Better formatted error message for ancestral alleles

Author: hyanwong

tests/test_sgkit.py

@@ -538,7 +538,10 @@ def test_ancestral_missingness(tmp_path):
     )
     ds = sgkit.load_dataset(str(zarr_path) + ".tmp")
     sd = tsinfer.SgkitSampleData(str(zarr_path) + ".tmp")
-    with pytest.warns(UserWarning, match="The following alleles were not found"):
+    with pytest.warns(
+        UserWarning,
+        match="Ancestral alleles not found in the variant_allele array for 3 sites",
+    ):
         inf_ts = tsinfer.infer(sd)
     for i, (
         inf_var,

tsinfer/formats.py

@@ -2405,11 +2405,16 @@ def sites_ancestral_allele(self):
             except IndexError:
                 unknown_alleles[allele] += 1
             ret[i] = allele_index
-        if sum(unknown_alleles.values()) > 0:
+        tot = sum(unknown_alleles.values())
+        if tot > 0:
             warnings.warn(
-                "The following alleles were not found in the variant_allele array "
-                "and will be treated as unknown:\n"
-                f"{unknown_alleles}"
+                "Ancestral alleles not found in the variant_allele "
+                f"array for the {tot} sites ({tot/len(string_allele)*100:.2f}%) "
+                "listed below. They will be treated as of unknown ancestral state:\n "
+                + "\n ".join(
+                    f"'{k}': {v} ({v/len(string_allele)*100:.2f}% of sites)"
+                    for k, v in unknown_alleles.items()
+                )
             )
         return ret