Fix irrelevant search results for short queries like GSoC

[MuhammadAashirAslam]

Fixes #733

Summary

Improve search precision for short single-word queries (e.g. gsoc) by preventing irrelevant matches from XML code tokens — implemented as a client-side query-time adjustment, without modifying indexing or documentation content.

---

Problem

While testing search locally, I noticed that short queries like gsoc were returning unrelated XML reference results.

This was happening due to Algolia's default typo tolerance behavior.

By default:

• Words with 4+ characters allow typos
• Prefix matching is also applied

So the query:

gsoc

was matching:

<m2n:sockets>

because:

gsoc → matched "soc" (prefix of "sockets")
nbTypos: 2
nbExactWords: 0

This was confirmed via Algolia ranking info:

"_rankingInfo": {
    "nbTypos": 2,
    "nbExactWords": 0,
    "firstMatchedWord": 2000
}

Meaning the result was purely a fuzzy match from code content, not real page relevance.

---

Change

A client-side rendering fix was implemented to make short queries stricter without touching indexing or documentation.

In both Algolia client entry points:

• _includes/algolia.html
• js/algolia-search.js

the following parameters were added inside the searchFunction:

helper.setQueryParameter('minWordSizefor1Typo', 5);
helper.setQueryParameter('minWordSizefor2Typos', 9);

---

Effect

• Words of length ≤4 (like gsoc) now require an exact match
• Words ≤8 characters allow at most 1 typo (instead of 2)

This prevents short queries from matching unrelated code tokens such as:

gsoc → sockets

---

Why this approach

• No changes to _config.yml
• No changes to indexing
• No documentation/reference files modified
• No global typo tolerance disabled

This is purely a query-time adjustment in client rendering, ensuring:

• Short queries stop returning noisy XML matches
• Normal documentation search remains unaffected

---

Files Changed

File	Change
_includes/algolia.html	Added query parameter override
js/algolia-search.js	Same change

Both updates were applied inside the searchFunction.

---

Validation

After this change:

• Searching gsoc no longer returns irrelevant XML reference matches
• Other queries like configuration continue returning correct results

---

Note

This change only improves relevance for short queries at the client level. Making pages containing terms like "GSoC" fully discoverable would additionally require adding keywords to searchableAttributes in _config.yml and rebuilding the Algolia index with bundle exec jekyll algolia. This is outside the scope of this PR since it involves indexing configuration.

---

Screenshot Before

Screenshot After

[MakisH]

Thank you for your contribution. However, I am confused. Testing locally, it does not seem to have any effect, and the screenshots don't demonstrate the expected result.

General note: the PR description is quite long for what the change is (I guess suggested by some tool). This makes it tricky to see what is going on.

I see that there is a similar discussion in #741. I am not going to interact with these PRs for now, and let the discussion converge.