fix error msg

.gitignore

@@ -1,3 +1,4 @@
+devel/data/pbmc3k_raw.h5ad
 *.kate-swp
 .vscode
 .pytest_cache

devel/sandbox_scanpy.py

@@ -0,0 +1,65 @@
+# Based on the code contributed by @gokceneraslan
+# see https://github.com/gagolews/genieclust/issues/71
+
+
+# pip install scanpy
+# pip install louvain
+# pip install scikit-misc
+
+
+import scanpy as sc
+sc.set_figure_params(dpi=100)
+ad = sc.datasets.pbmc3k()
+sc.pp.filter_genes(ad, min_cells=10)
+ad.layers['counts'] = ad.X.copy()
+sc.pp.normalize_total(ad, target_sum=10000)
+sc.pp.log1p(ad)
+sc.pp.highly_variable_genes(ad, n_top_genes=1000, flavor='seurat_v3', subset=True, layer='counts')
+sc.pp.scale(ad, max_value=8)
+sc.pp.pca(ad)
+sc.pp.neighbors(ad)
+
+sc.tl.umap(ad)
+
+sc.tl.louvain(ad, resolution=0.2)
+
+X_hidim = ad.X
+X_lodim = ad.obsm['X_pca']
+
+
+import genieclust
+import numpy as np
+
+g = genieclust.Genie(n_clusters=3, affinity='cosine')
+labels = g.fit_predict(X_hidim)
+ad.obs['genie_labels'] = labels.astype(str)
+sc.pl.umap(ad, color='genie_labels')
+
+g = genieclust.Genie(n_clusters=3, affinity='cosine')
+labels = g.fit_predict(X_lodim)
+ad.obs['genie_labels'] = labels.astype(str)
+sc.pl.umap(ad, color='genie_labels')
+
+sc.pl.umap(ad, color='louvain')
+
+
+mst = genieclust.internal.mst_from_distance(X_hidim, 'cosine')
+genieclust.plots.plot_segments(mst[1], ad.obsm["X_umap"])
+
+
+X_hidim_std = (X_hidim-X_hidim.mean(axis=0))/(X_hidim.std(axis=0))
+g = genieclust.Genie(n_clusters=3, affinity='cosine', M=5, postprocess="merge")
+labels = g.fit_predict(X_hidim_std)
+ad.obs['genie_labels_std'] = labels.astype(str)
+sc.pl.umap(ad, color='genie_labels_std')
+
+
+mst = genieclust.internal.mst_from_distance(X_hidim_std)
+genieclust.plots.plot_segments(mst[1], ad.obsm["X_umap"])
+
+
+
+mst = genieclust.internal.mst_from_distance(np.array(X_lodim, copy=True, order='C'))
+#sc.pl.umap(ad, color='genie_labels')
+genieclust.plots.plot_segments(mst[1], ad.obsm["X_umap"])
+

genieclust/genie.py

@@ -202,7 +202,7 @@ def _check_params(self, cur_state=None):
         _postprocess_options = ("boundary", "none", "all")
         cur_state["postprocess"] = str(self.postprocess).lower()
         if cur_state["postprocess"] not in _postprocess_options:
-            raise ValueError("`postprocess` should be one of %r" % _postprocess_options)
+            raise ValueError("`postprocess` should be one of %s" % repr(_postprocess_options))
 
         cur_state["affinity"] = str(self.affinity).lower()
         if cur_state["affinity"] in ["euclidean", "lp:p=2"]:
@@ -227,7 +227,7 @@ def _check_params(self, cur_state=None):
         _affinity_exact_options = (
             "l2", "l1", "cosinesimil", "precomputed")
         if cur_state["exact"] and cur_state["affinity"] not in _affinity_exact_options:
-            raise ValueError("`affinity` should be one of %r" % _affinity_exact_options)
+            raise ValueError("`affinity` should be one of %s" % repr(_affinity_exact_options))
 
         if type(self.mlpack_enabled) is str:
             cur_state["mlpack_enabled"] = str(self.mlpack_enabled).lower()

src/c_mst.h

@@ -315,24 +315,24 @@ ssize_t Cmst_from_nn(
 
 /*! Determine the first k nearest neighbours of each point.
  *
- *  Exactly n*(n-1) distance computations are performed.
+ *  Exactly n*(n-1)/2 distance computations are performed.
  *
  *  It is assumed that each query point is not its own neighbour.
  *
- * Worst-case time complexity: O(n*(n-1)/2*d*k)
+ *  Worst-case time complexity: O(n*(n-1)/2*d*k)
  *
  *
- * @param D a callable CDistance object such that a call to
- *        <T*>D(j, <ssize_t*>M, ssize_t l) returns an n-ary array
- *        with the distances from the j-th point to l points whose indices
- *        are given in array M
- * @param n number of points
- * @param k number of nearest neighbours,
- * @param dist [out]  a c_contiguous array, shape (n,k),
- *        dist[i,j] gives the weight of the (undirected) edge {i, ind[i,j]}
- * @param ind [out]   a c_contiguous array, shape (n,k),
- *        (undirected) edge definition, interpreted as {i, ind[i,j]}
- * @param verbose output diagnostic/progress messages?
+ *  @param D a callable CDistance object such that a call to
+ *         <T*>D(j, <ssize_t*>M, ssize_t l) returns an n-ary array
+ *         with the distances from the j-th point to l points whose indices
+ *         are given in array M
+ *  @param n number of points
+ *  @param k number of nearest neighbours,
+ *  @param dist [out]  a c_contiguous array, shape (n,k),
+ *         dist[i,j] gives the weight of the (undirected) edge {i, ind[i,j]}
+ *  @param ind [out]   a c_contiguous array, shape (n,k),
+ *         (undirected) edge definition, interpreted as {i, ind[i,j]}
+ *  @param verbose output diagnostic/progress messages?
  */
 template <class T>
 void Cknn_from_complete(CDistance<T>* D, ssize_t n, ssize_t k,