Point it at the folder
DISCOVER bronze.landing.orders
PATH 'incoming/orders';
After registration, query the folder like any other table:
SELECT *
FROM bronze.landing.orders
LIMIT 100;
Review the generated SQL first
Use PRINT to inspect the exact table definition without creating anything:
DISCOVER bronze.landing.orders
PATH 'incoming/orders'
PRINT;
Scan nested folders and retain file provenance
DISCOVER bronze.landing.events
PATH 'incoming/events'
WITH (
recursive = true,
file_metadata = true
);
What it detects
- CSV dialect, delimiter, quoting, headers, and inferred types.
- JSON, NDJSON, and XML structures with generated flattening options.
- Parquet, Avro, ORC, and Excel from their binary signatures.
- HL7, EDI, and FHIR text structures.
- Existing Delta Lake and Iceberg table layouts.
Local and cloud use the same command
The path stays relative whether the zone points to a local folder, ADLS, S3, or GCS. Storage credentials remain attached to the zone instead of being placed in the SQL statement.
Keep the folder coherent
A folder should contain one data format. DISCOVER samples files to verify agreement and unions compatible JSON, NDJSON, or XML shapes. For ambiguous data, pin the format explicitly:
DISCOVER bronze.landing.clicks
PATH 'incoming/clicks'
WITH (format = 'ndjson');
Why this matters
File exploration often starts with extensions, documentation, schema guesses, and hand-written table options. DISCOVER compresses that setup into one reviewable command, so raw landed data can become queryable in seconds.