(feat): allow join=outer for concat_on_disk

docs/release-notes/0.10.9.md

@@ -8,3 +8,5 @@
 
 ```{rubric} Performance
 ```
+
+* Support for `concat_on_disk` outer join {pr}`1504` {user}`ilan-gold`

src/anndata/experimental/merge.py

@@ -538,8 +538,6 @@ def concat_on_disk(
     # Argument normalization
     if pairwise:
         raise NotImplementedError("pairwise concatenation not yet implemented")
-    if join != "inner":
-        raise NotImplementedError("only inner join is currently supported")
 
     merge = resolve_merge_strategy(merge)
     uns_merge = resolve_merge_strategy(uns_merge)

tests/test_concatenate_disk.py

@@ -39,7 +39,7 @@ def array_type(request):
     return request.param
 
 
-@pytest.fixture(params=["inner"])
+@pytest.fixture(params=["inner", "outer"])
 def join_type(request):
     return request.param
 
@@ -126,10 +126,19 @@ def test_anndatas_without_reindex(
             sparse_fmt=sparse_fmt,
             **GEN_ADATA_OOC_CONCAT_ARGS,
         )
+        # ensure some names overlap, others do not, for the off-axis so that inner/outer is properly tested
         if axis == 0:
             a.obs_names = f"{i}-" + a.obs_names
+            a.var_names = [
+                f"{i}-{name}" if var_ind % 2 else name
+                for var_ind, name in enumerate(a.var_names)
+            ]
         else:
             a.var_names = f"{i}-" + a.var_names
+            a.obs_names = [
+                f"{i}-{name}" if obs_ind % 2 else name
+                for obs_ind, name in enumerate(a.obs_names)
+            ]
         adatas.append(a)
 
     assert_eq_concat_on_disk(
@@ -153,7 +162,7 @@ def test_anndatas_with_reindex(
     if axis == 0:
         sparse_fmt = "csr"
 
-    for _ in range(5):
+    for i in range(5):
         M = np.random.randint(1, 100)
         N = np.random.randint(1, 100)
 
@@ -166,11 +175,20 @@ def test_anndatas_with_reindex(
                 np.ndarray,
                 pd.DataFrame,
             ),
-            varm_types=(get_array_type("sparse", axis), np.ndarray, pd.DataFrame),
+            varm_types=(get_array_type("sparse", 1 - axis), np.ndarray, pd.DataFrame),
             layers_types=(get_array_type("sparse", axis), np.ndarray, pd.DataFrame),
         )
-        a.layers["sparse"] = get_array_type("sparse", axis)(a.layers["sparse"])
-        a.varm["sparse"] = get_array_type("sparse", 1 - axis)(a.varm["sparse"])
+        # ensure some names overlap, others do not, for the off-axis so that inner/outer is properly tested
+        if axis == 1:
+            a.obs_names = [
+                f"{i}-{name}" if obs_ind % 2 else name
+                for obs_ind, name in enumerate(a.obs_names)
+            ]
+        else:
+            a.var_names = [
+                f"{i}-{name}" if var_ind % 2 else name
+                for var_ind, name in enumerate(a.var_names)
+            ]
         adatas.append(a)
 
     assert_eq_concat_on_disk(
@@ -208,7 +226,7 @@ def test_concat_ordered_categoricals_retained(tmp_path, file_format):
 
 
 @pytest.fixture()
-def obsm_adatas():
+def xxxm_adatas():
     def gen_index(n):
         return [f"cell{i}" for i in range(n)]
 
@@ -256,8 +274,12 @@ def gen_index(n):
     ]
 
 
-def test_concatenate_obsm_inner(obsm_adatas, tmp_path, file_format):
-    assert_eq_concat_on_disk(obsm_adatas, tmp_path, file_format, join="inner")
+def test_concatenate_xxxm(xxxm_adatas, tmp_path, file_format, join_type):
+    if join_type == "outer":
+        for i in range(len(xxxm_adatas)):
+            xxxm_adatas[i] = xxxm_adatas[i].T
+            xxxm_adatas[i].X = sparse.csr_matrix(xxxm_adatas[i].X)
+    assert_eq_concat_on_disk(xxxm_adatas, tmp_path, file_format, join=join_type)
 
 
 def test_output_dir_exists(tmp_path):