Merge branch 'main' into ig/io_module

docs/release-notes/1685.dev.md

@@ -0,0 +1 @@
+Remove `shall_` from variable names in `settings` {user}`ilan-gold`

src/anndata/_core/anndata.py

@@ -696,7 +696,7 @@ def raw(self) -> Raw:
         The :attr:`raw` attribute is initialized with the current content
         of an object by setting::
 
-            adata.raw = adata
+            adata.raw = adata.copy()
 
         Its content can be deleted::
 

tests/test_base.py

@@ -277,7 +277,7 @@ def test_setting_dim_index(dim):
     mapping_attr = f"{dim}m"
 
     orig = gen_adata((5, 5))
-    orig.raw = orig
+    orig.raw = orig.copy()
     curr = orig.copy()
     view = orig[:, :]
     new_idx = pd.Index(list("abcde"), name="letters")
@@ -590,7 +590,7 @@ def test_convenience():
     adata = adata_sparse.copy()
     adata.layers["x2"] = adata.X * 2
     adata.var["anno2"] = ["p1", "p2", "p3"]
-    adata.raw = adata
+    adata.raw = adata.copy()
     adata.X = adata.X / 2
     adata_dense = adata.copy()
     adata_dense.X = adata_dense.X.toarray()

tests/test_concatenate.py

@@ -697,9 +697,9 @@ def test_concatenate_with_raw():
         layers=dict(Xs=X4),
     )
 
-    adata1.raw = adata1
-    adata2.raw = adata2
-    adata3.raw = adata3
+    adata1.raw = adata1.copy()
+    adata2.raw = adata2.copy()
+    adata3.raw = adata3.copy()
 
     adata_all = AnnData.concatenate(adata1, adata2, adata3)
     assert isinstance(adata_all.raw, Raw)
@@ -713,7 +713,7 @@ def test_concatenate_with_raw():
     assert_equal(adata_all.raw.to_adata().obs, adata_all.obs)
     assert np.array_equal(np.nan_to_num(adata_all.raw.X), np.nan_to_num(adata_all.X))
 
-    adata3.raw = adata4
+    adata3.raw = adata4.copy()
     adata_all = AnnData.concatenate(adata1, adata2, adata3, join="outer")
     assert isinstance(adata_all.raw, Raw)
     assert set(adata_all.raw.var_names) == set("abcdz")

tests/test_deprecations.py

@@ -25,7 +25,7 @@ def adata():
         obs=dict(obs_names=["s1", "s2"], anno1=["c1", "c2"]),
         var=dict(var_names=["a", "b", "c"]),
     )
-    adata.raw = adata
+    adata.raw = adata.copy()
     adata.layers["x2"] = adata.X * 2
     adata.var["anno2"] = ["p1", "p2", "p3"]
     adata.X = adata.X / 2

tests/test_get_vector.py

@@ -36,7 +36,7 @@ def test_amgibuous_keys():
         ),
     )
 
-    adata.raw = adata
+    adata.raw = adata.copy()
 
     for k in var_keys:
         # These are mostly to check that the test is working

tests/test_gpu.py

@@ -24,7 +24,7 @@ def test_adata_raw_gpu():
     adata = AnnData(
         X=cupy_sparse.random(500, 50, density=0.01, format="csr", dtype=cp.float32)
     )
-    adata.raw = adata
+    adata.raw = adata.copy()
     assert isinstance(adata.raw.X, sparse.csr_matrix)
 
 

tests/test_io_conversion.py

@@ -39,7 +39,7 @@ def test_sparse_to_dense_disk(tmp_path, mtx_format, to_convert):
     dense_from_mem_pth = tmp_path / "dense_mem.h5ad"
     dense_from_disk_pth = tmp_path / "dense_disk.h5ad"
     mem = gen_adata((50, 50), mtx_format)
-    mem.raw = mem
+    mem.raw = mem.copy()
 
     mem.write_h5ad(mem_pth)
     disk = ad.read_h5ad(mem_pth, backed="r")
@@ -66,7 +66,7 @@ def test_sparse_to_dense_disk(tmp_path, mtx_format, to_convert):
 def test_sparse_to_dense_inplace(tmp_path, spmtx_format):
     pth = tmp_path / "adata.h5ad"
     orig = gen_adata((50, 50), spmtx_format)
-    orig.raw = orig
+    orig.raw = orig.copy()
     orig.write(pth)
     backed = ad.read_h5ad(pth, backed="r+")
     backed.write(as_dense=("X", "raw/X"))
@@ -97,7 +97,7 @@ def test_sparse_to_dense_errors(tmp_path):
 def test_dense_to_sparse_memory(tmp_path, spmtx_format, to_convert):
     dense_path = tmp_path / "dense.h5ad"
     orig = gen_adata((50, 50), np.array)
-    orig.raw = orig
+    orig.raw = orig.copy()
     orig.write_h5ad(dense_path)
     assert not isinstance(orig.X, sparse.spmatrix)
     assert not isinstance(orig.raw.X, sparse.spmatrix)

tests/test_io_elementwise.py

@@ -374,7 +374,7 @@ def test_write_indptr_dtype_override(store, sparse_format):
 
 def test_io_spec_raw(store):
     adata = gen_adata((3, 2))
-    adata.raw = adata
+    adata.raw = adata.copy()
 
     write_elem(store, "adata", adata)
 

tests/test_raw.py

@@ -38,7 +38,7 @@ def adata_raw() -> ad.AnnData:
     adata = ad.AnnData(
         np.array(data, dtype="int32"), obs=obs_dict, var=var_dict, uns=uns_dict
     )
-    adata.raw = adata
+    adata.raw = adata.copy()
     # Make them different shapes
     adata = adata[:, [0, 1]].copy()
     return adata
@@ -131,7 +131,7 @@ def test_raw_as_parent_view():
     # https://github.com/scverse/anndata/issues/288
     a = ad.AnnData(np.ones((4, 3)))
     a.varm["PCs"] = np.ones((3, 3))
-    a.raw = a
+    a.raw = a.copy()
     # create a Raw containing views. This used to trigger #288.
     b = a.raw[:, "0"]
     # actualize
@@ -165,3 +165,10 @@ def test_to_adata_populates_obs():
     from_raw = adata_w_raw.raw.to_adata()
 
     assert_equal(adata, from_raw)
+
+
+def test_no_copy():
+    adata = gen_adata((20, 10), X_type=np.asarray)
+    adata.raw = adata  # no .copy() herer
+    np.log1p(adata.X, out=adata.X)
+    assert adata.X is adata.raw.X

tests/test_readwrite.py

@@ -133,7 +133,7 @@ def test_readwrite_kitchensink(tmp_path, storage, typ, backing_h5ad, dataset_kwa
     X = typ(X_list)
     adata_src = ad.AnnData(X, obs=obs_dict, var=var_dict, uns=uns_dict)
     assert not isinstance(adata_src.obs["oanno1"].dtype, pd.CategoricalDtype)
-    adata_src.raw = adata_src
+    adata_src.raw = adata_src.copy()
 
     if storage == "h5ad":
         adata_src.write(backing_h5ad, **dataset_kwargs)
@@ -242,7 +242,7 @@ def test_readwrite_equivalent_h5ad_zarr(tmp_path, typ):
 
     M, N = 100, 101
     adata = gen_adata((M, N), X_type=typ)
-    adata.raw = adata
+    adata.raw = adata.copy()
 
     adata.write_h5ad(h5ad_pth)
     adata.write_zarr(zarr_pth)

tests/test_transpose.py

@@ -24,7 +24,7 @@ def test_transpose_orig():
 
 def _add_raw(adata, *, var_subset=slice(None)):
     new = adata[:, var_subset].copy()
-    new.raw = adata
+    new.raw = adata.copy()
     return new
 
 

tests/test_views.py

@@ -525,7 +525,7 @@ def test_layers_view():
 # TODO: This can be flaky. Make that stop
 def test_view_of_view(matrix_type, subset_func, subset_func2):
     adata = gen_adata((30, 15), X_type=matrix_type)
-    adata.raw = adata
+    adata.raw = adata.copy()
     if subset_func is single_subset:
         pytest.xfail("Other subset generating functions have trouble with this")
     var_s1 = subset_func(adata.var_names, min_size=4)