Switch to / dimension separator by default

Aligning with V3 defaults. Seems to require zarr >= 2.17

Author: jeromekelleher

bio2zarr/vcf.py

@@ -1512,15 +1512,13 @@ def __init__(self, path, icf, schema, dimension_separator=None):
         self.path = pathlib.Path(path)
         self.icf = icf
         self.schema = schema
-        store = zarr.DirectoryStore(self.path)
         # Default to using nested directories following the Zarr v3 default.
-        self.dimension_separator = (
-            "/" if dimension_separator is None else dimension_separator
-        )
+        # This seems to require version 2.17+ to work properly
+        self.dimension_separator = "/" if dimension_separator is None else dimension_separator
+        store = zarr.DirectoryStore(self.path)
         self.root = zarr.group(store=store)
 
     def init_array(self, variable):
-        # print("CREATE", variable)
         object_codec = None
         if variable.dtype == "O":
             object_codec = numcodecs.VLenUTF8()
@@ -1865,6 +1863,7 @@ def encode(
     variants_chunk_size=None,
     samples_chunk_size=None,
     max_v_chunks=None,
+    dimension_separator=None,
     max_memory=None,
     worker_processes=1,
     show_progress=False,
@@ -1888,7 +1887,7 @@ def encode(
     if zarr_path.exists():
         logger.warning(f"Deleting existing {zarr_path}")
         shutil.rmtree(zarr_path)
-    vzw = VcfZarrWriter(zarr_path, icf, schema)
+    vzw = VcfZarrWriter(zarr_path, icf, schema, dimension_separator=dimension_separator)
     vzw.init()
     vzw.encode(
         max_v_chunks=max_v_chunks,

requirements/development.txt

@@ -8,4 +8,5 @@ pytest-coverage
 pytest-xdist
 sgkit
 tabulate
-tqdm
+tqdm
+zarr>=2.17

setup.cfg

@@ -28,7 +28,7 @@ include_package_data = True
 python_requires = >=3.9
 install_requires =
     numpy
-    zarr >= 2.10.0, != 2.11.0, != 2.11.1, != 2.11.2
+    zarr >= 2.17
     click
     tabulate
     tqdm

tests/test_vcf.py

@@ -99,6 +99,29 @@ def test_exploded_metadata_mismatch(self, tmpdir, icf_path, version):
             vcf.IcfMetadata.fromdict(d)
 
 
+class TestEncodeDimensionSeparator:
+
+    @pytest.mark.parametrize("dimension_separator", [None, "/"])
+    def test_directories(self, tmp_path, icf_path, dimension_separator):
+        zarr_path = tmp_path / "zarr"
+        vcf.encode(icf_path, zarr_path, dimension_separator=dimension_separator)
+        # print(zarr_path)
+        chunk_file = zarr_path / "call_genotype" / "0" / "0" / "0"
+        assert chunk_file.exists()
+
+    def test_files(self, tmp_path, icf_path):
+        zarr_path = tmp_path / "zarr"
+        vcf.encode(icf_path, zarr_path, dimension_separator=".")
+        chunk_file = zarr_path / "call_genotype" / "0.0.0"
+        assert chunk_file.exists()
+
+    @pytest.mark.parametrize("dimension_separator", ["\\", "X", []])
+    def test_bad_value(self, tmp_path, icf_path, dimension_separator):
+        zarr_path = tmp_path / "zarr"
+        with pytest.raises(ValueError):
+            vcf.encode(icf_path, zarr_path, dimension_separator=dimension_separator)
+
+
 class TestDefaultSchema:
     def test_format_version(self, schema):
         assert schema["format_version"] == vcf.ZARR_SCHEMA_FORMAT_VERSION