Raw tables

Raw tables accept JSON rows before you define a schema. RawTree stores each row's undeclared fields in the internal __raw_data JSON column.

Physical columns and JSON paths

Declared physical columns can exist beside paths in __raw_data. A physical column takes precedence when its name matches a JSON path.

RawTree exposes known JSON leaf paths as logical SQL columns. Query those paths by name, such as user.id, without reading the parent JSON value.

Use __raw_data.path when you need the JSON value at a path. Use __raw_data only when you need the complete parsed JSON object.

SELECT * expansion

SELECT * returns declared columns and known JSON leaf paths. It does not return the parent __raw_data column.

The set of known paths can change as rows arrive. Select named paths in reusable queries to keep the result shape stable.

SELECT event, user.id FROM events LIMIT 10

Missing and unknown paths

A known path that is absent from one row returns NULL. A path that does not exist in the table schema returns an UNKNOWN_IDENTIFIER exception.

Inspect the table or query known fields before you use a new path. The is_raw setting marks a raw table. The system.tables.is_raw value identifies a raw table.

Related

• Dynamic fields
• Query data
• Optimize queries
• Ingestion model
• Storage and compute