[
https://issues.apache.org/jira/browse/CALCITE-4500?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17286102#comment-17286102
]
Julian Hyde commented on CALCITE-4500:
--------------------------------------
Thanks for expanding the description, [~sirinath].
If we are looking for a compact and efficient format for record data in
Avatica, then we should also consider Apache Arrow. An advantage of Arrow is
that the data can be directly consumed by many data science libraries.
Snowflake has recently adopted Arrow as its driver's format, for this reason.
For RPC message packets in Avatica, I think Protobuf is working well.
Independent of whether SBE and FBE are are used in Avatica, there could be
Calcite adapters that convert SBE and FBE data into tables (just as we have
have adapters that read CSV, JSON and InnoDB data).
> Support Simple Binary Encoding (SBE) and Fast Binary Encoding (FBE)
> -------------------------------------------------------------------
>
> Key: CALCITE-4500
> URL: https://issues.apache.org/jira/browse/CALCITE-4500
> Project: Calcite
> Issue Type: New Feature
> Reporter: Suminda Dharmasena
> Priority: Major
>
> Can support be added for:
> - [Simple Binary Encoding
> (SBE)|https://github.com/real-logic/simple-binary-encoding]
> - [Fast Binary Encoding
> (FBE)|https://github.com/chronoxor/FastBinaryEncoding]
>
> {quote}Avatica’s wire protocols are JSON or Protocol Buffers over HTTP. The
> Java implementation of the JSON protocol uses
> [Jackson|https://github.com/FasterXML/jackson] to convert request/response
> command objects to/from JSON.
> {quote}
>
> Binery protocols are more efficient. 2 of he most efficient protocols are SBE
> and FBE.
> Also the best Json implementation is DSL JSON or Jsoniter as per benchmarks:
> https://github.com/fabienrenaud/java-json-benchmark. So it would be
> worthwhile to replace Jackson with one of them.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
