| 收益 | 查询免连接、口径集中 |
|---|---|
| 代价 | 存储放大、维度更新需回刷 |
| 适合 | 高频复用的核心分析主题 |
| 不适合 | 临时字段与长尾需求 |
宽表构建
宽表构建
INSERT OVERWRITE TABLE dws_trade_user_wide_1d PARTITION (dt='2026-10-11')
SELECT o.user_id, u.city, u.level, o.category,
COUNT(DISTINCT o.order_id) AS order_cnt,
SUM(o.pay_amount) AS pay_amt
FROM dwd_trade_order_detail_di o
LEFT JOIN dim_user u ON o.user_sk = u.user_sk
WHERE o.dt = '2026-10-11'
GROUP BY o.user_id, u.city, u.level, o.category;宽表字段要能说清来源与更新责任；字段越多回刷成本越高，建议分批扩展并保留回滚版本。