Simplify varying_between code.

Author: mstrazar

Orange/widgets/visualize/owvenndiagram.py

@@ -24,6 +24,7 @@
 from AnyQt.QtCore import pyqtSignal as Signal
 
 import Orange.data
+import Orange.statistics.util as util
 from Orange.widgets import widget, gui, settings
 from Orange.widgets.utils import itemmodels, colorpalette
 from Orange.widgets.utils.annotated_data import (create_annotated_table,
@@ -883,74 +884,33 @@ def unique_non_nan(ar):
     return uniq[~numpy.isnan(uniq)]
 
 
-def varying_matrix(X, idmap):
-    """
-    Enables efficient implementation of `varying_between`
-    for continuous and discrete variables.
-
-    Return indices of non-constant matrix columns between groups
-    defined by indices. Supports sparse data.
-    """
-    # Map rows to groups
-    inv_idmap = dict()
-    for g, rows in idmap.items():
-        for r in rows: inv_idmap[r] = g
-
-    # Group columns accordingly
-    vals_per_group = dict()
-    rows_per_group = dict()
-    varying = set()
-
-    # Remove columns with duplicate non-zero, non-nan values
-    for r, c in zip(*X.nonzero()):
-        g = inv_idmap[r]
-        rows_per_group[g, c] = \
-            rows_per_group.get((g, c), 0) + 1
-        val = X[r, c]
-        if isinstance(val, str) or not numpy.isnan(val):
-            if (g, c) not in vals_per_group:
-                vals_per_group[g, c] = set()
-            vals_per_group[g, c].add(val)
-            if len(vals_per_group[g, c]) > 1:
-                varying.add(c)
-
-    # Remove columns with mixed zero and non-zero values
-    for (g, c), cnt in rows_per_group.items():
-        if (g, c) in vals_per_group \
-                and len(vals_per_group[g, c]) \
-                and (0 < cnt < len(idmap[g])):
-            varying.add(c)
-    return varying
-
-
 def varying_between(table, idvar):
     """
     Return a list of all variables with non constant values between
     groups defined by `idvar`.
 
-    I.e. columns where values differ by group:
-        - If each example is its own group, this operation is trivial.
-
     """
+    all_possible = [var for var in table.domain.variables + table.domain.metas
+                    if var != idvar]
+    candidate_set = set(all_possible)
 
-    # idvar is not varying, so it would not be removed from the candidate_set
     idmap = group_table_indices(table, idvar)
-    varying = set()
-
-    # Trivial case
-    if len(idmap) == len(table):
-        return varying
-
-    varying_metas = varying_matrix(table.metas, idmap)
-    varying_data = varying_matrix(table.X, idmap)
 
-    varying = [var for vi, var in enumerate(table.domain.variables)
-                if var != idvar and vi in varying_data] + \
-              [var for vi, var in enumerate(table.domain.metas)
-                if var != idvar and vi in varying_metas]
+    varying = set()
+    for indices in idmap.values():
+        subset = table[indices]
+        for var in list(candidate_set):
+            column, _ = subset.get_column_view(var)
+            values = util.unique(column)
+
+            if var.is_string:
+                uniq = set(values)
+            else:
+                uniq = unique_non_nan(values)
 
-    all_possible = [var for var in (table.domain.variables + table.domain.metas)
-                    if var != idvar]
+            if len(uniq) > 1:
+                varying.add(var)
+                candidate_set.remove(var)
 
     return sorted(varying, key=all_possible.index)