feat(cmdb-import): show where each import pack comes from and link to its editor in OpenRegister

[WilcoLouwerse]

What

The "Mapping" block of the CMDB import section shows, per pack, where the mapping the next import runs comes from, and links to its editor in OpenRegister (ConductionNL/openregister#4602). The per-row mapping tables of this PR's first version are gone: viewing and editing the packs happens in OpenRegister. Stacked on #1288.

How

• CmdbImportMapping.vue: two tables, the export's sheets and columns, and the sheet packs with target, a source badge (OpenRegister / shipped file), version (stored and shipped when they differ), an "edited" marker, a "newer version shipped" hint, and Edit in OpenRegister (editUrl) for an admin. The municipality pack is not listed: it maps the municipality chosen in the import options, not a column. Without OpenRegister's pack store the block says the shipped mapping is used. A broken stored pack names the pack and links to its editor.
• CmdbImport.vue: the help under each import option is one short sentence; the full explanation of which fields a re-import overwrites and what publishing means is in the section's information panel.
• Opening the page writes nothing: the endpoint only reads since feat(cmdb-import): run the mapping packs stored in OpenRegister, seeded from the shipped files #1288's review fixes. The loading line is a status region.
• The sheet-name help and the section's loaded event stay (the help texts and hints name the profile's sheets).
• Unused helpers and strings removed; en + nl for the new strings.

Checks

At 3ddbffd: npm test (jest) 190 of 190, vitest cmdbImportMapping.spec.js 19 of 19 (the full vitest run has 6 failures that are also on #1288's head without this PR's UI: four manifest-page specs and two in settingsInfoPanels.spec.js), test:l10n and check:l10n-js OK, eslint and prettier clean, openspec validate --strict valid, npm run build OK. e2e on the rig (OpenRegister with #4602): cmdb-import-mapping.spec.ts and cmdb-import-packs.spec.ts 9 of 9. PHP checks: as #1288 (same lib/).

Merge order

After #1288, and after ConductionNL/openregister#4602. The link is conditional: editUrl is only answered for a pack the store holds, and the store is only used on an OpenRegister with #4602 (see #1288's probe). On an older OpenRegister the block shows "Shipped file" and no link, so there is no link into a missing editor. The docs name the requirement.

Jira: WOO-588 — plan: https://github.com/ConductionNL/readonly-mirror-wilco-claude-plans/tree/dev-mirror/issues/jira/WOO-586/sub-tasks/WOO-588/

🤖 Generated with Claude Code

[WilcoLouwerse]

Verdict: APPROVE (Thorough) — self-review posted as COMMENT (GitHub blocks self-APPROVE)

No blockers: the block renders everything as text, the collapsible has aria-expanded/aria-controls, every new string is in all four l10n files, and the 503 path shows the reason with the shipped sheet names as fallback. The concerns are the help-text fallback for profiles with other than two sheets and the overclaim in the docs.

Update 2026-10-12: All three concerns were fixed on the branch after posting; f4 (error hints) was fixed too.

🟡 Concerns (3)

• Help text and mapping block disagree for profiles with other than two sheets — src/utils/cmdbImport.js:1027. mappingSheetNames() returns the first two sheet names when the profile has two or more, and the shipped names ('Onbeh Applicaties CMDB', 'Beheerde Applicaties CMDB') when it has fewer than two. ✅ Addressed in 2e381f7b.
• Docs claim the block shows what the import does; several behaviours are not shown — docs/features/cmdb-import.md:352. 'The block reads the files through the same loader the import uses, so what it shows is what the import does' promises totality. ✅ Addressed in 7db0b2a0.
• Help-text wiring has no executed test — src/views/settings/sections/CmdbImport.vue:800. The headline change, sheet names in the help text coming from the endpoint, rests on onMappingLoaded -> mapping -> sheetNames -> two t() calls. ✅ Addressed in ae929478.

🟢 Minor (4)

• E2E fallback assertion passes before the 503 arrives — tests/e2e/spec-coverage/cmdb-import-mapping.spec.ts:182. In the 'endpoint fails' half, the loop asserting the help contains the shipped sheets runs before the error card is confirmed.
• Heading level, loading announcement and duplicated helpers — src/views/settings/sections/CmdbImportMapping.vue:80. (a) Pack headings are h4 but the block has no h3 of its own (its title is a button), so in the outline they nest under whichever h3 the report left before it (CmdbImport.vue uses h3 at 246 and 273), or skip a level when no report is shown.
• Error hint still names the shipped sheets — src/utils/cmdbImport.js:535 (outside the diff, no thread). The WORKBOOK_TOO_LARGE hint ('it reads only "{first}" and "{second}"') still takes PROFILE_DEFAULTS.sheets (lines 537-538), so after this slice one page can name two different sets of sheets. ✅ Addressed in f04bf823.
• Feature overview does not list the new endpoint — docs/features/README.md:167 (outside the diff, no thread). The CMDB import card (not in this diff) lists only 'Endpoint: POST /apps/stackiq/api/cmdb-import' and the CmdbImportController.

✅ Verified clean

• No v-html; interpolated t() calls use AS_TEXT.
• All new keys are in en.js, en.json, nl.js and nl.json with identical key sets.
• The UI reads only fields mappingOverview() returns; one call, matching routes.php.

Scope decision: slice 2 of 2 of the closed #1280, stacked on #1288; reviewed against its own base only.
Gates: this stacked PR gets no quality workflow until it targets development; hydra-gates ran locally scoped to its diff: 45 of 45 applicable gates green. Conflict markers and PHP syntax green on a3f4da3.
Jira context: WOO-588 — CMDB-import: mapping inzichtelijk en beheerbaar voor de beheerder (status: Refinement)

[WilcoLouwerse]

Verdict: APPROVE (Thorough) — self-review posted as COMMENT (GitHub blocks self-APPROVE)

No blockers at a1f91b0: 1 new concern (the editor link needs openregister#4602), 2 new minors, 2 earlier minors still open. 5 threads are unresolved and development requires every thread resolved; ce55ebc (the block rewritten as tables, pushed during this review) is not covered by this approval and needs its own round.

Round	Found	Status
1 (a3f4da3)	🟡 help text vs mapping block for other sheet counts · 🟡 docs overclaimed what the block shows · 🟡 help-text wiring untested · 🟢 e2e fallback passes before the 503 · 🟢 heading, loading announcement, duplicated helpers	3 resolved (f04bf82, 7db0b2a, ae92947), 1 partly, 1 open
2 (a1f91b0) — this review	1 🟡, 2 🟢	open

Finding	Was	Now	Where
🟡 Help text and mapping block disagree for profiles with other than two sheets	🟡 r1	resolved in f04bf82 (mappingSheetNames, fileHelpText, sheetsReadText handle 1 to N sheets)	thread
🟡 Docs claim the block shows what the import does	🟡 r1	resolved in 7db0b2a at a1f91b0; ce55ebc brings the tables back, so the claim needs a fresh look next round	thread
🟡 Help-text wiring has no executed test	🟡 r1	resolved in ae92947	thread
🟢 Heading level, loading announcement and duplicated helpers	🟢 r1	partly: the heading point is moot; the loading state still has no role="status"/aria-live, and the .cmdb-import__help/__code CSS is still duplicated from CmdbImport.vue	thread
🟢 E2E fallback assertion passes before the 503 arrives	🟢 r1	still open: the help assertion still runs before the error card is confirmed, so the loading state satisfies it	thread
🟡 The "Edit in OpenRegister" link and the help text depend on an editor that is not merged	new	new — 🟡	comment
🟢 Orphaned translation key	new	new — 🟢	comment
🟢 "Nothing is written" is not guaranteed	new	new — 🟢	comment

Details of this round

Scope decision: delta a3f4da3…a1f91b0, 8 non-merge commits, 16 files (frontend, l10n, docs, openspec, vitest, e2e). The base merges were read with git show --cc; their resolved conflicts are spec and docs text only. Not reviewed: 836107c and ce55ebc (the block shown as tables, help shortened), pushed during this review. The inline comments sit on ce55ebc, on lines that still carry the finding. A finding about the old "not editable" hint was dropped, because ce55ebc filters the municipality pack out instead.

✅ Verified clean (at a1f91b0)

• Every field the UI reads (packId, storedVersion, shippedVersion, edited, newerShipped, editable, editUrl, openRegisterPacks) is in #1288's mapping response, and both source values render.
• The 503 and 500 envelopes keep details.reason, details.pack and details.editUrl through normaliseError. The 403, network and no-reason cases have tests.
• editUrl is built server-side. The frontend accepts only http(s) or server-relative URLs, has no v-html, and renders every pack-supplied value as text.
• The four l10n files are in parity.

Gates: local run (stacked PR, no CI), 88 of 89 applicable gates ran. Not run: icon-vocabulary (vue-material-design-icons not installed). Gates 22, 53 and 116 failed for wiring only (ajv not resolvable in the worktree), so they count as not run, not as a pass. The gate-16 hit is on lib/Command/ReferencesAuditCommand.php, which is outside this PR's diff.
Jira context: WOO-588, CMDB-import mapping visible and manageable for the administrator.

Before the next round:

git grep -nF 'OpenRegister is missing or a mapping file is invalid.' -- l10n   # empty
gh pr view 4602 --repo ConductionNL/openregister --json state                   # MERGED, or the order in the PR body

[WilcoLouwerse]

Verdict: APPROVE (Thorough) — self-review posted as COMMENT (GitHub blocks self-APPROVE)

Nothing left to address: this round's 2 concerns and 2 minors, plus one concern from the pre-push check, were fixed on the branch before posting, and all open threads from rounds 1-2 are resolved. What still stands between this PR and a merge: 1 approval, its base #1288, and openregister#4602 for the editor link.

Round	Found	Status
1 (a3f4da3)	3 🟡 · 2 🟢	resolved in f04bf82, 7db0b2a, ae92947, 49bc84b, 9fd025a
2 (a1f91b0)	🟡 editor link needs openregister#4602 · 2 🟢	resolved via #1288's 3d8b789 and ce55ebc
3 (9fd025a) — this review	2 🟡 (caption promised a column→field mapping the tables lacked; docs overclaimed) · 2 🟢	fixed before posting in 7fc5172, 9fd025a

Finding	Was	Now	Where
🟡 Editor link depends on an editor that is not merged	🟡 r2	resolved: editUrl only for a pack in a store with the hardened editor (#1288 3d8b789)	thread
🟢 Orphaned translation key	🟢 r2	resolved	thread
🟢 "Nothing is written" not guaranteed	🟢 r2	resolved by #1288's read-only GET	thread
🟢 E2E fallback assertion before the 503	🟢 r1	resolved in 49bc84b	thread
🟢 Heading, loading announcement, duplicated helpers	🟢 r1	resolved in 49bc84b + 9fd025a	thread
🟡 Caption promises a column→field mapping the tables do not show	new	fixed before posting in 7fc5172	src/views/settings/sections/CmdbImportMapping.vue
🟡 Docs say the block shows what it does not	new	fixed before posting in 7fc5172	docs/features/cmdb-import.md

Details of this round

Scope decision: this PR's own commits since a1f91b0 (836107c, ce55ebc, 49bc84b: 13 files, 658 changed lines); base merges read with git show --cc. Also this run's merge 50cde19 (conflicts in the change's spec and tasks resolved), the fix 7fc5172, the merge d5a7795 and the fix 9fd025a.

✅ Fixed before posting (5)

• Caption promised a mapping the tables lacked (7fc5172): every sheet pack has a collapsible "Columns and fields" with a field table and a rules table. The field table shows column, field, required and transform type (a lookup with its size). The rules table shows defaults, create-only defaults, create-only fields and never-written fields. Everything is plain text, with captions and th scope.
• Docs overclaimed (7fc5172): docs/features/cmdb-import.md names what the block shows and what it does not (empty values, sheet constants, limits, the municipality pack).
• updatePending from #1288 shown (7fc5172): "Update pending: shipped {version} replaces stored {storedVersion} at the next import."
• The loading announcement reached no one (7fc5172, 9fd025a): one visually hidden role="status" region sits outside the collapsed block. It renders empty and changes after mount, and it also announces a failed load while the block is collapsed. Both tests go red without the fix.
• Unreachable municipality label (7fc5172): the branch and its l10n key are removed.

✅ Verified clean

• Every field the UI reads is in #1288's mappingOverview() at 0a349a8. A transform object never prints [object Object], and there is no v-html.

Gates: local run (stacked PR, no CI) on 3ddbffd, 88 of 89 applicable gates ran. Not run: icon-vocabulary. Gates 22, 53 and 116 failed for wiring only (ajv not resolvable), so they count as not run. The gate-16 hit is outside this diff.
Checks after the last edit: vitest cmdbImportMapping.spec.js 24/24; jest CmdbImport.spec.js + cmdbImport.spec.js 86/86; eslint, prettier and stylelint clean; test:l10n and check:l10n-js OK. Playwright was not run.
Jira context: WOO-588, CMDB-import mapping visible and manageable for the administrator.

Before the next round: this round's findings came mostly from the fixing. The next round's fixes go through the class sweep and the pre-push check first.

npx vitest run tests/vitest/cmdbImportMapping.spec.js   # 24 green