add fuzzy trigram searching to add-on store

[seanbudd]

Link to issue number:

None

Summary of the issue:

Add-on store list searching uses an exact search, i.e. it doesn't allow for typos or similar text to be found.
This also means that we cannot sort by search results relevance.
Instead a trigram search can be used to improve the search usability of the add-on store.

Description of user facing changes:

When searching add-ons, add-ons are filtered and ordered by search results relevance. The order can be updated after searching, and the previous order will restore if the search is cleared.

Description of developer facing changes:

None

Description of development approach:

• Introduced a trigram-based fuzzy search algorithm for add-on filtering, providing more relevant search results even with partial or misspelled queries. Added a new searchRank field to AddonListField and made it the default sorting method when a search is active. [1] [2] [3]
• Updated sorting logic throughout the GUI to use a new sortableFields property, which includes the new "Search relevance" option, and ensured that sorting and column selection are synchronized with the current search/filter state. [1] [2] [3] [4] [5]
• When a search filter is applied or cleared, the code now updates both the ViewModel and UI selection to avoid inconsistencies and ensures the sort field reflects the current context (e.g., switches to "Search relevance" when searching).
• Added logic to clear both ViewModel and UI selection when the search filter changes, preventing multiple rows from appearing selected and ensuring a single source of truth for selection.

Testing strategy:

Tested manually by filtering the add-on store

Known issues with pull request:

None

Code Review Checklist:

• Documentation:
  • Change log entry
  • User Documentation
  • Developer / Technical Documentation
  • Context sensitive help for GUI changes
• Testing:
  • Unit tests
  • System (end to end) tests
  • Manual testing
• UX of all users considered:
  • Speech
  • Braille
  • Low Vision
  • Different web browsers
  • Localization in other languages / culture than English
• API is compatible with existing add-ons.
• Security precautions taken.

[Copilot]

Pull request overview

This PR adds trigram-based fuzzy search functionality to the NVDA add-on store, improving search usability by allowing typos and partial matches. The search results are now filtered and ordered by relevance, with the ability to restore the previous sort order when the search is cleared.

Key changes:

• Implemented trigram-based similarity scoring for add-on search with a threshold of 0.3
• Added "Search relevance" as a sortable field that becomes the default when searching
• Automatically switches sort field to "Search relevance" when search is active and reverts to previous sort when cleared

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 5 comments.

File	Description
source/gui/addonStoreGui/viewModels/addonList.py	Adds trigram search implementation, searchRank field, searchableText property, and logic to handle sort field transitions
source/gui/addonStoreGui/controls/storeDialog.py	Updates UI to clear selection on search, switch to search relevance sorting, and synchronize column filter display
source/gui/addonStoreGui/controls/addonList.py	Updates column click handling to use sortableFields instead of presentedFields

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 5 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[CyrilleB79]

Warning: strxfrm is not a normalization function.
It just provides a string that can be used as a key for locale-aware comparisons. Did you really intend to use this function?

[seanbudd]

yes, I believe it works as expected when performing trigram comparisons. e.g. ensuring an accented character matches the character itself.

[CyrilleB79]

>>> strxfrm("Sean")
'\x0e\x91\x0e!\x0e\x02\x0ep\x01\x01\x12\x01\x01'
>>> strxfrm("Séan")
'\x0e\x91\x0e!\x0e\x02\x0ep\x01\x02\x02\x0e\x01\x12\x01\x01'

Will trigram search work on these strings?

Also calling the result of strxfrm normalized is misleading.

[seanbudd]

yes, it will match by each char, in theory it should provide much better matching

[seanbudd]

I don't think it's misleading. The conversion and stripping normalizes the text for the purpose of trigram searches

[CyrilleB79]

Here are points of feedback after my first tests:

• "福恩", the author of subtitle reader, cannot be found.
• Assertion error when typing 3 spaces in the search field, or "AI " (e.g. when typing "AI image generator")
• the add-on list is very slow to navigate with regular 1-second freeze
• Maybe, if not already done, a more important weight should be given to the add-on's name, e.g. with respect to description, so that, for example, "Check Input Gestures" or "Enhanced Touch Gestures" appear on the top of the results.
• Seems that you have implemented the algorithm manually. Why didn't you use a dedicated library for this?