Add 'keep_var' variable to make sure variable names don't disappear

Author: benjamin-james

benj/aggregate.py

@@ -134,6 +134,7 @@ def aggregate_concat(metadata=None, directory:Union[_PathLike, List[_PathLike]]=
                      sample_key="Sample", calc_qc:bool=True,
                      min_cells_per_sample:int=30,
                      sep="\t", verbose:bool=True,
+                     keep_var:bool=True
                      **kwargs):
     """Metadata+directory, or h5ad with or without metadata"""
     import os
@@ -190,7 +191,7 @@ def aggregate_concat(metadata=None, directory:Union[_PathLike, List[_PathLike]]=
             else:
                 if var_same is None:
                     var_same = adata.var.copy()
-                else:
+                elif not keep_var:
                     ### Idea: Keep one .var that keeps track of when the same items are there
                     ### If the column values are all the same, save and delete in all anndata
                     ###
@@ -222,8 +223,9 @@ def aggregate_concat(metadata=None, directory:Union[_PathLike, List[_PathLike]]=
         tk = list(adata_tbl.keys())
         del adata_tbl
         gc.collect()
-        for vn in var_same.columns:
-            adata.var[vn] = var_same[vn]
+        if not keep_var:
+            for vn in var_same.columns:
+                adata.var[vn] = var_same[vn]
         if len(tk) == len(scrub_tbl.keys()):
             adata.uns["scrublet"] = {"batches": scrub_tbl,
                                      "batched_by": sample_key}

benj/utils.py

@@ -6,15 +6,18 @@ def pg(adata, color_map="Reds"):
 
 def read_elems(path: str, elems: Union[str, List[str]], retry:int=2) -> Union[Dict[str, any], any]:
     import h5py
-    import anndata.experimental
+    try:
+        from anndata.io import read_elem
+    except ModuleNotFoundError:
+        from anndata.experimental import read_elem
     if retry < 0:
         return {}
     try:
         with h5py.File(path, "r") as F:
             if isinstance(elems, str):
-                return anndata.experimental.read_elem(F[elems])
+                return read_elem(F[elems])
             else:
-                return {elem: anndata.experimental.read_elem(F[elem]) for elem in elems}
+                return {elem: read_elem(F[elem]) for elem in elems}
     except KeyError:
         import time
         print("Key error for %s, sleeping 5 seconds and retrying..." % path)
@@ -140,6 +143,7 @@ def weighted_pearson_correlation(A, B, wt=None):
 
 def is_nonzero_combinatorial(adata, genes, label:str="Combination", prefix:str="Combo ", inplace:bool=True):
     import numpy as np
+    import pandas as pd
     import scipy.sparse
     I = adata.var_names.get_indexer(genes)
     X = adata.X[:, I]