Data grid

Only the rows near the viewport exist in the DOM; two spacer rows above and below keep the scrollbar the right length. Row heights are measured rather than assumed, so switching density in the customizer re-measures instead of leaving every row misaligned.

The dataset is generated from a seeded sequence rather than shipped as a file — zero bytes in the download, and the same rows on every load, so row 12,345 is always the same row.

The trade-offs are real.A virtualized table has no pagination, because scrolling replaces it. It has no Print action either: expanding fifty thousand rows for the printer produces a thousand pages. And Ctrl-F finds only what is rendered, which is why the search box filters the data instead of relying on the browser's.

The paginated grid at /products is the right default for ordinary datasets — it keeps pinning, inline editing, saved views, export and print. This page is for the case where there are too many rows to render at all.