgen(path)
# Check the sample IDs are the ones from the .sample file
assert ds["sample_id"].values.tolist() == ["s1", "s2", "s3", "s4", "s5"]
def test_read_bgen__contig_dtype(shared_datadir, dtype):
path = shared_datadir / "example.bgen"
ds = read_bgen(path, contig_dtype=dtype)
dtype = np.dtype(dtype)
if dtype.kind in {"U", "S"}:
assert ds["variant_contig"].dtype == np.int64
else:
assert ds["variant_contig"].dtype == dtype
def test_read_bgen__fancy_index(shared_datadir, chunks):
path = shared_datadir / "example.bgen"
ds = read_bgen(path, chunks=chunks)
npt.assert_almost_equal(ds["call_genotype_probability"][INDEXES, 0],
EXPECTED_PROBABILITIES,
decimal=3)
npt.assert_almost_equal(ds["call_dosage"][INDEXES, 0],
EXPECTED_DOSAGES,
decimal=3)
def test_read_bgen__with_no_samples(shared_datadir):
# The example file was generated using
# qctool -g sgkit_bgen/tests/data/example.bgen -og sgkit_bgen/tests/data/example-no-samples.bgen -os sgkit_bgen/tests/data/example-no-samples.sample -bgen-omit-sample-identifier-block -incl-samples sgkit_bgen/tests/data/samples
# Then deleting example-no-samples.sample
path = shared_datadir / "example-no-samples.bgen"
ds = read_bgen(path)
# Check the sample IDs are generated
assert ds["sample_id"].values.tolist() == [
b"sample_0",
b"sample_1",
b"sample_2",
b"sample_3",
b"sample_4",
]
def test_read_bgen__scalar_index(shared_datadir, chunks):
path = shared_datadir / "example.bgen"
ds = read_bgen(path, chunks=chunks)
for i, ix in enumerate(INDEXES):
npt.assert_almost_equal(ds["call_genotype_probability"][ix, 0],
EXPECTED_PROBABILITIES[i],
decimal=3)
npt.assert_almost_equal(ds["call_dosage"][ix, 0],
EXPECTED_DOSAGES[i],
decimal=3)
for j in range(3):
npt.assert_almost_equal(
ds["call_genotype_probability"][ix, 0, j],
EXPECTED_PROBABILITIES[i, j],
decimal=3,
)
def test_read_bgen(shared_datadir, chunks):
path = shared_datadir / "example.bgen"
ds = read_bgen(path, chunks=chunks)
# check some of the data (in different chunks)
assert ds["call_dosage"].shape == _shape("variants", "samples")
npt.assert_almost_equal(ds["call_dosage"].values[1][0], 1.987, decimal=3)
npt.assert_almost_equal(ds["call_dosage"].values[100][0], 0.160, decimal=3)
npt.assert_array_equal(ds["call_dosage_mask"].values[0, 0], [True])
npt.assert_array_equal(ds["call_dosage_mask"].values[0, 1], [False])
assert ds["call_genotype_probability"].shape == _shape(
"variants", "samples", "genotypes")
npt.assert_almost_equal(ds["call_genotype_probability"].values[1][0],
[0.005, 0.002, 0.992],
decimal=3)
npt.assert_almost_equal(ds["call_genotype_probability"].values[100][0],
[0.916, 0.007, 0.076],
decimal=3)
npt.assert_array_equal(ds["call_genotype_probability_mask"].values[0, 0],
[True] * 3)
npt.assert_array_equal(ds["call_genotype_probability_mask"].values[0, 1],
[False] * 3)
def test_read_bgen__invalid_chunks(shared_datadir):
path = shared_datadir / "example.bgen"
with pytest.raises(ValueError,
match="`chunks` must be tuple with 3 items"):
read_bgen(path, chunks=(100, -1)) # type: ignore[arg-type]
def test_read_bgen__invalid_contig_dtype(shared_datadir, dtype):
path = shared_datadir / "example.bgen"
with pytest.raises(ValueError,
match="`contig_dtype` must be of string or int type"):
read_bgen(path, contig_dtype=dtype)
def test_read_bgen__invalid_gp_dtype(shared_datadir, dtype):
path = shared_datadir / "example.bgen"
with pytest.raises(ValueError,
match="`gp_dtype` must be a floating point data type"):
read_bgen(path, gp_dtype=dtype)
def test_read_bgen__gp_dtype(shared_datadir, dtype):
path = shared_datadir / "example.bgen"
ds = read_bgen(path, gp_dtype=dtype)
dtype = np.dtype(dtype)
assert ds["call_genotype_probability"].dtype == dtype
assert ds["call_dosage"].dtype == dtype
