Dandelion io (#483)

* incorporate dandelion awkward array changes

* update docstrings

* merge master changes, update dandelion test fixture

* fix formatting

* enforce dandelion >=0.3.5 version in toml

* fixed faulty reference to dandelion function in docstring

pyproject.toml

@@ -79,7 +79,7 @@ test = [
     'black',
 ]
 dandelion = [
-    'sc-dandelion>=0.2.3',
+    'sc-dandelion>=0.3.5',
 ]
 diversity = [
     'scikit-bio>=0.5.7'

src/scirpy/io/_io.py

@@ -581,47 +581,32 @@ def write_airr(adata: DataHandler.TYPE, filename: Union[str, Path], **kwargs) ->
     writer.close()
 
 
-def to_dandelion(adata: DataHandler.TYPE, **kwargs):
+def to_dandelion(adata: DataHandler.TYPE):
     """Export data to `Dandelion <https://github.com/zktuong/dandelion>`_ (:cite:`Stephenson2021`).
 
     Parameters
     ----------
     adata
         annotated data matrix with :term:`IR` annotations.
-    **kwargs
-        additional arguments passed to :func:`~scirpy.io.to_airr_cells`
 
     Returns
     -------
     `Dandelion` object.
     """
     try:
-        import dandelion as ddl
+        from dandelion import from_scirpy
     except ImportError:
         raise ImportError("Please install dandelion: pip install sc-dandelion.") from None
-    airr_cells = to_airr_cells(adata, **kwargs)
-
-    contig_dicts = {}
-    for tmp_cell in airr_cells:
-        for i, chain in enumerate(tmp_cell.to_airr_records(), start=1):
-            # dandelion-specific modifications
-            chain.update(
-                {
-                    "sequence_id": f"{tmp_cell.cell_id}_contig_{i}",
-                }
-            )
-            contig_dicts[chain["sequence_id"]] = chain
 
-    data = pd.DataFrame.from_dict(contig_dicts, orient="index")
-    return ddl.Dandelion(ddl.load_data(data))
+    return from_scirpy(adata)
 
 
 @_doc_params(doc_working_model=doc_working_model)
-def from_dandelion(dandelion, transfer: bool = False, **kwargs) -> AnnData:
+def from_dandelion(dandelion, transfer: bool = False, to_mudata: bool = False, **kwargs) -> AnnData:
     """\
     Import data from `Dandelion <https://github.com/zktuong/dandelion>`_ (:cite:`Stephenson2021`).
 
-    Internally calls :func:`scirpy.io.read_airr`.
+    Internally calls `dandelion.to_scirpy`.
 
     {doc_working_model}
 
@@ -632,29 +617,22 @@ def from_dandelion(dandelion, transfer: bool = False, **kwargs) -> AnnData:
     transfer
         Whether to execute `dandelion.tl.transfer` to transfer all data
         to the :class:`anndata.AnnData` instance.
+    to_mudata
+        Return MuData object instead of AnnData object.
     **kwargs
-        Additional arguments passed to :func:`scirpy.io.read_airr`.
+        Additional arguments passed to `dandelion.to_scirpy`.
 
     Returns
     -------
     AnnData object with :term:`AIRR` data in `obsm["airr"]` for each cell. For more details see
     :ref:`data-structure`.
     """
     try:
-        import dandelion as ddl
+        from dandelion import to_scirpy
     except ImportError:
         raise ImportError("Please install dandelion: pip install sc-dandelion.") from None
 
-    dandelion_df = dandelion.data.copy()
-    # replace "unassigned" with None
-    for col in dandelion_df.columns:
-        dandelion_df.loc[dandelion_df[col] == "unassigned", col] = None
-
-    adata = read_airr(dandelion_df, **kwargs)
-
-    if transfer:
-        ddl.tl.transfer(adata, dandelion)  # need to make a version that is not so verbose?
-    return adata
+    return to_scirpy(dandelion, transfer=transfer, to_mudata=to_mudata, **kwargs)
 
 
 @_doc_params(doc_working_model=doc_working_model)

src/scirpy/tests/test_io.py

@@ -245,7 +245,7 @@ def test_write_airr_none_field_issue_454(tmp_path):
     write_airr(adata, tmp_path / "test.airr.tsv")
 
 
-@pytest.mark.xfail(reason="Dandelion still uses `duplicate_count` instead of `umi_count`", raises=AssertionError)
+# @pytest.mark.xfail(reason="Dandelion still uses `duplicate_count` instead of `umi_count`", raises=AssertionError)
 @pytest.mark.extra
 @pytest.mark.parametrize(
     "anndata_from_10x_sample",