Kamu

Changed
- The `/query` REST API endpoint will:
- return `404 Not Found` on not found datasets
- return `400 Bad Request` on invalid SQL
- return `422 Unprocessable Content` on unrecognized request body fields

Added
- New repository `AccessTokenRepository` to work with new access tokens
- Middleware now accept new token format `Bearer ka_*`
- New Gql APIs to manage new access tokens
- `listAccessTokens` to fetch access tokens by account
- `createAccessToken` to create new access token for account
- `revokeAccessToken` to revoke existing access token

Added
- The `/query` REST API endpoint now supports:
- POST requests with all parameters being passed via body
- Specifying schema format
- Specifying `aliases` to associate table names with specific dataset IDs
- Returning and providing state information to achieve full reproducibility of queries

Added
- New `EthereumLogs` polling source allows to stream and decode log data directly from any ETH-compatible blockchain node
- See the updated `examples/reth-vs-snp500` example
- See the new [`datafusion-ethers`](https://github.com/kamu-data/datafusion-ethers) crate for implementation details
- Added E2E test infrastructure
- Added necessary components for managed run -- for startup, operations, and shutdown
Changed
- Upgraded to `arrow 52` and `datafusion 39`
- Improved binary data formatting in CLI table output - instead of the `<binary>` placeholder it will display an abbreviated hex values e.g. `c47cf6…7e3755`
- JSON and CSV formatters can now output binary data - it will be `hex`-encoded by default
- Hidden arguments and options are excluded from [the CLI reference](resources/cli-reference.md)
Fixed
- JSON formatter now properly supports `Decimal` types
- Stabilized startup using connection to databases
- Added HTTP middleware that wraps each request into a separate transaction
- Also added wrapping for some commands, in particular `kamu system generate-token`
- The structure of services that required lazy access to databases was reorganized:
- Extracted `PredefinedAccountsRegistrator` & `DatasetOwnershipServiceInMemoryStateInitializer`
- Fixed potential crash when attempting to rollback a transaction if the connection fails to establish

Fixed
- Fixed support of `--force` mode for pull/push actions using Smart Transfer Protocol

Added
- New `--reset-derivatives-on-diverged-input` flag to `kamu pull` command, which will trigger
compaction for derived dataset if transformation fails due to root dataset compaction and retry transformation
- Initial support for ingestion via file uploads, with local FS and S3-based storage for temporary files
Changed
- `AddPushSource` event may omit specifying a schema. In this case, the very first push ingestion invocation
would try to make a best-effort auto-inference of the data schema.
Fixed
- Fixed issue with smart protocol transfer operations upon empty datasets