Status

Current state: "Under Discussion"

Discussion thread: TBD

JIRA: TBD

Released: TBD

Motivation

Apache Thrift (along with protocol-buffers )is widely adopted as a high throughput network traffic protocol. Historically, Pinterest has been utilizing thrift to encode strongly typed Kafka messages as well as HDFS sequence files in the data warehouse. On one hand, versioned thrift schema files repo served as schema registry where producers and consumers across languages could encode/decode with the latest schema. It also poses a challenge of maintenance and upgrades have given the lack of backward-compatible thrift encoding/decoding support in Flink, lack of inference Table schema DDL support and linage table schema upgrade automation.

After multiple large-scale high-tier production use cases, we propose to work closely with community and industry peers to upstream our patches. Including but not limited to

• efficient partial thrift encoding/decoding format, make room for future columns pruning optimization in both streaming and batch
• backward/forward compatible thrift binary <-> Row converter that can handle highly nested sophisticated typed schema definition without worry about upstream schema changes or state restore
• extend Table DDL and potentially View DDL to inference schema

Public Interfaces

TBD

Proposed Changes

Describe the new thing you want to do in appropriate detail. This may be fairly extensive and have large subsections of its own. Or it may be a few sentences. Use judgement based on the scope of the change.

Compatibility, Deprecation, and Migration Plan

• What impact (if any) will there be on existing users?
• If we are changing behavior how will we phase out the older behavior?
• If we need special migration tools, describe them here.
• When will we remove the existing behavior?

Test Plan

Describe in few sentences how the FLIP will be tested. We are mostly interested in system tests (since unit-tests are specific to implementation details). How will we know that the implementation works as expected? How will we know nothing broke?

Rejected Alternatives

If there are alternative ways of accomplishing the same thing, what were they? The purpose of this section is to motivate why the design is the way it is and not some other way.