SUMMARIZE

SUMMARIZE profiles a table or a query. It returns one row for each column. Run it first on a new dataset.

SUMMARIZE obs;

column_name	column_type	min	max	distinct	avg	std	count	null_percentage
temperature	Float64	10.0	30.0	3	20.0	10.0	3	25.0
depth	Int64	0	100	3	62.5	47.9	4	0.0
platform	Utf8	A	B	2			4	0.0

Beacon returns the columns in the order of the source.

Forms

SUMMARIZE obs;                                   -- a table
SUMMARIZE public.obs;                            -- schema-qualified
SUMMARIZE SELECT depth FROM obs WHERE depth > 0; -- any query
SUMMARIZE (SELECT * FROM read_parquet('obs/*.parquet'));

::: warning A table function needs a query
`SUMMARIZE` takes a **name** or a **query**. A bare
[table function](/docs/2.0.0/sql/table-functions) is neither, so this does not parse:

```sql
SUMMARIZE read_netcdf('argo/*.nc');   -- error

Wrap it in a query instead:

SUMMARIZE (SELECT * FROM read_netcdf('argo/*.nc'));

:::

## What each column means

- **min / max**: for columns with an order, such as numbers, strings and timestamps. Beacon returns them as text.
- **distinct**: the exact number of distinct non-null values.
- **avg / std**: for numeric columns only. Other columns get `NULL`.
- **count**: the number of non-null values.
- **null_percentage**: the share of `NULL` values, from `0` to `100`.

## Notes

`SUMMARIZE` becomes an ordinary aggregate query with one pass. It therefore works on a
read-only connection. It needs no special
privileges. It scans the source once.