Fokko commented on code in PR #9695:
URL: https://github.com/apache/iceberg/pull/9695#discussion_r1689996232
##########
open-api/rest-catalog-open-api.yaml:
##########
@@ -537,6 +537,113 @@ paths:
5XX:
$ref: '#/components/responses/ServerErrorResponse'
+ /v1/{prefix}/namespaces/{namespace}/tables/{table}/preplan:
Review Comment:
I'm not sure if we really should force the client into certain directions.
For example, I think PyIceberg will mainly go to `plan` directly as it is more
likely to consume smaller tables than your typical ETL workloads.
> I think no, a server can decide to only implement one API. The server can
require the client to go through the preplan API by indicating a server
capability (e.g. preplanning), and in that case if the client still directly
hits the plan endpoint without a plan task, then a 421 Misdirected Request
should be thrown.
So then a redirect would be sufficient?
> And did we align for scan-planning capability that a server can choose to
implement none of these endpoints, can implement both preplan and plan, or just
implement plan?
I think there are situations, like Amogh described, where you don't want to
accumulate too much state on the server-side (for example, data sharing
servers), so in that case I would expect that it would only expose the plan
endpoint. But in such a case, the pre-plan could also be a no-op like Ryan
suggested.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
