CodingCat commented on PR #4795:
URL: https://github.com/apache/iceberg/pull/4795#issuecomment-1138954434
Hi, @rdblue , thanks for the reply
> The approach you linked to doesn't work because that is in the SQLConf,
which is at the session level. You'd need a separate session for every writer
thread. If that's what you'd like to do, then we can help you get that feature
in. Is that what you want?
yes, this is pretty much what we did with delta lake, every thread creates a
new SparkSession inherited from the parent one, and then add metadata in
SQLConf...I am not a big fan of this approach since it imposed creation of a
bunch of new SparkSessions across threads, that's also why I had the way in the
current PR
>
> I think that we may want to introduce a different way than using write
properties to set additional metadata, because of the SQL problem. We could add
a way to pass this information, like this:
>
> ```java
> public class CommitMetadata {
> private static final ThreadLocal<Map<String, String>>
COMMIT_PROPERTIES = ThreadLocal.withInitial(ImmutableMap::of);
>
> public static <R> R withCommitProperties(Map<String, String>
properties, Callable<R> callable) throws Exception {
> COMMIT_PROPERTIES.set(properties);
> try {
> return callable.call();
> } finally {
> COMMIT_PROPERTIES.set(ImmutableMap.of());
> }
> }
>
> static Map<String, String> commitProperties() {
> return COMMIT_PROPERTIES.get();
> }
> }
> ```
>
yeah, this looks neat! I can work on a PR for this
--
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]
