Unified search: shared render-type resolution rule

[habdelra]

Phase 0 of Unify prerendered-card-search and card-search — the single, pure render-type resolution rule (Decision #1) that the server and host both call so they pick the same type. Additive; no behavior change on existing paths.

Stacked on #5150 (CS-11429) — it uses that PR's SearchRenderType. Review the top commit here; the base will be retargeted to main once #5150 merges.

The rule — resolveRenderType({ renderType, filterOn, types }) (runtime-common search-utils)

• an explicit renderType CodeRef → use it;
• the "native" escape valve → the result's own most-derived type (types[0]);
• omitted → the query's filter.on (the common ancestor searched on); when the query has no filter.on, fall back to the most-derived type.

It returns the CodeRef used both to select the HTML column and echo meta.renderType (server, CS-11433) and to drive the live CardRenderer @codeRef (host, CS-11434). The default is filter.on — an intended change from today's silent types[0] that only takes effect on the unified path; the legacy compat shim (CS-11433) preserves today's types[0]-unless-renderType behavior until callers migrate. The rule governs only the prefer-HTML path; dataOnly results are the actual types, so the request carries no render spec to resolve.

Tests

Shared contract tests run via the realm-server runner (5 cases): explicit CodeRef wins; "native" → types[0]; omitted → filter.on; the missing-filter.on fallback; and that a dataOnly request carries no render (so the rule is inapplicable).

🤖 Generated with Claude Code

[habdelra]

If you are only testing on the realm server, then don't put teh test in common. that pattern is only used for tests that need to execute in both host and realm -server

[Copilot]

Pull request overview

This PR introduces a shared, pure “render-type resolution” rule in @cardstack/runtime-common so both the realm-server and host can resolve renderType consistently for unified search (prefer-HTML) results, and adds shared contract tests to lock the behavior down.

Changes:

• Added resolveRenderType({ renderType, filterOn, types }) to packages/runtime-common/search-utils.ts.
• Added runtime-common shared contract tests covering the resolution matrix and the dataOnly “no render spec” invariant.
• Wired the new shared tests into the realm-server test runner.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File	Description
packages/runtime-common/tests/render-type-resolution-test.ts	Adds shared contract tests for resolveRenderType and verifies dataOnly requests omit render.
packages/runtime-common/search-utils.ts	Adds the shared resolveRenderType rule used to pick a rendering CodeRef consistently across server/host.
packages/realm-server/tests/render-type-resolution-test.ts	Executes the shared contract tests via runSharedTest in realm-server.
packages/realm-server/tests/index.ts	Registers the new realm-server test module so it runs in CI.

---

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

[habdelra]

[Claude Code 🤖] Done — both suites moved out of runtime-common/tests into self-contained realm-server qunit modules (importing from @cardstack/runtime-common), shared files + consumers removed: render-type resolution here (2f27551) and the unified-search contracts on #5150 (b5f1def).

[github-actions]

Host Test Results

    1 files  ±0      1 suites  ±0   1h 51m 23s ⏱️ -53s
2 947 tests ±0  2 932 ✅ ±0  15 💤 ±0  0 ❌ ±0 
2 966 runs  ±0  2 951 ✅ ±0  15 💤 ±0  0 ❌ ±0 

Results for commit 1528dda. ± Comparison against earlier commit 2f27551.

Realm Server Test Results

    1 files  ±0      1 suites  ±0   10m 28s ⏱️ -16s
1 566 tests ±0  1 566 ✅ ±0  0 💤 ±0  0 ❌ ±0 
1 657 runs  ±0  1 657 ✅ ±0  0 💤 ±0  0 ❌ ±0 

Results for commit 1528dda. ± Comparison against earlier commit 2f27551.

[backspace]

Is this “Decision #1” reference something we want in the comments forever?

[habdelra]

[Claude Code 🤖] Agreed — no, it should not live in the comment forever. Removed in 1528dda; the rule itself stays documented, just without the decision-number reference (comments describe the enduring contract, not the rollout). Swept the same kind of reference (a CS-11433 mention) out of the stacked #5157 too.