| 粒度 | 业务过程最细 |
|---|---|
| 处理 | 清洗+标准化 |
| 关联 | 维度代理键 |
| 命名 | dwd_域_过程_周期 |
DWD 加工骨架
DWD 加工骨架
INSERT OVERWRITE TABLE dwd_trade_order_di PARTITION (dt='2026-10-12')
SELECT o.order_id,
o.user_id,
COALESCE(d.customer_sk, -1) AS customer_sk,
CAST(o.pay_amt AS DECIMAL(18,2)) AS pay_amt,
COALESCE(m.std_code,'UNKNOWN') AS order_status
FROM (SELECT *, row_number() OVER (PARTITION BY order_id ORDER BY etl_load_time DESC) AS rn
FROM ods_order WHERE dt='2026-10-12') o
LEFT JOIN dim_customer d ON o.user_id = d.customer_id AND d.is_current=1
LEFT JOIN dim_code_map m ON o.order_status = m.src_code AND m.code_type='order_status'
WHERE o.rn = 1;枚举映射落成映射表，SQL 引用映射表而不是在每个脚本里硬编码 CASE。