Revert to Zstd as ICF encoder.

Author: jeromekelleher

bio2zarr/cli.py

@@ -75,7 +75,7 @@ def list_commands(self, ctx):
     "--compressor",
     type=click.Choice(["lz4", "zstd"]),
     default=None,
-    help="Codec to use for compressing column chunks",
+    help="Codec to use for compressing column chunks (Default=zstd)."
 )
 
 # Note: -l and -w were chosen when these were called "width" and "length".

bio2zarr/vcf.py

@@ -151,7 +151,7 @@ class VcfPartition:
 
 ICF_METADATA_FORMAT_VERSION = "0.2"
 ICF_DEFAULT_COMPRESSOR = numcodecs.Blosc(
-    cname="lz4", clevel=7, shuffle=numcodecs.Blosc.NOSHUFFLE
+    cname="zstd", clevel=7, shuffle=numcodecs.Blosc.NOSHUFFLE
 )
 
 
@@ -1505,7 +1505,9 @@ def __init__(self, path, icf, schema, dimension_separator=None):
         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
+        self.dimension_separator = (
+            "/" if dimension_separator is None else dimension_separator
+        )
         self.root = zarr.group(store=store)
 
     def init_array(self, variable):
@@ -1521,7 +1523,7 @@ def init_array(self, variable):
             compressor=numcodecs.get_codec(variable.compressor),
             filters=[numcodecs.get_codec(filt) for filt in variable.filters],
             object_codec=object_codec,
-            dimension_separator=self.dimension_separator
+            dimension_separator=self.dimension_separator,
         )
         # Dimension names are part of the spec in Zarr v3
         a.attrs["_ARRAY_DIMENSIONS"] = variable.dimensions

tests/test_icf.py

@@ -275,6 +275,16 @@ def test_compressor_explode(self, tmp_path, compressor):
         icf = self.run_explode(tmp_path, compressor=compressor)
         assert icf.metadata.compressor == compressor.get_config()
 
+    def test_default_compressor_explode(self, tmp_path):
+        icf = self.run_explode(tmp_path)
+        assert icf.metadata.compressor == vcf.ICF_DEFAULT_COMPRESSOR.get_config()
+        assert icf.metadata.compressor["cname"] == "zstd"
+
+    def test_default_compressor_dexplode(self, tmp_path):
+        icf = self.run_dexplode(tmp_path)
+        assert icf.metadata.compressor == vcf.ICF_DEFAULT_COMPRESSOR.get_config()
+        assert icf.metadata.compressor["cname"] == "zstd"
+
     @pytest.mark.parametrize(
         "compressor",
         [
@@ -340,7 +350,8 @@ def test_empty_chunk_file(self, tmp_path):
         with pytest.raises(RuntimeError, match="blosc"):
             icf["POS"].values
 
-    @pytest.mark.parametrize("length", [10, 100, 200, 210])
+    # Chunk file is 195 long
+    @pytest.mark.parametrize("length", [10, 100, 190, 194])
     def test_truncated_chunk_file(self, tmp_path, length):
         icf_path = tmp_path / "icf"
         vcf.explode(icf_path, [self.data_path])
@@ -359,7 +370,7 @@ def test_chunk_incorrect_length(self, tmp_path):
         icf_path = tmp_path / "icf"
         vcf.explode(icf_path, [self.data_path])
         chunk_file = icf_path / "POS" / "p0" / "2"
-        compressor = numcodecs.Blosc(cname="lz4")
+        compressor = numcodecs.Blosc(cname="zstd")
         with open(chunk_file, "rb") as f:
             pkl = compressor.decode(f.read())
         x = pickle.loads(pkl)

tests/test_vcf_examples.py

@@ -841,7 +841,7 @@ def test_split_explode(tmp_path):
     pcvcf = vcf.IntermediateColumnarFormat(out)
     assert pcvcf.columns["POS"].vcf_field.summary.asdict() == {
         "num_chunks": 3,
-        "compressed_size": 630,
+        "compressed_size": 587,
         "uncompressed_size": 1008,
         "max_number": 1,
         "max_value": 1235237,