Merge pull request #28 from jeromekelleher/bgen-unphased

Bgen unphased

Author: jeromekelleher

biofuse/cli.py

@@ -111,7 +111,9 @@ def mount_bgen(vcz_url, mount_dir, basename, access_log_path, **kwargs):
     FUSE process's memory; only ``.bgen`` reads cross the wire.
     ``--no-sample-file`` and ``--no-bgi`` suppress the corresponding
     sidecar; ``--no-header-samples`` drops the sample identifiers from
-    the ``.bgen`` header block.
+    the ``.bgen`` header block. ``--unphased`` ignores the input's
+    ``call_genotype_phased`` field and encodes every variant with the
+    BGEN phased flag clear.
 
     The bcftools-view-style filter / backend / log options are inherited
     from ``vcztools view-bgen``; see ``vcztools view-bgen --help`` for the

biofuse/formats.py

@@ -129,7 +129,11 @@ def _plink_encoder_factory(reader, opts):
 
 
 def _bgen_encoder_factory(reader, opts):
-    return vcztools.BgenEncoder(reader, embed_header_samples=not opts.no_header_samples)
+    return vcztools.BgenEncoder(
+        reader,
+        embed_header_samples=not opts.no_header_samples,
+        unphased=opts.unphased,
+    )
 
 
 PLINK_SPEC = FormatSpec(

tests/test_bgen_apps.py

@@ -182,6 +182,19 @@ async def test_no_header_samples_stable_across_opens(
                 data = await trio.to_thread.run_sync(bgen_path.read_bytes)
                 assert data == expected, f"cycle {cycle} differed from reference"
 
+    @pytest.mark.parametrize("unphased", [True, False])
+    async def test_unphased_stable_across_opens(self, tmp_path, fx_small_vcz, unphased):
+        opts = dataclasses.replace(vcztools.ViewBgenOptions(), unphased=unphased)
+        ref_reader = opts.make_reader(str(fx_small_vcz.path))
+        with vcztools.BgenEncoder(ref_reader, unphased=unphased) as ref:
+            expected = ref.read(0, ref.total_size)
+
+        async with _mount_bgen(tmp_path, fx_small_vcz, opts=opts) as (mnt, basename):
+            bgen_path = mnt / f"{basename}.bgen"
+            for cycle in range(3):
+                data = await trio.to_thread.run_sync(bgen_path.read_bytes)
+                assert data == expected, f"cycle {cycle} differed from reference"
+
 
 def _pread_sync(path: pathlib.Path, off: int, size: int) -> bytes:
     with path.open("rb") as f:

tests/test_formats.py

@@ -313,6 +313,34 @@ def test_embed_default_matches_in_process(self, fx_reader, fx_small_vcz):
         assert data == ref_data
 
 
+class TestBgenUnphasedToggle:
+    """``--unphased`` flows through to ``BgenEncoder(unphased=True)``.
+
+    With the flag set, the encoder must ignore ``call_genotype_phased``
+    and produce the same bytes as an in-process ``BgenEncoder`` built
+    with ``unphased=True`` on the same reader.
+    """
+
+    def test_unphased_matches_in_process(self, fx_reader, fx_small_vcz):
+        opts = vcztools.ViewBgenOptions(unphased=True)
+        with formats.BGEN_SPEC.encoder_factory(fx_reader, opts) as encoder:
+            data = encoder.read(0, encoder.total_size)
+        ref_reader = _open_reader(fx_small_vcz.path)
+        with vcztools.BgenEncoder(ref_reader, unphased=True) as ref:
+            ref_data = ref.read(0, ref.total_size)
+        assert data == ref_data
+
+    def test_unphased_default_matches_in_process(self, fx_reader, fx_small_vcz):
+        opts = vcztools.ViewBgenOptions()
+        assert opts.unphased is False
+        with formats.BGEN_SPEC.encoder_factory(fx_reader, opts) as encoder:
+            data = encoder.read(0, encoder.total_size)
+        ref_reader = _open_reader(fx_small_vcz.path)
+        with vcztools.BgenEncoder(ref_reader, unphased=False) as ref:
+            ref_data = ref.read(0, ref.total_size)
+        assert data == ref_data
+
+
 class TestSpecsRegistry:
     def test_specs_dict_has_both_entries(self):
         assert formats.SPECS == {"plink": formats.PLINK_SPEC, "bgen": formats.BGEN_SPEC}