Table API & SQL

Apache Flink features two relational APIs - the Table API and SQL - for unified stream and batch processing. The Table API is a language-integrated query API for Java, Scala, and Python that allows the composition of queries from relational operators such as selection, filter, and join in a very intuitive way.

This documentation is intended for contributors of the table modules, and not for users. If you want to use Table API & SQL, check out the documentation.

Modules

Common

• flink-table-common:
  • Type system definition
  • UDF stack and built-in function definitions
  • Internal data definitions
  • Extension points for catalogs, formats, connectors
  • Core APIs for extension points such as Schema
  • Utilities to deal with type system, internal data types and printing

API

• flink-table-api-java:
  • Java APIs for Table API and SQL
  • Package org.apache.flink.table.delegation, which serves as entrypoint for all planner capabilities
• flink-table-api-scala: Scala APIs for Table API and SQL
• flink-table-api-bridge-base: Base classes for APIs to bridge between Table API and DataStream API
• flink-table-api-java-bridge:
  • Java APIs to bridge between Table API and DataStream API
  • Connectors that are developed using DataStream API, usually need to depend only on this module.
• flink-table-api-scala-bridge: Scala APIs to bridge between Table API and DataStream API
• flink-table-api-java-uber:
  • Uber JAR bundling flink-table-common and all the Java API modules, including the bridging to DataStream API and 3rd party dependencies.
  • This module is intended to be used by the flink-dist, rather than from the users directly.
• flink-table-calcite-brdge: Calcite dependencies for writing planner plugins (e.g. SQL dialects) that interact with Calcite API

Runtime

• flink-table-code-splitter: Tool to split generated Java code so that each method does not exceed the limit of 64KB.
• flink-table-runtime:
  • Operator implementations
  • Built-in functions implementations
  • Type system implementation, including readers/writers, converters and utilities
  • Raw format
  • The produced jar includes all the classes from this module and flink-table-code-splitter, including 3rd party dependencies

Parser and planner

• flink-sql-parser: Default ANSI SQL parser implementation
• flink-table-planner:
  • AST and Semantic tree
  • SQL validator
  • Query planner, optimizer and rules implementation
  • Code generator
  • The produced jar includes all the classes from this module together with the two parsers, including 3rd party dependencies (excluding Scala dependencies).
• flink-table-planner-loader-bundle Bundles flink-table-planner, including Scala dependencies.
• flink-table-planner-loader: Loader for flink-table-planner that loads the planner and it's Scala dependencies in a separate classpath using flink-table-planner-loader-bundle, isolating the Scala version used to compile the planner.

SQL client

• flink-sql-client: CLI tool to submit queries to a Flink cluster

Testing

• flink-table-test-utils: Brings in transitively all the dependencies you need to execute Table pipelines and provides some test utilities such as assertions, mocks and test harnesses.

Notes

No module except flink-table-planner should depend on flink-table-runtime in production classpath, no module except flink-table-planner-loader should depend on flink-table-planner-loader-bundle in production classpath, and similarly no module should depend on flink-table-planner or flink-table-planner-loader in production classpath. For testing, you should depend on flink-table-planner-loader and flink-table-runtime. These are already shipped by the Flink distribution.