Warnings

On this page

Warnings run after data is committed. They log results without failing the model. See Validation for how warnings fit into the three-stage model.

-- @warning: name(args)

Available Macros

Macro	Example	Description
`freshness(col, duration)`	`freshness(updated_at, 24h)`	Data staleness
`row_count(op, n)`	`row_count(>=, 100)`	Row count check
`row_count_delta(max_pct)`	`row_count_delta(20)`	Change percentage between runs
`sum_delta(col, max_pct)`	`sum_delta(amount, 10)`	Sum change ratio between runs
`mean_delta(col, max_pct)`	`mean_delta(amount, 15)`	Mean change ratio between runs
`approaching_limit(col)`	`approaching_limit(amount)`	Nearing configured threshold
`null_percent(col, max_pct)`	`null_percent(email, 5)`	NULL percentage
`not_null(col)`	`not_null(customer_id)`	NULL count
`unique(col)`	`unique(order_id)`	Duplicate detection
`min(col, op, val)`	`min(amount, >, 0)`	Minimum value
`compare(col, op, val)`	`compare(amount, >=, 0)`	Value comparison
`mean(col, op, val)`	`mean(amount, >=, 100)`	Average check
`low_entropy(col, min)`	`low_entropy(status, 1.0)`	Data diversity
`cardinality(col, op, n)`	`cardinality(region, >=, 3)`	Distinct value count
`in_list(col, 'vals')`	`in_list(status, 'active,inactive')`	Allowed values

Delta macros (row_count_delta, sum_delta, mean_delta) use DuckLake’s table_changes() to compare against the previous run.

Custom Macros

Defined in config/macros/warnings.sql. Convention: ondatra_warning_{name}(t, ...).

CREATE OR REPLACE MACRO ondatra_warning_high_refund_rate(t, col, max_pct) AS TABLE
  SELECT printf('high refund rate: %.1f%%', 100.0 * SUM(CASE WHEN c < 0 THEN 1 ELSE 0 END) / COUNT(*))
  FROM query(printf('SELECT %s AS c FROM %s', col, t))
  HAVING 100.0 * SUM(CASE WHEN c < 0 THEN 1 ELSE 0 END) / COUNT(*) > max_pct;

-- @warning: high_refund_rate(amount, 5)