fix(postman): fall back to English (id=1) when _getdefault returns invalid id (#35780)

[dsolistorres]

Summary

Follow-up to PR #35795. The prerequest script we added there accepts any non-falsy value from GET /api/v2/languages/_getdefault and stores it as defaultLanguageId. In some test environments the endpoint returns the LANG__404 sentinel with id=-1, which our guard treated as valid. Test bodies then sent \"languageId\": -1 and the server FK-failed on contentlet insert.

Evidence from run 26312707306 (PR #35771)

GET http://localhost:8080/api/v2/languages/_getdefault [200 OK, 925B, 28ms]
'defaultLanguageId = -1'
...
ERROR:  insert or update on table "contentlet" violates foreign key constraint "fk_contentlet_lang"
Detail: Key (language_id)=(-1) is not present in table "language".

The failing test (Page API / Test DotAsset as titleImage / Create DotAsset image) had defaultLanguageId = -1 substituted into its body. Server's _getdefault returned the sentinel, our script blindly trusted it.

Fix

Tighten the existing guard and add a fallback to English (id=1) — always present in dotCMS starter data, matches the same approach as #35813:

-const id = body && body.entity && body.entity.id;
-if (!id) {
-    throw new Error('Could not resolve default language: entity.id missing in response');
+let id = body && body.entity && body.entity.id;
+if (!id || id <= 0) {
+    // _getdefault can return the LANG__404 sentinel (id=-1) in fresh
+    // test environments. Fall back to English (id=1), which is always
+    // present in dotCMS starter data. See issue #35780.
+    console.log('Default language returned invalid id (' + id + '), falling back to English (id=1)');
+    id = 1;
 }

Network errors and non-200 responses still throw (no silent swallowing) — only invalid-id payloads fall back.

Scope

• Workflow_Resource_Tests.json + GraphQLTests.json (both currently on main).
• AI uses the same script via still-open PR fix(postman): pass explicit language in AI Search Related by identifier (#35780) #35816 — I'll push the same fallback to that branch in a follow-up so both PRs are consistent before merge.

Test plan

• JSON validity verified for both files.
• Diff is surgical: 14 inserts / 6 deletes total.
• CI verifies — PR Test / Postman Tests - GraphQL and Workflow should pass even when _getdefault returns the sentinel.

🤖 Generated with Claude Code

[claude]

Claude finished @dsolistorres's task in 1m 18s —— View job

---

Review

• Read changed files
• Verify diff matches the description
• Check for edge cases / risks
• Post review

Verdict: looks correct and minimal. The guard change (!id || id <= 0) catches the -1 LANG__404 sentinel, the fallback to id=1 is justified by the inline comment, and the original throws for network/HTTP errors are preserved so non-payload failures still surface. JS coercion makes id <= 0 work whether the server returns -1 as a number or "-1" as a string. Both files were kept in sync.

Worth flagging (non-blocking)

1. Script duplication is growing. This prerequest script now lives in GraphQLTests.json:14860 and Workflow_Resource_Tests.json:19833, and the PR mentions a third copy is coming via fix(postman): pass explicit language in AI Search Related by identifier (#35780) #35816. Three identical 30-line embedded JS blocks in Postman JSON is a future drift hazard — the next bug fix has to touch all three, and there is no compiler to catch a missed file. Not something to fix in this PR, but worth a tracking issue: either move this to a Postman collection-level pre-request (parent folder), or generate the JSON from a shared snippet at build time.

2. id = 1 is an environment assumption, not a fact. The comment says English is "always present in dotCMS starter data," which is true for fresh installs, but a test environment that has deleted or reseeded languages without id=1 would still FK-fail with exactly the same symptom — just less obviously, because the fallback masks the misconfig. Acceptable trade-off for tests, but if anyone hits this again the script's console.log is the only breadcrumb. Consider also asserting via pm.test so the run fails loudly rather than silently picking 1, e.g.:

   console.warn('Default language returned invalid id (' + id + '), falling back to English (id=1) — server may be misconfigured');

   console.warn over console.log would make this stand out in Newman output.

3. Type of id is not normalized. entity.id comes back as a JSON number, but if it ever returned a numeric string the env var stored would be a string. That has not been a problem historically, just noting that id = Number(id) (or parseInt(id, 10)) before storing would make the contract explicit. Low priority.

4. No coverage for the sentinel path itself. There's no test that asserts the fallback is exercised when _getdefault returns id=-1 — only the in-the-wild CI failure validates it. If the sentinel value ever changes (e.g. becomes 0 or null), the guard might miss it again. A Postman test in LanguageResourceTests.json that calls _getdefault on a fresh env and asserts the script's behavior would be defensive. Probably overkill for now.

Nits

• The block is now 14/6 instead of 7/3 because the comment is included in both copies. Fine.
• let id is needed (reassigned) — correct change from const.
  · branch issue-35780-postman-language-fallback