- Extract the complete import logic (label-scan reading, page
find-or-create, slot filling, content-idempotent image attach) from
ImportSubmarketWorkbooksCommand into the shared service
SubmarketWorkbookImporter; the CLI command and the new controller are
thin shells around importFile(), so both entry points cannot drift.
CLI regression verified: identical dry-run report for the reference
workbook (12 elements, page 443)
- New "Sub-Markets" tab in the VITEC Import module
(SubmarketImportController + Import/Submarkets.html): upload multiple
filled workbooks into the SAME inbox the CLI uses
(migrations/submarkets_xlsx_in/), per-file checkboxes, dry-run and
apply buttons, result cards with the familiar report and a direct
link into the page module; backend dry-run verified against the CLI
- Submitted filenames are whitelisted against the actual directory
listing (no path traversal); uploads restricted to sanitized .xlsx
basenames
- Routes submarkets/submarket_upload/submarket_process in Modules.php,
Sub-Markets tab link in the sibling module views, ok/danger badge
colors in backend-import.css
- Docs section 5y (incl. operations note: apply in small batches in the
module, mass runs via CLI - image downloads run inside the web
request)
per sub-market (25, grouped by the 9 main markets): VITEC-styled Read me,
a Page-layout sheet embedding the Figma design with section markers,
DB-live dropdowns, soft character limits, a hidden Meta sheet and named
input cells. Names must NOT start with a cell-reference token - Excel
silently strips s1_-style names as repairs.
vitec:import-submarket-workbooks (dry-run default, idempotent) finds or
creates the page below its main market (hidden, layout 21), fills the six
colPos slots updating in place and never deleting, links detail_page on
the market record, downloads asset URLs into fileadmin/submarket_import/
for editorial review and writes the flexforms as DataHandler arrays.
Reading is three-layered after a bruising day of real-world Excel: named
ranges first, then a label scan over the locked column A with section
tracking - a customer save stripped ALL defined names and shifted rows,
so fixed addresses are not a contract.
Learned and fixed along the way: DuplicationBehavior moved to
Resource\Enum in v14; sys_file_reference rows via DataHandler need
pid/tablenames/uid_foreign or are dropped silently; handing DataHandler a
fresh NEW reference per run STACKS live references on re-import (caught
by a double-run test), so attachments are now content-idempotent - the
existing live reference is reused or retargeted and surplus duplicates
are deleted via cmdmap; RequestFactory replaces GeneralUtility::getUrl so
download failures carry their reason - which surfaced that the dev basic
auth blocked the server itself (server IP + loopback now allowlisted,
which a future SSR needs anyway) and that the core resource-consistency
check rightly rejects HTML masquerading as PNG.
Verified end to end on Broadcast & Media from a clean slate, twice in a
row: 12 elements in colPos 110-160, layout-21 payload, both images
attached exactly once with publicUrl in the JSON; second run reports
nothing but unchanged.
orasche