Use this skill when the OpenART registry selects tool.sql_queries.1652b99c183e58c4 for the current task.

SQL Queries Skill

Write correct, performant, readable SQL across all major data warehouse dialects.

Dialect-Specific Reference

PostgreSQL

• Date arithmetic: date_column + INTERVAL '7 days', DATE_TRUNC('month', created_at)
• String: ILIKE, regex with ~, SPLIT_PART()
• JSON: data->>'key', data#>>'{path,to,key}'
• Arrays: ARRAY_AGG(), ANY(), @> containment
• Performance: EXPLAIN ANALYZE, partial indexes, EXISTS over IN

Snowflake

• Date arithmetic: DATEADD(day, 7, date_column), DATEDIFF()
• Semi-structured: column:key::string dot notation, LATERAL FLATTEN()
• Performance: clustering keys, partition pruning, RESULT_SCAN()

BigQuery

• Date arithmetic: DATE_ADD(), DATE_DIFF(), DATE_TRUNC(date, MONTH)
• No ILIKE — use LOWER(), REGEXP_CONTAINS()
• Arrays: UNNEST(), ARRAY_AGG()
• Performance: always filter on partition columns, APPROX_COUNT_DISTINCT(), avoid SELECT *

Redshift

• Date arithmetic: DATEADD(), DATEDIFF(), DATE_TRUNC()
• String: ILIKE, LISTAGG()
• Performance: DISTKEY, SORTKEY, ANALYZE, VACUUM

Databricks SQL

• Delta Lake: time travel with TIMESTAMP AS OF, MERGE INTO for upserts
• Performance: OPTIMIZE, ZORDER, CACHE TABLE

Common SQL Patterns

Window Functions

ROW_NUMBER() OVER (PARTITION BY user_id ORDER BY created_at DESC)
SUM(revenue) OVER (ORDER BY date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW)
LAG(value, 1) OVER (PARTITION BY entity ORDER BY date)
revenue / SUM(revenue) OVER () as pct_of_total

CTEs for Readability

WITH
base_users AS (SELECT ...),
user_metrics AS (SELECT ... FROM base_users ...),
summary AS (SELECT ... FROM user_metrics ...)
SELECT * FROM summary;

Cohort Retention

WITH cohorts AS (
    SELECT user_id, DATE_TRUNC('month', first_activity_date) as cohort_month
    FROM users
),
activity AS (
    SELECT user_id, DATE_TRUNC('month', activity_date) as activity_month
    FROM user_activity
)
SELECT cohort_month, COUNT(DISTINCT user_id) as cohort_size,
    COUNT(DISTINCT CASE WHEN activity_month = cohort_month THEN user_id END) as month_0,
    COUNT(DISTINCT CASE WHEN activity_month = cohort_month + INTERVAL '1 month' THEN user_id END) as month_1
FROM cohorts LEFT JOIN activity USING (user_id)
GROUP BY cohort_month;

Funnel Analysis

Use conditional aggregation with MAX(CASE WHEN event = 'X' THEN 1 ELSE 0 END) per step, then compute conversion rates between steps.

Deduplication

WITH ranked AS (
    SELECT *, ROW_NUMBER() OVER (PARTITION BY entity_id ORDER BY updated_at DESC) as rn
    FROM source_table
)
SELECT * FROM ranked WHERE rn = 1;

Error Handling and Debugging

1. Syntax errors: Check dialect-specific syntax differences
2. Column not found: Verify column names, check case sensitivity
3. Type mismatches: Cast explicitly (CAST(col AS DATE))
4. Division by zero: Use NULLIF(denominator, 0)
5. Ambiguous columns: Always qualify with table alias in JOINs
6. Group by errors: All non-aggregated columns must be in GROUP BY