SQL workspace

You need: an analyst or admin login, and at least one project assigned to you (an admin does this on the Analysts page).

Open SQL (/sql). Your datasets appear in the panel on the side, grouped by vendor and project. Folders that have no processed file yet are hidden unless you tick the box.

Names you can query

SELECT * FROM lake.gold.chdc_health__cleaning__hospitals LIMIT 20;

• lake is the catalog, gold / silver / bronze the layer, then the dataset table vendor__project__folder.
• The default is lake.gold, so FROM chdc_health__cleaning__hospitals also works.
• Your own views are view__<name> in each layer.
• Click a dataset to get a starter query.

The editor

• Auto-complete suggests SQL keywords and functions, your dataset names, lake.<layer>. paths, and columns (with types) after table. or an alias.
• Validate checks the query without running it. Run (or Ctrl+Enter) runs it.
• Only SELECT, WITH, VALUES and TABLE statements are accepted. Nothing can change data.
• Layout (stacked or side by side), the divider position and your draft are remembered in your browser. Double-click a pane title to zoom it.

Limits that protect everyone

Limit	Default
Rows shown	1,000
Size of a shown result	10 MB
Run time	60 seconds
Query text	20,000 characters
Running queries	2 per person, 6 in total (more get a polite “try again”)

Cancel stops a running query. Leaving the page or losing the connection also stops it on the server.

Export the full result to CSV

Export CSV downloads the whole result, not just the 1,000 shown rows. It streams straight from Trino to your browser’s download manager, so nothing large is held in memory. A progress chip shows rows and megabytes so far; Cancel stops it.

Export limit	Default
Rows	1,000,000
Size	500 MB
Time	10 minutes
At the same time	1 per person, 2 in total

When a limit is hit the file ends cleanly and the chip says so. The file is UTF-8 with a byte-order mark (so Excel reads Devanagari and similar scripts), uses CRLF line ends and standard quoting.

Formula guard

A text cell that starts with =, +, - or @ gets a leading ' so Excel cannot run it as a formula. This changes such values in the file (a phone number +91 98765 43210 becomes '+91 98765 43210). Real numbers are untouched.

What you cannot see, and why errors look the same

You can query only datasets of projects assigned to you. A dataset that does not exist and a dataset you may not read give the same error text, so a message never reveals what other vendors have. Every attempt, including refused ones, is written to the audit log.

Next

Happy with a query? Save it as a view and publish it as an API.