Warn when using defaults that aren't explict and not metadata

tests/test_variantdata.py

@@ -218,7 +218,11 @@ def test_variantdata_accessors_defaults(tmp_path, in_mem):
     ds = data if in_mem else sgkit.load_dataset(data)
 
     default_schema = tskit.MetadataSchema.permissive_json().schema
-    assert vdata.sequence_length == ts.sequence_length
+    with pytest.warns(
+        UserWarning,
+        match="`sequence_length` was not found as an attribute in the dataset",
+    ):
+        assert vdata.sequence_length == ts.sequence_length
     assert vdata.sites_metadata_schema == default_schema
     assert vdata.sites_metadata == [{} for _ in range(ts.num_sites)]
     for time in vdata.sites_time:
@@ -234,17 +238,32 @@ def test_variantdata_accessors_defaults(tmp_path, in_mem):
     assert vdata.individuals_metadata == [
         {"variant_data_sample_id": sample_id} for sample_id in ds.sample_id[:]
     ]
-    for time in vdata.individuals_time:
-        assert tskit.is_unknown_time(time)
-    assert np.array_equal(
-        vdata.individuals_location, np.array([[]] * ts.num_individuals, dtype=float)
-    )
-    assert np.array_equal(
-        vdata.individuals_population, np.full(ts.num_individuals, tskit.NULL)
-    )
-    assert np.array_equal(
-        vdata.individuals_flags, np.zeros(ts.num_individuals, dtype=int)
-    )
+    with pytest.warns(
+        UserWarning, match="`individuals_time` was not found as an array in the dataset"
+    ):
+        for time in vdata.individuals_time:
+            assert tskit.is_unknown_time(time)
+    with pytest.warns(
+        UserWarning,
+        match="`individuals_location` was not found as an array in the dataset",
+    ):
+        assert np.array_equal(
+            vdata.individuals_location, np.array([[]] * ts.num_individuals, dtype=float)
+        )
+    with pytest.warns(
+        UserWarning,
+        match="`individuals_population` was not found as an array in the dataset",
+    ):
+        assert np.array_equal(
+            vdata.individuals_population, np.full(ts.num_individuals, tskit.NULL)
+        )
+    with pytest.warns(
+        UserWarning,
+        match="`individuals_flags` was not found as an array in the dataset",
+    ):
+        assert np.array_equal(
+            vdata.individuals_flags, np.zeros(ts.num_individuals, dtype=int)
+        )
 
 
 @pytest.mark.skipif(sys.platform == "win32", reason="No cyvcf2 on windows")

tsinfer/formats.py

@@ -2520,6 +2520,12 @@ def sequence_length(self):
         try:
             return self.data.attrs["sequence_length"]
         except KeyError:
+            warnings.warn(
+                "`sequence_length` was not found as an attribute in the dataset, so"
+                " the largest position has been used. It can be set with"
+                " ds.attrs['sequence_length'] = 1337; ds.to_zarr('path/to/store',"
+                " mode='a')"
+            )
             return int(np.max(self.data["variant_position"])) + 1
 
     @property
@@ -2653,6 +2659,12 @@ def individuals_time(self):
         try:
             return self.data["individuals_time"][:][self.individuals_select]
         except KeyError:
+            warnings.warn(
+                "`individuals_time` was not found as an array in the dataset, so "
+                "tskit.UNKNOWN_TIME has been used. It can be apppended to the dataset "
+                "with data_array.to_zarr('path/to/store', append_dim='samples', "
+                "mode='a')"
+            )
             return np.full(self.num_individuals, tskit.UNKNOWN_TIME)
 
     @functools.cached_property
@@ -2696,20 +2708,35 @@ def individuals_location(self):
         try:
             return self.data["individuals_location"][:][self.individuals_select]
         except KeyError:
+            warnings.warn(
+                "`individuals_location` was not found as an array in the dataset, "
+                "so [] has been used. It can be apppended to the dataset with "
+                "data_array.to_zarr('path/to/store', append_dim='samples', mode='a')"
+            )
             return np.array([[]] * self.num_individuals, dtype=float)
 
     @functools.cached_property
     def individuals_population(self):
         try:
             return self.data["individuals_population"][:][self.individuals_select]
         except KeyError:
+            warnings.warn(
+                "`individuals_population` was not found as an array in the dataset, "
+                "so tskit.NULL has been used. It can be apppended to the dataset with "
+                "data_array.to_zarr('path/to/store', append_dim='samples', mode='a')"
+            )
             return np.full((self.num_individuals), tskit.NULL, dtype=np.int32)
 
     @functools.cached_property
     def individuals_flags(self):
         try:
             return self.data["individuals_flags"][:][self.individuals_select]
         except KeyError:
+            warnings.warn(
+                "`individuals_flags` was not found as an array in the dataset, so 0 "
+                "has been used. It can be apppended to the dataset with "
+                "data_array.to_zarr('path/to/store', append_dim='samples', mode='a')"
+            )
             return np.full((self.num_individuals), 0, dtype=np.int32)
 
     @staticmethod