The grid renders the rows in view and fetches the next window as you scroll; sorting, filtering and aggregation are pushed down to the source or the bed.
| material | description | plant | stock | value | lead time | rule |
|---|---|---|---|---|---|---|
| M-1183920 | resistor 470 Ω 1% 0603 | 1210 | 128,400 | €2,337 | 21 d | pass |
| M-2200417 | housing, cast, 220 mm | 1210 | 3,180 | €412,300 | 42 d | lead time drift |
| M-0091114 | bearing 6204-2RS | 2000 | 18,900 | €27,405 | 14 d | pass |
| M-3391005 | controller board rev C | 1210 | 412 | €184,210 | 70 d | over 12 weeks safety |
Columns come from the golden view and from the transactions in the same bed: description and lead time are master data, stock and value are live facts. The rule column shows the latest rule results per row, and a click opens the rule's evidence.
The grid holds only a window of the table. Every sort, filter, group and aggregate is a query the engine pushes to the source or runs on the bed, and only the rows in view travel. Keys stay stable across windows, so an edit on row 3 million lands on the right row after a re-sort.
The same component renders a steward's cluster of fourteen records, a drill-down from a pivot cell into 214,000 order lines, and a governed extract of the whole table; each is one query with the same security.
Hub interfaces page results and cap exports; inline and bulk editing are common; spreadsheet-style editing appears in some products. Virtual scrolling with pushdown is rarer, and where it exists it covers the hub's own tables, because the transactions sit elsewhere.
Here the grid scrolls the federated result set, so a transaction table of tens of millions of rows and a master table of thousands look and behave the same, and an export has no fixed row cap.
Where the domain allows it and the owning system accepts a write-back: a planner edits a forecast row; an order line stays read-only unless the process app owns the change.
The window request carries a time budget; a source over budget answers from its cached or replicated contract, and the column says so.
The workbench exposes the same grid through SQL to spreadsheets; a native add-in is on the roadmap (see the review notes).
Bring two systems that hold the same customers, suppliers or materials. In one working day we connect them, run the match on live rows, review the queue with your steward and publish the first golden view with its API.