单表复杂场景 Teradata 查询答案

【问题标题】：Complex Scenario Teradata Query From Single Table单表复杂场景 Teradata 查询
【发布时间】：2021-09-16 15:00:07
【问题描述】：

问题：

获取具有 type='Onsite Repair' 的记录以及具有仅当同一日期同一 Act_ID 中还存在 Type='Travel' 的行时，才键入（'Calibration'、'Interface Troubleshooting'、'Setup/Configuration'、'Customer Applications'）。否则忽略。（应始终显示现场维修类型的记录）

Type = "现场维修" 要么 Type = ("Calibration", "Interface Troubleshooting", "Setup/Configuration","Customer Applications") 仅当在同一日期同一 Act_ID 中还存在 Type="Travel" 的行时。否则忽略类型。

来源数据

ACT_ID	TYPE	START_TS	END_TS
ACTID1	Travel	7/20/2016 13:00	7/20/2016 15:30
ACTID1	Interface Troubleshooting	7/20/2016 15:30	7/20/2016 19:00
ACTID1	Travel	7/20/2016 19:00	7/20/2016 21:00
ACTID1	Travel	9/20/2016 13:00	9/20/2016 15:30
ACTID1	Onsite Repair	9/20/2016 15:30	9/20/2016 23:30
ACTID1	Travel	9/21/2016 13:00	9/21/2016 15:30
ACTID1	Onsite Repair	9/21/2016 15:30	9/21/2016 23:30
ACTID1	Travel	9/22/2016 13:00	9/22/2016 15:30
ACTID1	Onsite Repair	9/22/2016 15:30	9/22/2016 23:30

预期数据

ACT_ID	TYPE	START_TS	END_TS
ACTID1	Interface Troubleshooting	7/20/2016 15:30	7/20/2016 19:00
ACTID1	Onsite Repair	9/20/2016 15:30	9/20/2016 23:30
ACTID1	Onsite Repair	9/21/2016 15:30	9/21/2016 23:30
ACTID1	Onsite Repair	9/22/2016 15:30	9/22/2016 23:30

【问题讨论】：

到目前为止你尝试过什么？你知道怎么做吗：获取 type='Onsite Repair' 的记录以及 Type in ('Calibration', 'Interface Troubleshooting', 'Setup/Configuration', '客户应用'）
嗨@Nick.McDermaid，我在下面写了查询及其工作，但我正在寻找更简单的查询而不是执行自联接。 SELECT * FROM WHERE TYPE = 'Onsite Repair' UNION SELECT * FROM TABLE A INNER JOIN TABLE B ON A.ACT_ID = B.ACT_ID WHERE A.PROJ_NAM IN ('Calibration', 'Interface Troubleshooting', 'Setup/Configuration', '客户应用') AND B.TYPE='旅行' AND CAST(A.START_TS AS DATE FORMAT 'YYYY-MM-DD') = CAST(B.START_TS AS DATE FORMAT 'YYYY-MM-DD');跨度>

标签： sql teradata teradata-sql-assistant scenarios teradatasql

【解决方案1】：

Windowed Aggregate 的简单任务加上限定：

select *
from table
qualify Type = 'Onsite Repair'
     OR (Type = in ('Calibration','Interface Troubleshooting', 'Setup/Configuration','Customer Applications')
        -- row with Type='Travel'
          AND count(case when Type='Travel' then 1 end)
                         -- same Act_ID on the same date
              over (partition by Act_ID, cast(START_TS as date)  > 0
        )

【讨论】：

谢谢@dnoeth。该查询的答案集与我编写的查询答案集完全匹配，并且我认为是正确的。你能帮我了解一下合格部分的处理吗？我知道qualify 像过滤条件一样工作，但我对如何处理这部分更好奇（Type = in ('Calibration','Interface Troubleshooting', 'Setup/Configuration','Customer Applications') AND count(case when Type ='Travel' then 1 end) over (partition by Act_ID, cast(START_TS as date)) > 0
QUALIFY 在 WHERE/GROUP BY/HAVING 之后处理，Group Count 应用聚合逻辑而不丢失详细信息行。