Hi Gunnar, Thanks for your time and response... I think the problem you want to solve is the exposure of the ExecutionConfig (that can be mutated) no? The configuration is not mutated, we only need to know if objectReuse is enable. This is already expose on RuntimeContext we think to keep it similar to it to simplify any migrations, but as said, for this migration from ExecutionConfig we only need the isObjectReuseEnabled, and we could expose only this configuration..
Best regards, On 2023/01/13 15:50:09 Gunnar Morling wrote: > Hey Joao, > > Thanks for this FLIP! One question on the proposed interface changes: > is it expected that the configuration is *mutated* via the InitContext > passed to Sink::createWriter()? If that's not the case, how about > establishing a read-only contract representing the current > configuration and passing in that one instead? That would probably > deserve its own FLIP upon which yours here then would depend. Later > on, other contracts which effectively shouldn't modify a config could > use that one, too. > > Note I don't mean to stall your efforts here, but I thought it'd be a > good idea to bring it up and gauge the general interest in this. > > Best, > > --Gunnar > > Am Fr., 13. Jan. 2023 um 15:17 Uhr schrieb Joao Boto <b...@boto.pro>: > > > > Hi flink devs, > > > > I'd like to start a discussion thread for FLIP-287[1]. > > This comes from an offline discussion with @Lijie Wang, from FLIP-239[2] > > specially for the sink[3]. > > > > Basically to expose the ExecutionConfig and JobId on SinkV2#InitContext. > > This changes are necessary to correct migrate the current sinks to SinkV2 > > like JdbcSink, KafkaTableSink and so on, that relies on RuntimeContext > > > > Comments are welcome! > > Thanks, > > > > [1] > > https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=240880853 > > [2] > > https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=217386271 > > [3] https://issues.apache.org/jira/browse/FLINK-25421 >
