Table-Valued Functions¶

Table-Valued Functions (TFVs) are functions that return a table (as a LogicalPlan) that can be used anywhere that a regular (scalar) table is allowed.

Table functions behave similarly to views, but, as functions in general, table functions accept parameters.

Table-Valued Functions are represented as UnresolvedTableValuedFunction logical operators in logical plans (after AstBuilder is requested to parse a table-valued function in a SQL statement).

Google BigQuery Documentation

Read up on table-valued functions in the official documentation of Google BigQuery (for a lack of a better documentation).

Spark SQL supports Logical Operators and Generator expressions as table-valued functions.

Name	Logical Operator	Generator Expression
`range`	`Range`
`explode`		`Explode`
`explode_outer`		`Explode`
`inline`		Inline
`inline_outer`		Inline
`json_tuple`		`JsonTuple`
`posexplode`		`PosExplode`
`posexplode_outer`		`PosExplode`
`stack`		`Stack`

Custom table-valued functions can be registered using SparkSessionExtensions.

Spark SQL uses the SimpleTableFunctionRegistry to manage the built-in table-valued functions.