[FLINK-16555] Preflight check for known unstable hashCodes.

• Type: Improvement
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: 1.11.0
• Component/s: API / DataStream, API / Type Serialization System
• Labels: pull-request-available usability
• Assignee: Lu Niu
• Reporter: Stephan Ewen
• Created: 11/Mar/20 20:57
• Updated: 08/Apr/20 11:42
• Resolved: 08/Apr/20 11:42

Description

Data types can only be used as keys, if they have a stable hash code implementation that is deterministic across JVMs. Otherwise, the keyBy() operations will result in incorrect data routing.

We should eagerly check the key type information for known cases types with unstable hash code, such as

• arrays
• enums
• anything that does not override Object.hashCode

Issue Links

• relates to: FLINK-5874 Reject arrays as keys in DataStream API to avoid inconsistent hashing
• relates to: FLINK-16596 Support Enum-Values as part of a Key
• links to: GitHub Pull Request #11518