Wish she'd send me a snapshot of them toes

https://github.com/trinodb/trino/pull/27038

Here's the pull request

Hey everyone, I was actually taking another quick review of the code here, and my solution isn't the cleanest one.  You can just filter the offsets that you want to read in the query.  However, I was looking at the offset filtering code itself, and it takes a "summary" of any offset predicate that you give it (meaning the min and max across all partitions) and reads those from each partition.  Since the offsets from each partition are unrelated, my guess is that in practice you end up just reading them all.  What I did here is technically more efficient in "pushing the filter to Kafka", albeit with a worse UX.  I may have to come back to it.  That being said, the actual solution involves doing some custom parsing of WHERE statements in SQL to figure out which offsets to read for each topic partition, and then applying those, which is complicated.

Jordan has no life

A developer shares progress on adding snapshot read support to the Apache Trino Kafka connector. The core change introduces a new session variable and a `TopicPartitionOffsetProvider` interface to the `KafkaSplitManager` class, allowing users to specify exact start and end offsets per partition instead of always reading the full live topic. This enables repeatable, consistent reads from Kafka — a prerequisite for building a transactionally isolated data harness across multiple sources. The video also covers the open source contribution workflow (forking, PRs, CLAs), using Guice vs Dagger for dependency injection in Trino, and plans to next tackle time-travel/snapshot queries in the Trino Hudi connector.

Building Apache Trino Snapshot Reads Of Kafka Topics | Personal Projects With Ex-Google SWE, Ep 2