Fix search and filter bugs against the Ballerina Central API

[daneshk]

Summary

Fixes #1853 and two related filtering bugs found while auditing search/filter behavior against the live Ballerina Central search API.

• Multi-word search returned nothing (the reported issue) — escapeSolrQuery escaped the space character, turning a query like "SAP Business" into the wildcard term *sap\ business*. The API's *...* wildcard is a literal, non-analyzed pattern, so the escaped backslash-space never matches. Space is no longer escaped.
• Area filters could silently return 0 results when combined with another filter — Area values containing & (e.g. "Finance & Accounting") make the API's query parser return 0 results when followed by another AND keyword:... clause. Verified live that the identical clauses in the opposite order parse correctly, so Area filters are now always added last in the built query.
• The API's keyword: field isn't an exact match — it can silently mix in wrong connectors. Verified live that keyword:Vendor/OpenAI also matches azure.openai.text (real vendor: Microsoft), purely because it carries an unrelated bare keyword "Azure OpenAI". The same effect made a "Finance & Accounting" + "Microsoft" filter return 30 connectors instead of the correct 11, including several from unrelated areas (Accounts Payable, Tax, HR, etc). Added filterByExactKeywords(), a client-side post-filter that re-checks the exact Area//Vendor//Type/ tag, mirroring the existing excludeHidden/filterByRelevance post-processing pattern.
• The "HRMS" sidebar area matched zero live connectors — the actual tag Ballerina Central uses is Area/Human Resources (0 packages have Area/HRMS, 2 have Area/Human Resources). Renamed the hardcoded sidebar entry.

All of the above were verified directly against the live api.central.ballerina.io search endpoint before and after the fix, not just against mocked tests.

Test plan

• npm run typecheck passes
• npm run lint passes
• npm test — 87 tests pass, including 3 new regression tests:
  • multi-word query no longer escapes the space in the wildcard term
  • Area filter is placed after Vendor/Type filters in the built query
  • a loosely-matched false positive (azure.openai.text under a Vendor/OpenAI filter) is dropped by filterByExactKeywords
• Manually reproduced and re-verified all four issues against the live API (see commit message / PR description for the exact before/after query results)

🤖 Generated with Claude Code

[coderabbitai]

📝 Walkthrough

Summary

• Fixed multi-word wildcard search handling for the Ballerina Central API.
• Ordered area filters after other filters to prevent parser failures.
• Added exact client-side matching for Area/, Vendor/, and Type/ tags.
• Added complete-result filtering and client-side pagination for filtered searches.
• Renamed the HRMS sidebar area to Human Resources.
• Added regression tests for search, filter ordering, exact matching, and pagination.
• Type checking, linting, formatting, and all 89 tests pass.

Walkthrough

The search client preserves spaces in Solr terms and splits non-wildcard multi-word searches into separate AND clauses. It places area filters after vendor and type filters. Filtered searches fetch complete result sets before exact tag filtering, deduplication, sorting, relevance filtering, and pagination. Unfiltered searches retain server-side pagination. Regression tests cover these behaviors and the predefined HRMS area label is renamed to Human Resources.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 inconclusive)

Check name	Status	Explanation	Resolution
Description check	❓ Inconclusive	The check is inconclusive because multiple pull request templates are available, but pre-merge checks support only one description template.	Configure a single supported pull request description template, then rerun this check.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title clearly summarizes the primary search and filtering fixes in the pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/lib/rest-client/rest-client.test.ts`:
- Around line 173-184: Add a Connector entry to the types filter in the
searchPackages call, then capture the URL index for keyword:Type/Connector and
assert it appears before the existing area clause, while preserving the
vendor-before-area ordering assertions.

In `@src/lib/rest-client/rest-client.ts`:
- Around line 526-541: Apply pagination only after exact-keyword filtering:
update the single-combination path around executeSingleSearch to retrieve and
filter the complete logical result set, then calculate the exact count and slice
the requested page. Apply the same ordering to each multi-combination result
path at src/lib/rest-client/rest-client.ts lines 574-575 before merged
pagination; both sites require changes, and remove the
fixed-buffer/proportional-count approach.

🪄 Autofix

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: b5e89c4e-0465-4039-849c-9b0b49e95eeb

📥 Commits

Reviewing files that changed from the base of the PR and between 83c4c9f and 13e46ec.

📒 Files selected for processing (3)

• src/lib/rest-client/rest-client.test.ts
• src/lib/rest-client/rest-client.ts
• src/pages/HomePage.tsx

[coderabbitai]

🧹 Nitpick comments (1)

src/lib/rest-client/rest-client.test.ts (1)

285-319: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick win

Make complete-fetch regression mocks request-sensitive.

The mocks return all packages regardless of request parameters. These tests can pass if batch requests use a page-sized limit.

• src/lib/rest-client/rest-client.test.ts#L285-L319: make the mock apply the requested limit and offset, so partial fetches cannot produce the expected pages.
• src/lib/rest-client/rest-client.test.ts#L321-L340: use a count above the normal page size and assert each batch request has a limit sufficient for the complete result set.

As per path instructions, provide concise, actionable feedback focused on correctness and best practices.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/lib/rest-client/rest-client.test.ts` around lines 285 - 319, Update the
regression tests around searchPackages in
src/lib/rest-client/rest-client.test.ts:285-319 and 321-340. Make the mockFetch
response honor each request’s limit and offset so page-sized partial fetches
cannot satisfy the expected results; in the second test, use a dataset larger
than the normal page size and assert every batch request requests enough results
to cover the complete set.

Source: Path instructions

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@src/lib/rest-client/rest-client.test.ts`:
- Around line 285-319: Update the regression tests around searchPackages in
src/lib/rest-client/rest-client.test.ts:285-319 and 321-340. Make the mockFetch
response honor each request’s limit and offset so page-sized partial fetches
cannot satisfy the expected results; in the second test, use a dataset larger
than the normal page size and assert every batch request requests enough results
to cover the complete set.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: f8de1fcd-a856-4224-a1f3-f61d5f2a9888

📥 Commits

Reviewing files that changed from the base of the PR and between 13e46ec and 6735244.

📒 Files selected for processing (2)

• src/lib/rest-client/rest-client.test.ts
• src/lib/rest-client/rest-client.ts

🚧 Files skipped from review as they are similar to previous changes (1)

• src/lib/rest-client/rest-client.ts

[RDPerera]

PR body's fourth bullet says the sidebar entry was renamed HRMS → Human Resources. Seems that is missing or re-aligned for ealier state for some reason ?

[daneshk]

PR body's fourth bullet says the sidebar entry was renamed HRMS → Human Resources. Seems that is missing or re-aligned for ealier state for some reason ?

I reverted the change, as we can't change the category now