- Fewer yields and thus less congestion during record handling.
- Doubles the insert throughput and results in roughly 10% faster reads.
- Enable file writes as part of transactions and batch-operations.
- Batch insertions now return a fake "skipped" id, when an insertion conflict was ignored based on the API's policy.
- This is more in line with the transaction endpoint's behavior but technically a breaking change. This will only matter to you if you're using batch insertions, with an ignore policy and relying on ids of ignored insertions being stripped from the response.
- In the future, we may want to remove batch insertions entirely in favor of the newer and more general transactions.
- Internal: allow more flexible error types from SQL executor reducing the need for (un)packing.
- Update dependencies.
Full Changelog: v0.34.7...v0.35.0