sl_capture

1# Semantic Layer - Capture
2 
3This skill covers **when** and **how** to capture new patterns into the semantic layer. For schema reference and query grammar, load the `sl` skill first.
4 
5When the current turn produces a reusable pattern (business metric, derived view, join pattern, computed dimension), capture it so future queries can reach for it instead of rediscovering it.
6 
7## SQL dialect
8 
9The user-facing prompt includes a `Warehouse:` line under the SL Sources index
10(e.g. `Warehouse: BIGQUERY`). All `expr` strings - measure expressions, segment
11predicates, computed-column SQL - execute on that warehouse and must use its
12syntax. Date arithmetic in particular varies by dialect:
13 
14- **BigQuery**: `transaction_date >= TIMESTAMP_SUB(CURRENT_TIMESTAMP(), INTERVAL 90 DAY)` (when the column is `TIMESTAMP`); `event_date >= DATE_SUB(CURRENT_DATE(), INTERVAL 90 DAY)` (when `DATE`).
15- **Postgres / Redshift**: `transaction_date >= current_date - interval '90 days'`.
16- **Snowflake**: `transaction_date >= dateadd(day, -90, current_timestamp())`.
17 
18Match the column's manifest type (`type: time` → TIMESTAMP/DATETIME on the
19warehouse) - comparing TIMESTAMP to a DATE-arithmetic result fails on
20BigQuery. After every `sl_edit_source`/`sl_write_source`, the inline validator runs a
21`LIMIT 1` warehouse probe per measure and surfaces dialect mismatches; if
22you see an error trailer, fix the expression and retry rather than leaving
23the source for the post-squash gate to revert.
24…