Postgres列必须出现在GROUP BY子句中，或者用于聚合函数错误。

Question

我有下表：

portal=# \d+ accounts_version
                                      Table "public.accounts_version"
          Column          |            Type             | Modifiers | Storage  | Stats target | Description 
--------------------------+-----------------------------+-----------+----------+--------------+-------------
 id                       | integer                     | not null  | plain    |              | 
 status                   | account_status              |           | plain    |              | 
 adwords_id               | character varying(20)       |           | extended |              | 
 nickname                 | character varying(48)       |           | extended |              | 
 account_budget           | double precision            |           | plain    |              | 
 remaining_account_budget | double precision            |           | plain    |              | 
 daily_budget             | double precision            |           | plain    |              | 
 currency                 | character varying(12)       |           | extended |              | 
 exchange_rate            | double precision            |           | plain    |              | 
 login                    | character varying(48)       |           | extended |              | 
 password                 | character varying(48)       |           | extended |              | 
 billing                  | character varying(48)       |           | extended |              | 
 auto_tag_on              | boolean                     |           | plain    |              | 
 mcc_id                   | integer                     |           | plain    |              | 
 vps_id                   | integer                     |           | plain    |              | 
 client_id                | integer                     |           | plain    |              | 
 transaction_id           | bigint                      | not null  | plain    |              | 
 end_transaction_id       | bigint                      |           | plain    |              | 
 operation_type           | smallint                    | not null  | plain    |              | 
 country                  | character varying(24)       |           | extended |              | 
 external_comment         | text                        |           | extended |              | 
 internal_comment         | text                        |           | extended |              | 
 batch                    | character varying(48)       |           | extended |              | 
 updated_at               | timestamp without time zone |           | plain    |              | 
Indexes:
    "accounts_version_pkey" PRIMARY KEY, btree (id, transaction_id)
    "ix_accounts_version_end_transaction_id" btree (end_transaction_id)
    "ix_accounts_version_operation_type" btree (operation_type)
    "ix_accounts_version_transaction_id" btree (transaction_id)

我正在尝试查询每个adwords_id的最新条目。

所以我开始写这样的SQL查询：

SELECT * FROM accounts_version GROUP BY adwords_id; 

我收到以下错误：

ERROR:  column "accounts_version.id" must appear in the GROUP BY clause or be used in an aggregate function
LINE 1: SELECT * FROM accounts_version

           ^

我正在运行postgres 9.5。知道是什么导致了这个吗？

Answer 1

PostgreSQL GROUP BY将列组表达式作为输入，当列为text类型且未在聚合函数中使用时会引发错误请参阅SELECT - GROUP BY- Transact-SQL

Answer 2

GROUP BY不是正确的方法。你想要的是DISTINCT ON。要获取每个updated_at的最新adwords_id行，请尝试以下操作：

SELECT DISTINCT ON (adwords_id) *
FROM accounts_version
ORDER BY updated_at DESCENDING;