当 unpivot 未检测到不同类型的字段时，Snowflake 如何投射 select 查询的所有字段？

Question

根据这个，我制作了以下脚本：

select
    OBSERVATION_DATE,
    name_col,
    md5(
        md5(
            concat_ws(
                '',
                val_col,
                DATE(OBSERVATION_DATE, 'DD/MM/YYYY'),
                'CAMP',
                CAMPNO
            )
        )
    ),
    current_timestamp(),
    NULL,
    'ONA',
    val_col,
    md5(
        concat_ws(
            '',
            'name_col',
            DATE(OBSERVATION_DATE, 'DD/MM/YYYY'),
            'CAMP',
            CAMPNO
        )
    )
from
    (
        select
            T.*
        FROM
            TEMP_TABLE_NAME T
    ) unpivot (val_col for name_col in (FACILITY_ID, CAMPNO));

它将帮助我转置一个临时 table 以将所有相关字段添加到另一个 table，其中每个 field/value 对代表一行。

我从这个查询中得到的错误是：

SQL compilation error: error line 2 at position 77 invalid identifier 'CAMPNO'

我想可能是因为 campNo 是一个数字，所有字段都应该转换为 varchar，所以我尝试了这个查询：

select cast(campno as varchar) as CAMPNO FROM TEMP_TABLE_NAME;

它起作用了，但是在第二个 select 语句中我得到了一个错误：

...from (select cast(CAMPNO as varchar) as campno, T.* FROM TEMP_TABLE_NAME T)

并得到一个错误：

SQL compilation error: ambiguous column name 'CAMPNO'

我需要将 select * from myTable 中的所有字段转换为 varchar。

Answer 1

您收到此错误的原因是，一旦您取消透视，就没有名为 campno 的列。您将该列取消透视为行：

设置示例

create or replace transient table source_table
(
    id               number,
    observation_date varchar,
    facility_id      varchar,
    campno           number
);

insert overwrite into source_table
values (1, '01/01/2020', 2, 23),
       (2, '02/01/2020', 3, 44),
       (3, '03/01/2020', 1, 123),
       (4, '04/01/2020', 1, 2233)
;

如果您从此来源 table 执行 select *，您将有一个名为 campno 和 facility_id:

的列

select st.* from source_table st;

-- Resulting table:
-- +--+----------------+-----------+------+
-- |ID|OBSERVATION_DATE|FACILITY_ID|CAMPNO|
-- +--+----------------+-----------+------+
-- |1 |01/01/2020      |2          |23    |
-- |2 |02/01/2020      |3          |44    |
-- |3 |03/01/2020      |1          |123   |
-- |4 |04/01/2020      |1          |2233  |
-- +--+----------------+-----------+------+

但是一旦你在 campno 和 facility_id 字段上 unpivot table，列名就变成了行，你就不能再 select campno, facility_id .您还需要确保要取消透视的列的数据类型相同（因此是子查询）：

select *
from (
    select
        id,
        observation_date,
        facility_id,
        campno::varchar as campno
    from source_table) unpivot ( val_col for name_col in (facility_id, campno) );

-- Resulting table:
-- +--+----------------+-----------+-------+
-- |ID|OBSERVATION_DATE|NAME_COL   |VAL_COL|
-- +--+----------------+-----------+-------+
-- |1 |01/01/2020      |FACILITY_ID|2      |
-- |1 |01/01/2020      |CAMPNO     |23     |
-- |2 |02/01/2020      |FACILITY_ID|3      |
-- |2 |02/01/2020      |CAMPNO     |44     |
-- |3 |03/01/2020      |FACILITY_ID|1      |
-- |3 |03/01/2020      |CAMPNO     |123    |
-- |4 |04/01/2020      |FACILITY_ID|1      |
-- |4 |04/01/2020      |CAMPNO     |2233   |
-- +--+----------------+-----------+-------+

看看上面怎么没有名为 campno 或 facility_id 的列？它们已被逆透视成行，不再是列...

我认为（但不确定）您可能正在寻找的是：

select
    id,
    observation_date,
    name_col,
    val_col,
    'ONA',
    md5(concat_ws('', id, name_col, val_col, observation_date)),
    current_timestamp()
from (
    select
        id,
        to_date(observation_date, 'DD/MM/YYYY') as observation_date,
        facility_id::varchar                    as facility_id,
        campno::varchar                         as campno
    from source_table
) unpivot ( val_col for name_col in (facility_id, campno) )

-- Resulting table:
-- +--+----------------+-----------+-------+-----+----------------------------------------------------------------+------------------------------------+
-- |ID|OBSERVATION_DATE|NAME_COL   |VAL_COL|'ONA'|MD5(MD5(CONCAT_WS('', ID, NAME_COL, VAL_COL, OBSERVATION_DATE)))|CURRENT_TIMESTAMP()                 |
-- +--+----------------+-----------+-------+-----+----------------------------------------------------------------+------------------------------------+
-- |1 |2020-01-01      |FACILITY_ID|2      |ONA  |19baf986df81f1818afae848cd14fc87                                |2021-03-01 09:59:45.919000000 -08:00|
-- |1 |2020-01-01      |CAMPNO     |23     |ONA  |1fcb518697772362a0dabcba7aacfa8a                                |2021-03-01 09:59:45.919000000 -08:00|
-- |2 |2020-01-02      |FACILITY_ID|3      |ONA  |60a82dbc3d1b78d09519fc50b26026cd                                |2021-03-01 09:59:45.919000000 -08:00|
-- |2 |2020-01-02      |CAMPNO     |44     |ONA  |cb03dc5d1df4e2548f26284c5ff339c2                                |2021-03-01 09:59:45.919000000 -08:00|
-- |3 |2020-01-03      |FACILITY_ID|1      |ONA  |fe0dd77e601f6f3bac4cde8da537eb3d                                |2021-03-01 09:59:45.919000000 -08:00|
-- |3 |2020-01-03      |CAMPNO     |123    |ONA  |95604e260fe1a69bc54100b08fee6d87                                |2021-03-01 09:59:45.919000000 -08:00|
-- |4 |2020-01-04      |FACILITY_ID|1      |ONA  |a94029663591b1c942e9f3be1467e04f                                |2021-03-01 09:59:45.919000000 -08:00|
-- |4 |2020-01-04      |CAMPNO     |2233   |ONA  |1d90ae0854a9042bf44906511e90ced8                                |2021-03-01 09:59:45.919000000 -08:00|
-- +--+----------------+-----------+-------+-----+----------------------------------------------------------------+------------------------------------+

md5(md5()) 毫无意义，我想我在您的另一篇帖子中看到过，所以不确定您为什么要这样做。

当 unpivot 未检测到不同类型的字段时，Snowflake 如何投射 select 查询的所有字段？

Snowflake how can we cast all fields of select query when unpivot is not detecting fields with a different types?

sql

unpivot

snowflake-cloud-data-platform