Merge pull request #121 from jeromekelleher/trim-empty-regions

Trim empty regions

CHANGELOG.md

@@ -1,9 +1,10 @@
-# 0.0.5 2024-04-XX
+# 0.0.5 2024-04-17
 
 - Fix bug in schema handling (compressor settings ignored)
 - Move making ICF field partition directories into per-partition processing.
   Remove progress on the init mkdirs step.
 - Turn off progress monitor on dexplode-partition
+- Fix empty partition bug
 
 # 0.0.4 2024-04-08
 

bio2zarr/vcf_utils.py

@@ -435,6 +435,22 @@ def variants(self, region):
             if var.POS >= start:
                 yield var
 
+    def _filter_empty(self, regions):
+        """
+        Return all regions in the specified list that have one or more records.
+
+        Sometimes with Tabix indexes these seem to crop up:
+
+        - https://github.com/sgkit-dev/bio2zarr/issues/45
+        - https://github.com/sgkit-dev/bio2zarr/issues/120
+        """
+        ret = []
+        for region in regions:
+            variants = self.variants(region)
+            if next(variants, None) is not None:
+                ret.append(region)
+        return ret
+
     def partition_into_regions(
         self,
         num_parts: Optional[int] = None,
@@ -511,4 +527,4 @@ def partition_into_regions(
             if self.index.record_counts[ri] > 0:
                 regions.append(Region(self.sequence_names[ri]))
 
-        return regions
+        return self._filter_empty(regions)

tests/test_vcf_utils.py

@@ -9,13 +9,7 @@
 
 
 def assert_part_counts_non_zero(part_counts, index_file):
-    # We may have one zero count value at the end in Tabix indexes.
-    # Should probably try to get rid of it, but probably no harm
-    # https://github.com/jeromekelleher/bio2zarr/issues/45
-    if index_file.endswith(".tbi"):
-        assert np.all(part_counts[:-1] > 0)
-    else:
-        assert np.all(part_counts > 0)
+    assert np.all(part_counts > 0)
 
 
 class TestIndexedVcf: