Runtime: Code quality burndown

[x] Support measure filters in metricsview package
[x] Deprecate metrics queries in queries package
[x] In runtime/metricsview, push unnest handling to dimension expression
[x] Use jsonval for metricsResolver
[ ] Migrate MetricsViewRows and MetricsViewSearch to use runtime/metricsview for query generation
[ ] Setup test fixtures for easily testing metrics queries against all backends
[ ] Migrate old metrics tests
[ ] Validation for advanced measures (including unnest)
[ ] Cache metrics view schema in state in reconciler
[ ] Migration of OLAP connector:
- Check connector.druid.dsn and olap_connector is set for every Druid project
- Remove DB driver and DSN from admin-side
- Change OLAP config logic in local.NewApp
[ ] Test unnest for Druid and DuckDB (must not add prefix on Druid)
[ ] Test windowed measure partitioned by unnested dimension
[ ] Fix dynamic time range of P1Y becomes narrower than no time range
[ ] Clean up parse_metrics_view.go code
[ ] Deprecate rillv1beta compiler
[ ] In resolverResult, the rowLimit can be changed to SetCap
[ ] In resolverResult, the MarshalJSON can easily be optimized
[ ] Rename OLAPStore.Execute to OLAPStore.Query
[ ] Remove Measures in Subquery, replace it with a recursive scan of names in Having
[ ] Fix ordering of messages in api.proto
[ ] Rename ServiceToken to ServiceAuthToken for consistency
[ ] Incorporate Security.QueryFilter into metricssql
[ ] Drop V2 suffixes in resources.proto
[ ] Remove ReadProfiling and ReadRepo permissions in JWT on cloud
[ ] Add E2E tests for security policies
[ ] Add an option to custom APIs about whether or not to skip security policies on the underlying resolver
[ ] In the admin server, cleanup and reduce duplication in the calls to NewToken
[ ] Implement pagination for ListResources
[ ] Fix the exports in runtime/resolvers/sql.go
[ ] Simplify trigger API
[ ] Support triggers for alerts and reports
[ ] Fix +1 second issue in alert execution times (thread)
[ ] Remove the DDL functions from OLAPStore and make internal functions
[ ] Cleanup CPU/memory config (spread multiple places – DuckDB should get from the instance, the instance should discover CPU/memory instead of being set by the creator)
[ ] Change CORS on local to only allow requests from frontend (code here)
[ ] Support DataYAML in reports
[ ] Remove unused features in runtime/drivers/blob and move to runtime/pkg
[ ] Change ObjectStore.ListObjects to return an iterator
[ ] Change ObjectStore.DownloadFiles to accept a fixed options struct
[ ] Rename ObjectStore.DownloadFiles to DownloadObjects
[ ] Change SQLStore.Query[AsFiles] to accept a fixed options struct
[ ] Change globutil.URL to have Bucket instead of Host
[ ] Remove the EmbedCatalog option
[ ] Remove implementation of CatalogStore on the DuckDB driver (always use SQLite)
[ ] Panics from reparse are not caught by controller because the watcher is invoked in a different goroutine (causes a crash)
[ ] Change runtime/server/auth to use the external URL as the audience. And use auth_service_url for issuer.
[ ] Support rill env [<project>] instead "rill env --project" for consistency
[ ] Parse mock_users: in rill.yaml on the backend for correct validation
[ ] Rename "organization" to "org" everywhere
[ ] Rename resource "kind" to "type" internally
[ ] Support exports for custom APIs
[ ] Support returning user-defined 4xx errors from custom APIs (maybe with a {{ error ... }} template func)
[ ] Finalize support for multi-level aggregations ("per" key in advanced measures)
[ ] Support join_to_spine for time series
[ ] Remove ModelMaterializeDelaySeconds
[ ] Remove code marked deprecated for a long time
[ ] Remove BucketPlanner, PullTrigger and other deprecated/orphaned resources
[ ] Document resources.proto
[ ] Add a /v1 prefix to the ad-hoc endpoints (auth, Github, etc.) on the admin and runtime servers
[ ] The rill project logs command fails with ugly error messages when it times out. Here's an example, but it sometimes looks different (maybe if enforced at the ingress level instead?): Error: failed to watch logs: rpc error: code = Unavailable desc = connection error: desc = "transport: Error while dialing: dial tcp XXX.XXX.XXX.XXX:443: i/o timeout" (Unavailable)
[ ] When running rill user list on an org you're not a member of, it still partially succeeds as a superuser, which is confusing. It should fail immediately instead.
[ ] Validate the measure format_presets on the backend during metrics view validation.
[ ] Avoid upgrade in rill upgrade if version is already latest.

rilldata / rill

Runtime: Code quality burndown #5053