[BEAM-12664] Improve textio: Write sharding

• Type: Improvement
• Status: Open
• Priority: P3
• Resolution: Unresolved
• Component/s: sdk-go
• Assignee: Unassigned
• Reporter: Robert Burke
• Created: 27/Jul/21 04:21
• Updated: 04/Jun/22 20:56

Description

The other SDKs have implementations that shard files on write. So should the Go SDK. The feature is mentioned in the Beam Programming Guide:

https://beam.apache.org/documentation/programming-guide/#file-based-writing-multiple-files

It would be expedient to provide an Xlang TextIO implementation for the Go SDK compared to replicating the implementation in Go, at cost of some execution time performance.

Ideally it would be similarly generalized to simplify writing File Sinks.  File sinks are necessarily complex to provide a robust and reliable implementation

Current Go implementation.

https://github.com/apache/beam/blob/master/sdks/go/pkg/beam/io/textio/textio.go#L119

Python FileIO implementation:

https://github.com/apache/beam/blob/master/sdks/python/apache_beam/io/filebasedsink.py 

(Note iobase.Sink is deprecated, but is still suitable for file io.)

Java TextIO & FileIO:

https://github.com/apache/beam/blob/f8fbbfa309ac88848057de694d4cc1cba3eaa92a/sdks/java/core/src/main/java/org/apache/beam/sdk/io/TextIO.java#L1259 

https://github.com/apache/beam/blob/master/sdks/java/core/src/main/java/org/apache/beam/sdk/io/FileIO.java 

 

KafkaIO (example of writing Go SDK side wrapper for a xlang Java IO):

https://github.com/apache/beam/blob/master/sdks/go/pkg/beam/io/xlang/kafkaio/kafka.go 

 

General docs on writing sinks: https://beam.apache.org/documentation/io/developing-io-overview/#sinks