Data viewer

Ask the warehouse a question it has never been asked. This page reads 5,884,331 observations straight out of 111 Parquet files, in your browser, over byte ranges. There is no server behind it and no database. Every chart it draws carries the query that produced it, including which files were read and which were skipped.

Loading the menu.

1. Source

Publisher and table

Choosing a source fetches its measure list and nothing else.

2. Measure

3. Places

Appearance

Colours, sizes, spacing and your own watermark. Kept on this device. The whomatters seal moves where you put it; what it says is ours.

The query behind this chart

Statistics

Every procedure states what it is standing on, and refuses when those conditions do not hold rather than returning a number that looks the same either way.

This run

How to read it

A gap in a line is a gap. Where the publisher recorded that no value is available, an amber tick appears below the axis and the count is stated in words. Nothing that was never published is drawn as a zero.

While a chart is resolving it says how many rows it has read and is marked partial, so a screenshot taken halfway through cannot pass as a finished one.

How this was measured

The viewer reads the same Parquet export offered on the download page: 5,884,331 rows across 111 files, one per place, written from canonical on 2026-08-12. It fetches only the byte ranges a query needs, and it skips whole row groups whose recorded minimum and maximum show they cannot contain the source being asked for. That skipping is checkable from the page: the button beside the query reads every group anyway and reports how many matching rows were inside the ones it would have skipped. The answer has to be zero, and it is shown rather than asserted.

216 sources are queryable here. Sources whose licence permits display and refuses redistribution are deliberately absent, because a browser query is a copy. They remain charted and cited elsewhere on this site from aggregates.