> I think we should just accept a string like "bk:ledger:entry:batch_index" that represents a MessageID instead.
Sure, this is a good idea. I can follow this format without `bk:`. What do you think? Thanks, Zixuan Yunze Xu <x...@apache.org> 于2023年12月29日周五 12:34写道： > IMO, exposing the ledger id and entry id directly through the API is > not a good idea. For example, what if you want to get a chunked > message? You would have to add another field again and deprecate the > existing method. > > I think we should just accept a string like > "bk:ledger:entry:batch_index" that represents a MessageID instead. If > we're going to replace BK with other storage systems in future (though > I think it's far way from now), we can still reuse this API. > > Thanks, > Yunze > > On Thu, Dec 28, 2023 at 5:21 PM Zixuan Liu <zix...@apache.org> wrote: > > > > Hi Pulsar Community, > > > > I want to discuss the return value of the > > `org.apache.pulsar.client.admin.Topics#getMessageById`, which returns > > `Message<byte[]>`. If the message is a batch message, we only get the > first > > message from the batch message. > > > > So I want to change this behavior by returning `List<Message<byte[]>`, > and > > in the CLI, we also need to print multiple messages when using > > `get-message-by-id`. > > > > Now, I am adding a new method with the same name and a `batchIndex` > > parameter. When "batchIndex" is -1, we can retrieve all messages from the > > batch message. Otherwise, we get one message by giving the index. For > this > > implementation, please see https://github.com/apache/pulsar/pull/21813. > > > > In long time ago, there was also a PR to fix this issue, please see > > https://github.com/apache/pulsar/pull/19827. > > > > Two PR uses different ways to fix this issue, please let me know your > > thoughts! > > > > Thanks, > > Zixuan >
