# TASK 6.3: Implement Curated SQL Catalog, Execution, And Diagnostics

## Summary

Implement the execution boundary for the first `staging_to_curated` stage: run target-specific SQL for the resolved curated target set, apply the chosen rerun-safe curated replacement strategy, and publish safe per-target diagnostics that later publish-stage work can consume.

## Dependencies

- [TASK 6.2](./TASK6.2.md) - curated target dependency matrix and v1 field behavior

## Files to Create

### 1. Curated SQL catalog and diagnostics models

Create or finalize the components that select curated SQL templates, execute them, and return target-level diagnostics.

Recommended logical services:

- `BrinkPosCuratedSqlTemplateCatalog`
- `BrinkPosCuratedSqlExecutor`

Required diagnostics details:

- target name
- required staging tables
- SQL template identifier
- affected-row count or equivalent execution summary
- timing
- safe status summary

## Files to Modify

### 2. Warehouse curated execution path

Modify the tenant warehouse curated execution path so `staging_to_curated` can run the chosen replacement strategy and then execute target SQL for every required curated table.

### 3. Staging-to-curated batch details

Modify or finalize the `staging_to_curated` `details_json` shape so it preserves upstream load-to-staging references and per-target curated execution results.

## Required Behavior

- the chosen v1 curated replacement strategy is applied consistently before or alongside target SQL as required
- curated SQL executes for every required target in the resolved upstream inventory
- if any required target fails, the endpoint fails and the full `staging_to_curated` batch fails
- upstream staging state remains for replay and diagnostics
- diagnostics must not store secrets
- batch metadata must preserve enough detail for `TASK7.md` to resolve `publish_curated` inputs confidently

## Testing

### Unit tests

Add unit tests for:

- SQL template selection per target
- execution ordering of replacement strategy and target SQL
- per-target success and failure diagnostics recording
- failure behavior when one required target fails
- rerun-safe stage behavior and lifecycle integration
- diagnostics sanitization

## Acceptance Criteria

- [x] Curated SQL executes for every required target in the resolved upstream inventory
- [x] The chosen replacement strategy is applied consistently for reruns
- [x] Target failures fail the endpoint and stage correctly
- [x] Diagnostics stay secret-free and downstream-readable
- [x] Unit tests cover execution, rerun semantics, and diagnostics behavior

## Notes

- This subtask creates the first business-facing Redshift curated outputs for the slice.
- Keep curated SQL execution ownership in shared Warehouse code rather than in the Brink adapter.