fixed logging for representation choice in UMAP

scanpy/__init__.py

@@ -28,6 +28,7 @@
 del get_versions, check_versions
 
 # the actual API
+from ._settings import settings  # start with settings as several tools are using it
 from . import tools as tl
 from . import preprocessing as pp
 from . import plotting as pl
@@ -37,7 +38,6 @@
 from anndata import read_h5ad, read_csv, read_excel, read_hdf, read_loom, read_mtx, read_text, read_umi_tools
 from .readwrite import read, read_10x_h5, read_10x_mtx, write
 from .neighbors import Neighbors
-from ._settings import settings
 
 set_figure_params = settings.set_figure_params
 

scanpy/tools/_umap.py

@@ -121,7 +121,10 @@ def umap(
     n_epochs = 0 if maxiter is None else maxiter
     verbosity = _VERBOSITY_LEVELS_FROM_STRINGS.get(settings.verbosity, settings.verbosity)
     neigh_params = adata.uns['neighbors']['params']
-    X = choose_representation(adata, neigh_params.get('use_rep', None), neigh_params.get('n_pcs', None))
+    X = choose_representation(
+        adata, neigh_params.get('use_rep', None), neigh_params.get('n_pcs', None), silent=True)
+    # the data matrix X is really only used for determining the number of connected components
+    # for the init condition in the UMAP embedding
     X_umap = simplicial_set_embedding(
         X,
         adata.uns['neighbors']['connectivities'].tocoo(),

scanpy/tools/_utils.py

@@ -2,6 +2,7 @@
 from .. import logging as logg
 from ._pca import pca
 from ..preprocessing._simple import N_PCS
+from .. import settings
 
 doc_use_rep = """\
 use_rep : {`None`, 'X'} or any key for `.obsm`, optional (default: `None`)
@@ -16,7 +17,10 @@
 """
 
 
-def choose_representation(adata, use_rep=None, n_pcs=None):
+def choose_representation(adata, use_rep=None, n_pcs=None, silent=False):
+    verbosity = settings.verbosity
+    if silent and settings.verbosity > 1:
+        settings.verbosity = 1
     if use_rep is None and n_pcs == 0:  # backwards compat for specifying `.X`
         use_rep = 'X'
     if use_rep is None:
@@ -26,9 +30,7 @@ def choose_representation(adata, use_rep=None, n_pcs=None):
                     raise ValueError(
                         '`X_pca` does not have enough PCs. Rerun `sc.pp.pca` with adjusted `n_comps`.')
                 X = adata.obsm['X_pca'][:, :n_pcs]
-                logg.info('    using \'X_pca\' with n_pcs = {}'
-                          .format(X.shape[1]))
-                return X
+                logg.info('    using \'X_pca\' with n_pcs = {}'.format(X.shape[1]))
             else:
                 logg.warn(
                     'You\'re trying to run this on {} dimensions of `.X`, '
@@ -37,19 +39,20 @@ def choose_representation(adata, use_rep=None, n_pcs=None):
                     .format(adata.n_vars))
                 X = pca(adata.X)
                 adata.obsm['X_pca'] = X[:, :n_pcs]
-                return X
         else:
             logg.info('    using data matrix X directly')
-            return adata.X
+            X = adata.X
     else:
         if use_rep in adata.obsm.keys():
-            return adata.obsm[use_rep]
+            X = adata.obsm[use_rep]
         elif use_rep == 'X':
-            return adata.X
+            X = adata.X
         else:
             raise ValueError(
                 'Did not find {} in `.obsm.keys()`. '
                 'You need to compute it first.'.format(use_rep))
+    settings.verbosity = verbosity  # resetting verbosity
+    return X
 
 
 def preprocess_with_pca(adata, n_pcs=None, random_state=0):