SQL

Query engines

Which tool runs your query, and what that means for speed, limits and writing data.

Several tools read the lakehouse, on different engines. This page explains why a large query behaves differently in a notebook than in the SQL Editor, and which tool can change data. For the wider picture, see core concepts.

Rule of thumb: notebooks pull data to your code; Trino pushes your query to the data. Small and exploratory work fits a notebook; large or repeated SQL belongs on Trino.

Compare the tools

Catalog PreviewNotebooksSQL EditorDashboardsMap Viewer
What it isA fixed 20-row sample of a table.pyiceberg reads tables into your session; DuckDB runs SQL over what you loaded.The portal’s SQL editor.Superset charts, each backed by a query.A map of a Zarr dataset.
EngineTrinoYour notebook sessionTrinoTrinoYour browser
What movesA handful of rows.The data you read, pulled into your session.Only the result, up to 1,000 rows.Only each chart’s result.The parts of the dataset you look at.
Limits20 rows.Your session’s memory.Stops after 60 seconds.Chart queries stop after 300 seconds.Zarr generic tables only.
Can writeNoYes, with pyicebergYesNoNo
App permissionNoneUse notebooksUse SQL editorUse dashboardsUse map viewer
Reach for it whenYou want to check a table holds what you expect.Python, iterating on joins and aggregations, plots in the notebook.Ad-hoc SQL, checking results, creating or filling a table.Sharing results with people who won’t write a query.Exploring gridded data on a map.
Read moreTable detailsQuerying dataSQL EditorBuilding dashboardsMap Viewer

DuckDB in a notebook runs SQL over Arrow tables and parquet files. To query an Iceberg table with it, load the table with the lakehouse helpers first; see querying data.

Permissions are the same everywhere

Every tool on this page reads as you, with your own data grants. Switching tools never widens what you can read. See how access works.

Writing works the same way: the SQL Editor and pyiceberg in a notebook can change a table only when your grants allow it. Tables you create through the SQL Editor are the exception to “as you”; see tables you create are not yours.