@@ -395,68 +395,6 @@ def test_dotplot_style_no_reset():
395395 assert plot .cmap == "winter" , "style() should not reset unspecified parameters"
396396
397397
398- @pytest .mark .parametrize (
399- "plot_fn" ,
400- [sc .pl .dotplot , sc .pl .matrixplot ],
401- ids = ["dotplot" , "matrixplot" ],
402- )
403- @pytest .mark .parametrize (
404- ("values" , "expected_suffixes" ),
405- [
406- pytest .param (
407- pd .Categorical (np .tile ([1 , 2 ], 350 )), ("_1" , "_2" ), id = "int_categorical"
408- ),
409- pytest .param (np .tile ([0.5 , 2.0 ], 350 ), ("_0.5" , "_2.0" ), id = "numeric" ),
410- ],
411- )
412- def test_multi_groupby_nonstring (plot_fn , values , expected_suffixes ):
413- """Multi-column `groupby` accepts non-string columns."""
414- adata = pbmc68k_reduced ()
415- adata .obs ["num" ] = values
416- plot = plot_fn (adata , adata .var_names [:4 ], ["bulk_labels" , "num" ], return_fig = True )
417- categories = list (plot .categories )
418- assert len (categories ) == 2 * adata .obs ["bulk_labels" ].nunique ()
419- first_label = adata .obs ["bulk_labels" ].cat .categories [0 ]
420- assert categories [:2 ] == [f"{ first_label } { s } " for s in expected_suffixes ]
421-
422-
423- @pytest .mark .parametrize (
424- "plot_fn" ,
425- [sc .pl .dotplot , sc .pl .matrixplot ],
426- ids = ["dotplot" , "matrixplot" ],
427- )
428- def test_multi_groupby_partial_nan (plot_fn ):
429- """Rows missing any of several `groupby` values get no joined category."""
430- adata = pbmc68k_reduced ()
431- labels = adata .obs ["bulk_labels" ].astype (object )
432- labels .iloc [:10 ] = np .nan
433- adata .obs ["part_nan" ] = pd .Categorical (labels )
434- plot = plot_fn (adata , adata .var_names [:4 ], ["part_nan" , "phase" ], return_fig = True )
435- assert not any ("nan" in str (c ) for c in plot .categories )
436-
437-
438- @pytest .mark .parametrize (
439- "plot_fn" ,
440- [sc .pl .dotplot , sc .pl .matrixplot ],
441- ids = ["dotplot" , "matrixplot" ],
442- )
443- def test_multi_groupby_mixed_type (plot_fn ):
444- """Multi-column `groupby` accepts object columns mixing types and missing values."""
445- adata = pbmc68k_reduced ()
446- mixed = np .empty (len (adata ), dtype = object )
447- mixed [::2 ] = "control"
448- mixed [1 ::2 ] = 1
449- mixed [:10 ] = np .nan
450- adata .obs ["mixed" ] = mixed
451- plot = plot_fn (
452- adata , adata .var_names [:4 ], ["bulk_labels" , "mixed" ], return_fig = True
453- )
454- categories = [str (c ) for c in plot .categories ]
455- assert not any ("nan" in c for c in categories )
456- assert any (c .endswith ("_control" ) for c in categories )
457- assert any (c .endswith ("_1" ) for c in categories )
458-
459-
460398def test_dotplot_add_totals (plot_cmp ):
461399 pbmc = pbmc68k_reduced ()
462400 markers = {"T-cell" : "CD3D" , "B-cell" : "CD79A" , "myeloid" : "CST3" }
@@ -1750,10 +1688,27 @@ def test_groupby_index(plot_cmp):
17501688 plot_cmp ("dotplot_groupby_index" )
17511689
17521690
1753- def test_groupby_list () -> None :
1754- """Test category order when groupby is a list.
1691+ def _mixed_with_nan (n : int ) -> np .ndarray :
1692+ col = np .array (["control" , 1 ] * (n // 2 ) + ["control" ] * (n % 2 ), dtype = object )
1693+ col [:10 ] = np .nan
1694+ return col
1695+
1696+
1697+ @pytest .mark .parametrize (
1698+ "make_other" ,
1699+ [
1700+ pytest .param (lambda a : a .obs ["cell_type" ], id = "categorical" ),
1701+ pytest .param (
1702+ lambda a : pd .Categorical (np .arange (a .n_obs ) % 2 + 1 ), id = "int_categorical"
1703+ ),
1704+ pytest .param (lambda a : np .where (np .arange (a .n_obs ) % 2 , 2.0 , 0.5 ), id = "float" ),
1705+ pytest .param (lambda a : _mixed_with_nan (a .n_obs ), id = "mixed_with_nan" ),
1706+ ],
1707+ )
1708+ def test_groupby_list (make_other : Callable [[AnnData ], object ]) -> None :
1709+ """Test category order when groupby is a list, incl. non-string and missing values.
17551710
1756- See <https://github.com/scverse/scanpy/issues/1735>
1711+ See <https://github.com/scverse/scanpy/issues/1735> and #4384.
17571712 """
17581713 rng = np .random .default_rng ()
17591714 adata = krumsiek11 ()
@@ -1762,25 +1717,28 @@ def test_groupby_list() -> None:
17621717 cats = adata .obs .cell_type .cat .categories .tolist ()
17631718 rng .shuffle (cats )
17641719 adata .obs ["rand_cat" ] = pd .Categorical (cat_val , categories = cats )
1720+ adata .obs ["other" ] = make_other (adata )
17651721
17661722 plot = sc .pl .dotplot (
17671723 adata ,
17681724 ["Gata1" , "Gata2" ],
1769- ["rand_cat" , "cell_type " ],
1725+ ["rand_cat" , "other " ],
17701726 swap_axes = True ,
17711727 show = False ,
17721728 return_fig = True ,
17731729 )
17741730 plot .make_figure ()
17751731 ticks = [t .get_text () for t in plot .get_axes ()["mainplot_ax" ].get_xticklabels ()]
17761732
1777- # each group’s categories determine the tick order, unobserved pairs are dropped
1778- observed = set (zip (adata .obs .rand_cat , adata .obs .cell_type , strict = True ))
1733+ # each group’s categories determine the tick order,
1734+ # unobserved pairs and rows with missing values are dropped
1735+ other = adata .obs ["other" ]
1736+ observed = set (zip (adata .obs .rand_cat , other , strict = True ))
17791737 assert ticks == [
1780- f"{ rand } _{ cell } "
1738+ f"{ rand } _{ o } "
17811739 for rand in cats
1782- for cell in adata . obs . cell_type .cat .categories
1783- if (rand , cell ) in observed
1740+ for o in other . astype ( "category" ) .cat .categories
1741+ if (rand , o ) in observed
17841742 ]
17851743
17861744
0 commit comments