Meltano

Hello,

I'm using custom tap/mapper/targets to replicate our Postgres database to Snowflake. We're not using off-the-shelf plugins because they were too slow for our use-cases.

Now, &gt;90% of our time is spent serializing/deserializing JSON. I'd like to give `BATCH` messages a try, but the doc say it's not supported by mappers.

I'm not sure _why_, because from `singer_sdk` it seems like my `InlineMapper` can override the `map_batch_message` method. So is it really impossible to do, or can I give it a shot?

You can give it a shot. The main obstacle is that mappers work on the Singer stream message-by-message, so if you need to transform records in a BATCH sync, you'll need to read the file, map it, then write the records back to batch. So, I would imagine you lose some of the benefits of BATCH mode that way.

You could of course also try writing a custom mapper that does not need to process records individually

Thanks, I'm gonna try. My mapper runs conditionally based on the stream, which means I can often skip it based on the `stream` property of the `BATCH` message!