[orca] Improve logic to filter redundant names

[Joanmarie Diggs <joanied src gnome org>]

commit 6915e1ea55fe59d3e55eb302b95e2048eca1340f
Author: Joanmarie Diggs <jdiggs igalia com>
Date:   Tue Apr 12 12:32:21 2022 +0200

    Improve logic to filter redundant names
    
    The logic to filter out names which were redundant to labels handled
    the following conditions:
    
    * The name and label only differed with respect to whitespace
    * The label started with the text of the name (e.g. because the
      author/developer wanted to append some helpful information)
    
    It did not, however, handle the following:
    
    * The name and label only differed with respect to punctuation
    * The name itself had duplicate text (e.g. repeats the label twice)
    
    This failure makes Orca quite chatty with Gtk 4 widgets where the
    use_underline attribute is present because Gtk 4 exposes a name
    consisting of the displayed text with the underline included
    followed by that same text without the underline included.
    
    This commit handles those two missing scenarios.
    
    Fixes #237.

 src/orca/generator.py | 25 ++++++++++++++++++++-----
 1 file changed, 20 insertions(+), 5 deletions(-)
---
diff --git a/src/orca/generator.py b/src/orca/generator.py
index ebcdaaa22..54aa2710e 100644
--- a/src/orca/generator.py
+++ b/src/orca/generator.py
@@ -28,6 +28,7 @@ __copyright__ = "Copyright (c) 2009 Sun Microsystems Inc." \
 __license__   = "LGPL"
 
 import pyatspi
+import re
 import sys
 import time
 import traceback
@@ -396,12 +397,26 @@ class Generator:
             descendant = self._script.utilities.realActiveDescendant(obj)
             name = self._generateName(descendant)
 
+        # If we don't have a label, always use the name.
+        if not label:
+            return name
+
         result.extend(label)
-        if not len(label):
-            result.extend(name)
-        elif len(name) and name[0].split() != label[0].split() \
-             and not label[0].startswith(name[0]):
-            result.extend(name)
+        if not name:
+            return result
+
+        # Try to eliminate names which are redundant to the label.
+        # Convert all non-alphanumeric characters to space and get the words.
+        nameWords = re.sub(r"[\W_]", " ", name[0]).split()
+        labelWords = re.sub(r"[\W_]", " ", label[0]).split()
+
+        # If all of the words in the name are in the label, the name is redundant.
+        if set(nameWords).issubset(set(labelWords)):
+            msg = "GENERATOR: name '%s' is redundant to label '%s'" % (name[0], label[0])
+            debug.println(debug.LEVEL_INFO, msg, True)
+            return result
+
+        result.extend(name)
         return result
 
     def _generateLabelOrName(self, obj, **args):