Lattice Grid 1.78: filter menus and pivots run in the engine
A grid over a hundred-million-row table should feel like a grid over a hundred rows. This release closes three places where a pushdown source used to show its seams: the filter menu, the pivot, and select-all.
A filter menu that opens at once
Click a column’s filter button on a pushdown source and the value list now comes straight from the engine: the top 200 values by count, most frequent first, with a search box that asks again for anything further down the list. The same route serves the tool panel’s filter section too. Nothing in the browser scans a row to build it, so the popup that used to freeze the tab on a large table now opens as fast as it does on a hundred rows.
Pivots that run where the data lives
A pivot on DuckDB, ClickHouse, Elasticsearch or Splunk now answers from grouped aggregates in the engine itself. The pivot columns come from the engine’s own reply, a group opens its children on demand, and no leaf row crosses the wire to produce the totals. Pivoting a ten-million-row table by country and category moves the handful of combinations that actually occur, not the ten million rows behind them. A source that genuinely cannot pivot, a plain REST endpoint without a query language of its own, says so plainly instead of drawing grouped rows and calling them a pivot.
Select-all means every matching row
Tick the header checkbox on a filtered pushdown grid and the selection now covers every row the filter matches, counted by the engine, not only the rows already loaded. Untick a few and the header shows the difference. Export or delete that selection and the grid streams the whole thing and hands back a Promise, so “select all, export” does exactly what it sounds like on a table far larger than the page.
Splunk exports with a ceiling
A Splunk export in streaming mode now stops at the row cap you set instead of reading an unbounded search to its end, and says plainly when it has been capped. A finished search job is released the moment its last page is read, so a read-only role’s search-artifact quota lasts through a whole session’s questions rather than one.
ClickHouse answers more of the query
A filter or a second statistic over an already-aggregated column, price summed and price filtered in the same request, now answers from ClickHouse instead of being refused. See it running on ClickHouse, live on 28 million rows: open a filter menu and watch the value list arrive from the engine.
See it running
Reading ClickHouse live shows the filter menu, the pivot and the generated SQL together. The pushdown guide covers pivoting, filter values and select-all on a pushdown source in full, and the Splunk guide has the export cap. The full list of what changed is in the changelog.