Kamu

Added
- GraphQL API exposes current vocabulary as a part of dataset's metadata

Fixed
- Fixed GraphQL API issues with TransformInput structures with the alias name that is different from dataset name

Changed
- Improved reaction of `kamu inspect schema` and `kamu tail` on datasets without schema yet
- GraphQL schema and corresponding schema/tail responses now assume there might not be a dataset schema yet
- Web-platform version upgraded after GraphQL API changes

Fixed
- Lineage and transform now respect names of datasets in the `SetTransform` metadata events, that may be different from names in a workspace

Fixed
- Upgrade to Flink engine prevents it from crashing on checkpoints over 5 MiB large

Changed
- Major upgrade of Apache Flink version
- No updates to existing datasets needed, but the verifiability of some datasets may be broken (since we don't yet implement engine versioning as per ODF spec)
Added
- Flink now supports a much nicer temporal table join syntax:
sql
SELECT
t.event_time,
t.symbol,
p.volume as volume,
t.price as current_price,
p.volume * t.price as current_value
FROM tickers as t
JOIN portfolio FOR SYSTEM_TIME AS OF t.event_time AS p
WHERE t.symbol = p.symbol

- We recommend using `FOR SYSTEM_TIME AS OF` join syntax as replacement for old `LATERAL TABLE` joins
- Determinism of Flink computation should be improved