bigquery-local（本地大查询）

BigQuery Local Skill

Project-specific BigQuery conventions. Standard GoogleSQL syntax, bq CLI usage, JOIN patterns, window functions, BQML, and external tables are assumed knowledge.

1. Cost Guardrails

• YOU MUST: Use --dry_run before large queries
• YOU MUST: Avoid SELECT *; specify columns explicitly
• YOU MUST: Always filter on partition columns
• IMPORTANT: Partition pruning fails when functions wrap the partition column

- Bad: WHERE DATE(created_at) = '2024-01-01' - Good: WHERE created_at >= '2024-01-01' AND created_at < '2024-01-02'

• IMPORTANT: Use APPROX_COUNT_DISTINCT over COUNT(DISTINCT) when exact counts are not required

2. Table Design Conventions

• Combine partitioning (date) + clustering (up to 4 high-cardinality columns)
• Set require_partition_filter = TRUE on large tables
• Put smaller table on right side of JOIN (broadcast optimization)

3. Pricing Reference

• On-demand: $5/TB scanned
• Storage: Active $0.02/GB, Long-term $0.01/GB

4. Slot Usage Check

SELECT
  job_id,
  total_bytes_processed,
  total_slot_ms,
  TIMESTAMP_DIFF(end_time, start_time, SECOND) AS duration_sec
FROM `region-us`.INFORMATION_SCHEMA.JOBS
WHERE creation_time > TIMESTAMP_SUB(CURRENT_TIMESTAMP(), INTERVAL 1 DAY)
ORDER BY total_slot_ms DESC
LIMIT 10;

5. Reference Links

• Official docs: https://cloud.google.com/bigquery/docs
• Pricing: https://cloud.google.com/bigquery/pricing