ENH: fill missing variables during concat by reindexing

xarray/core/concat.py

@@ -378,7 +378,9 @@ def process_subset_opt(opt, subset):
 
             elif opt == "all":
                 concat_over.update(
-                    set(getattr(datasets[0], subset)) - set(datasets[0].dims)
+                    set().union(
+                        *list((set(getattr(d, subset)) - set(d.dims) for d in datasets))
+                    )
                 )
             elif opt == "minimal":
                 pass
@@ -553,16 +555,35 @@ def get_indexes(name):
                     data = var.set_dims(dim).values
                     yield PandasIndex(data, dim, coord_dtype=var.dtype)
 
+    # preserve variable order for variables in first dataset
+    data_var_order = list(datasets[0].variables)
+    # append additional variables to the end
+    data_var_order += [e for e in data_names if e not in data_var_order]
+    # create concatenation index, needed for later reindexing
+    concat_index = list(range(sum(concat_dim_lengths)))
+
     # stack up each variable and/or index to fill-out the dataset (in order)
     # n.b. this loop preserves variable order, needed for groupby.
-    for name in datasets[0].variables:
+    for name in data_var_order:
         if name in concat_over and name not in result_indexes:
-            try:
-                vars = ensure_common_dims([ds[name].variable for ds in datasets])
-            except KeyError:
-                raise ValueError(f"{name!r} is not present in all datasets.")
-
-            # Try concatenate the indexes, concatenate the variables when no index
+            variables = []
+            variable_index = []
+            for i, ds in enumerate(datasets):
+                if name in ds.variables:
+                    variables.append(ds.variables[name])
+                    # add to variable index, needed for reindexing
+                    variable_index.extend(
+                        [sum(concat_dim_lengths[:i]) + k for k in range(concat_dim_lengths[i])]
+                    )
+                else:
+                    # raise if coordinate not in all datasets
+                    if name in coord_names:
+                        raise ValueError(
+                            f"coordinate {name!r} not present in all datasets."
+                        )
+            vars = list(ensure_common_dims(variables))
+
+            # Try to concatenate the indexes, concatenate the variables when no index
             # is found on all datasets.
             indexes: list[Index] = list(get_indexes(name))
             if indexes:
@@ -586,9 +607,28 @@ def get_indexes(name):
                     )
                     result_vars[k] = v
             else:
-                combined_var = concat_vars(
-                    vars, dim, positions, combine_attrs=combine_attrs
-                )
+                # if variable is only present in one dataset of multiple datasets,
+                # then do not concat
+                if len(variables) == 1 and len(datasets) > 1:
+                    combined_var = variables[0]
+                # only concat if variable is in multiple datasets
+                # or if single dataset (GH1988)
+                else:
+                    combined_var = concat_vars(
+                        vars, dim, positions, combine_attrs=combine_attrs
+                    )
+                    # reindex if variable is not present in all datasets
+                    if len(variable_index) < len(concat_index):
+                        try:
+                            fill = fill_value[name]
+                        except (TypeError, KeyError):
+                            fill = fill_value
+                        combined_var = (
+                            DataArray(data=combined_var, name=name)
+                            .assign_coords({dim: variable_index})
+                            .reindex({dim: concat_index}, fill_value=fill)
+                            .variable
+                        )
                 result_vars[name] = combined_var
 
         elif name in result_vars:

xarray/tests/test_concat.py

@@ -50,8 +50,6 @@ def test_concat_compat() -> None:
         ValueError, match=r"coordinates in some datasets but not others"
     ):
         concat([ds1, ds2], dim="q")
-    with pytest.raises(ValueError, match=r"'q' is not present in all datasets"):
-        concat([ds2, ds1], dim="q")
 
 
 class TestConcatDataset:
@@ -776,15 +774,15 @@ def test_concat_merge_single_non_dim_coord():
         actual = concat([da1, da2], "x", coords=coords)
         assert_identical(actual, expected)
 
-    with pytest.raises(ValueError, match=r"'y' is not present in all datasets."):
+    with pytest.raises(ValueError, match=r"'y' not present in all datasets."):
         concat([da1, da2], dim="x", coords="all")
 
     da1 = DataArray([1, 2, 3], dims="x", coords={"x": [1, 2, 3], "y": 1})
     da2 = DataArray([4, 5, 6], dims="x", coords={"x": [4, 5, 6]})
     da3 = DataArray([7, 8, 9], dims="x", coords={"x": [7, 8, 9], "y": 1})
     for coords in ["different", "all"]:
         with pytest.raises(ValueError, match=r"'y' not present in all datasets"):
-            concat([da1, da2, da3], dim="x")
+            concat([da1, da2, da3], dim="x", coords=coords)
 
 
 def test_concat_preserve_coordinate_order() -> None: