scripts: kconfig: improve search results ordering using scoring

This implements the same improved sorting algorithm as was recently
added to the documentation (see zephyrproject-rtos#98016),
namely:

- A match in a Kconfig symbol's name is given more weight than a match
  in its prompt.
- Field-length normalization is applied so that the shorter the field,
  the higher its relevance (e.g. searching for "sensor" will now
  basically yield CONFIG_SENSOR as the top result as the query basically
  matches 100% of the symbol's name.

Signed-off-by: Benjamin Cabé <[email protected]>

Author: kartben

scripts/kconfig/config_utils.py

@@ -0,0 +1,74 @@
+#!/usr/bin/env python3
+
+# SPDX-FileCopyrightText: Copyright The Zephyr Project Contributors
+# SPDX-License-Identifier: Apache-2.0
+
+"""
+Common utilities for Kconfig configuration interfaces.
+
+This module provides shared functionality for menuconfig.py and guiconfig.py.
+"""
+
+import re
+
+from kconfiglib import Symbol, Choice
+
+
+def score_search_matches(search_str, nodes):
+    """
+    Scores and sorts search results for Kconfig nodes based on relevance.
+
+    This implements a basic scoring system where:
+    - A match in a symbol's name is given more weight than a match in its prompt
+    - Field-length normalization is applied so that the shorter the field, the higher its relevance
+
+    Args:
+        search_str: The search string (space-separated regexes)
+        nodes: List of MenuNode objects to search through
+
+    Returns:
+        List of tuples (node, score) sorted by score (highest first)
+    """
+    # Parse the search string into regexes
+    try:
+        regexes = [re.compile(regex.lower()) for regex in search_str.split()]
+        # If no regexes, all nodes match, order is unchanged
+        if len(regexes) == 0:
+            return [(node, 1) for node in nodes]
+    except re.error:
+        # Invalid regex - return empty results
+        return []
+
+    NAME_WEIGHT = 2.0
+    PROMPT_WEIGHT = 1.0
+
+    scored_results = []
+
+    for node in nodes:
+        # Get lowercase versions for matching
+        name = ""
+        if isinstance(node.item, (Symbol, Choice)):
+            name = node.item.name.lower() if node.item.name else ""
+
+        prompt = node.prompt[0].lower() if node.prompt else ""
+
+        # Check if all regexes match in either name or prompt
+        name_matches = name and all(regex.search(name) for regex in regexes)
+        prompt_matches = prompt and all(regex.search(prompt) for regex in regexes)
+
+        if not (name_matches or prompt_matches):
+            continue
+
+        # Apply field-length normalization (shorter fields = higher relevance)
+        score = 0
+        if name_matches:
+            score += NAME_WEIGHT / (len(name) ** 0.5)
+        if prompt_matches:
+            score += PROMPT_WEIGHT / (len(prompt) ** 0.5)
+
+        scored_results.append((node, score))
+
+    # Sort by score (highest first)
+    scored_results.sort(key=lambda x: x[1], reverse=True)
+
+    return scored_results

scripts/kconfig/guiconfig.py

@@ -87,6 +87,8 @@
                        TRI_TO_STR, TYPE_TO_STR, \
                        standard_kconfig, standard_config_filename
 
+from config_utils import score_search_matches
+
 
 # If True, use GIF image data embedded in this file instead of separate GIF
 # files. See _load_images().
@@ -1859,12 +1861,11 @@ def _update_jump_to_matches(msglabel, search_string):
     _jump_to_tree.selection_set(())
 
     try:
-        # We could use re.IGNORECASE here instead of lower(), but this is
-        # faster for regexes like '.*debug$' (though the '.*' is redundant
-        # there). Those probably have bad interactions with re.search(), which
-        # matches anywhere in the string.
-        regex_searches = [re.compile(regex).search
-                          for regex in search_string.lower().split()]
+        scored_sc_nodes = score_search_matches(search_string, _sorted_sc_nodes())
+        scored_menu_comment_nodes = score_search_matches(search_string, _sorted_menu_comment_nodes())
+
+        _jump_to_matches = [node for node, _ in scored_sc_nodes + scored_menu_comment_nodes]
+
     except re.error as e:
         msg = "Bad regular expression"
         # re.error.msg was added in Python 3.5
@@ -1875,39 +1876,6 @@ def _update_jump_to_matches(msglabel, search_string):
         _jump_to_tree.set_children("")
         return
 
-    _jump_to_matches = []
-    add_match = _jump_to_matches.append
-
-    for node in _sorted_sc_nodes():
-        # Symbol/choice
-        sc = node.item
-
-        for search in regex_searches:
-            # Both the name and the prompt might be missing, since
-            # we're searching both symbols and choices
-
-            # Does the regex match either the symbol name or the
-            # prompt (if any)?
-            if not (sc.name and search(sc.name.lower()) or
-                    node.prompt and search(node.prompt[0].lower())):
-
-                # Give up on the first regex that doesn't match, to
-                # speed things up a bit when multiple regexes are
-                # entered
-                break
-
-        else:
-            add_match(node)
-
-    # Search menus and comments
-
-    for node in _sorted_menu_comment_nodes():
-        for search in regex_searches:
-            if not search(node.prompt[0].lower()):
-                break
-        else:
-            add_match(node)
-
     msglabel["text"] = "" if _jump_to_matches else "No matches"
 
     _update_jump_to_display()

scripts/kconfig/menuconfig.py

@@ -223,6 +223,8 @@
                        TRI_TO_STR, TYPE_TO_STR, \
                        standard_kconfig, standard_config_filename
 
+from config_utils import score_search_matches
+
 
 #
 # Configuration variables
@@ -2078,54 +2080,13 @@ def select_prev_match():
             prev_s = s
 
             try:
-                # We could use re.IGNORECASE here instead of lower(), but this
-                # is noticeably less jerky while inputting regexes like
-                # '.*debug$' (though the '.*' is redundant there). Those
-                # probably have bad interactions with re.search(), which
-                # matches anywhere in the string.
-                #
-                # It's not horrible either way. Just a bit smoother.
-                regex_searches = [re.compile(regex).search
-                                  for regex in s.lower().split()]
-
-                # No exception thrown, so the regexes are okay
+                # Use the scoring function for symbols and choices
                 bad_re = None
 
-                # List of matching nodes
-                matches = []
-                add_match = matches.append
-
-                # Search symbols and choices
-
-                for node in _sorted_sc_nodes():
-                    # Symbol/choice
-                    sc = node.item
-
-                    for search in regex_searches:
-                        # Both the name and the prompt might be missing, since
-                        # we're searching both symbols and choices
-
-                        # Does the regex match either the symbol name or the
-                        # prompt (if any)?
-                        if not (sc.name and search(sc.name.lower()) or
-                                node.prompt and search(node.prompt[0].lower())):
-
-                            # Give up on the first regex that doesn't match, to
-                            # speed things up a bit when multiple regexes are
-                            # entered
-                            break
-
-                    else:
-                        add_match(node)
-
-                # Search menus and comments
+                scored_sc_nodes = score_search_matches(s, _sorted_sc_nodes())
+                scored_menu_comment_nodes = score_search_matches(s, _sorted_menu_comment_nodes())
 
-                for node in _sorted_menu_comment_nodes():
-                    for search in regex_searches:
-                        if not search(node.prompt[0].lower()):
-                            break
-                    else:
-                        add_match(node)
+                matches = [node for node, _ in scored_sc_nodes + scored_menu_comment_nodes]
 
             except re.error as e:
                 # Bad regex. Remember the error message so we can show it.