[FLINK-12711] Separate function implementation and definition

• Type: Sub-task
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.9.0
• Component/s: Table SQL / API
• Labels: pull-request-available
• Assignee: Timo Walther
• Reporter: Timo Walther
• Created: 03/Jun/19 13:24
• Updated: 18/Jun/19 14:11
• Resolved: 18/Jun/19 14:10

Description

This issue continues the work that was started in FLINK-11449. It distinguishes between function with implementation (UDFs) and function with no implementation.

The rough design looks as follows:

1. `interface FunctionDefinition`
   --> general interface for describing a function
   --> goal: separation of runtime and planning/optimization
   --> long-term methods: `getKind()` (aggregate, scalar, table), `getTypeInference()`
2. `interface UserDefinedFunctionDefinition extends FunctionDefinition`
   --> interface for describing a function with implementation
   --> methods: `createImplementation(): UserDefinedFunction`
   --> default: getTypeInference() = Util.DEFAULT_INFERENCE // future work
3. `class BuiltInFunctionDefinition implements FunctionDefinition`
   --> class for describing a function where the planner provides an implementation
   --> methods: `getName(): String`
4. Add `getKind` to `AggregateFunction`, `ScalarFunction`, `TableFunction`

Issue Links

• links to: GitHub Pull Request #8661