Check for duplicate alleles

Fixes tskit-dev#927

tests/test_variantdata.py

@@ -754,3 +754,20 @@ def test_empty_alleles_not_at_end(self, tmp_path):
         samples = tsinfer.VariantData(path, "variant_ancestral_allele")
         with pytest.raises(ValueError, match="Empty alleles must be at the end"):
             tsinfer.infer(samples)
+
+    def test_unique_alleles(self, tmp_path):
+        path = tmp_path / "data.zarr"
+        ds = sgkit.simulate_genotype_call_dataset(n_variant=3, n_sample=3, n_ploidy=1)
+        ds["variant_allele"] = (
+            ds["variant_allele"].dims,
+            np.array([["A", "C", "T"], ["A", "C", ""], ["A", "A", ""]], dtype="S1"),
+        )
+        ds["variant_ancestral_allele"] = (
+            ["variants"],
+            np.array(["A", "A", "A"], dtype="S1"),
+        )
+        sgkit.save_dataset(ds, path)
+        with pytest.raises(
+            ValueError, match="Duplicate allele values provided at site 2"
+        ):
+            tsinfer.VariantData(path, "variant_ancestral_allele")

tsinfer/formats.py

@@ -2398,6 +2398,12 @@ def __init__(
                     f"The ancestral allele {ancestral_allele} was not"
                     f" found in the dataset."
                 )
+        for i, (alleles, num_alleles) in enumerate(
+            zip(self.sites_alleles, self.num_alleles())
+        ):
+            if len(set(alleles) - {b"", "", None}) != num_alleles:
+                raise ValueError(f"Duplicate allele values provided at site {i}")
+
         self._sites_ancestral_allele = self._sites_ancestral_allele.astype(str)
         unknown_alleles = collections.Counter()
         converted = np.zeros(self.num_sites, dtype=np.int8)