使用 postgres 通过一个查询将数据插入到多个表中

Question

我想一次性将数据插入到多个相关表中 - 在单个事务中。我正在使用 Postgres 和 Dapper ORM。这是我的查询：

WITH userins AS (
    INSERT INTO public."user"
(
    "FirstName", 
    "LastName", 
    "Email", 
    "PasswordHash", 
    "Address", 
    "City", 
    "State", 
    "Zip", 
    "Phone", 
    "IsEnabled", 
    "IsVerified", 
    "IsDeleted",  
    "CreatedDate", 
)
    VALUES (@FirstName, @LastName, @Email, @PasswordHash, @Address, @City, @State, @Zip, @Phone, @IsEnabled, ?IsVerified, ?IsDeleted, @CreatedDate);
    RETURNING id AS user_id
   )
, clientins AS (
    INSERT INTO public.client( "Description", "PlanCustomerId", "IsActive", "IsDeleted", "CreatedBy", "CreatedDate", userid)
                           VALUES(@Description, @PlanCustomerId, @IsActive,@IsDeleted, user_id, @CreatedDate,user_id) RETURNING id as client_id;
   )
   , clientsubins
   (
    INSERT INTO public.client_subscription(
    "PlanId", 
    "IsActive",
    "StartDate", 
    "EndDate", 
    "IsDeleted", 
    "CreatedBy", 
    "CreatedDate", 
    clientid, 
    subscriptionid)
    VALUES (@PlanId, @IsActive, @StartDate, @EndDate, @IsDeleted, user_id, @CreatedDate, client_id, @subscriptionid);
   )
   RETURNING client_id

此查询是否有效或我需要更改什么？

Answer 1

postgreSql 对此有不同的语法。几天前我也陷入了困境，想出了如何处理它。您可以使用 BEGIN 块将数据插入到多个表中，请按照此查询进行操作。

BEGIN;
                WITH userins AS (
                INSERT INTO public.user(
                ""FirstName"", ""LastName"", ""Email"", ""PasswordHash"",""VerificationCode"", ""Phone"", ""IsEnabled"", ""IsVerified"", ""IsDeleted"",  ""CreatedDate"")VALUES(@FirstName, @LastName, @Email, @PasswordHash, @VerificationCode, @Phone, @IsEnabled, @IsVerified, @IsDeleted, @CreatedDate)
                    RETURNING id AS user_id)
                ,clientins AS(
                INSERT INTO public.client(""Description"", ""PlanCustomerId"", ""IsActive"", ""IsDeleted"", ""CreatedBy"", ""CreatedDate"", UserId) 
                        VALUES(@Description, @PlanCustomerId, @IsActive, @IsDeleted, (SELECT user_id from userins), @CreatedDate, (SELECT user_id from userins)) RETURNING id as client_id)
                ,clientsubins AS
                (
                   INSERT INTO public.client_subscription(""PlanId"", ""IsActive"",""StartDate"", ""EndDate"", ""IsDeleted"", ""CreatedBy"", ""CreatedDate"", ClientId, SubscriptionId)
                   VALUES(@PaymentMethodPlanId, @IsActive, @StartDate, @EndDate, @IsDeleted, (SELECT user_id from userins), @CreatedDate, (SELECT client_id from clientins), @PlanId)
                )
                ,clientpurchaseins AS
                (
                   INSERT INTO public.client_purchase_history(""PlanId"", ""InvoiceId"",""StartDate"", ""EndDate"", ""IsDeleted"", ""CreatedBy"", ""CreatedDate"", ClientId)
                   VALUES(@PlanId, @InvoiceId, @StartDate, @EndDate, @IsDeleted, (SELECT user_id from userins), @CreatedDate, (SELECT client_id FROM clientins))
                )
                SELECT client_id from clientins;
        COMMIT;

如果发生任何错误，Begin block 将处理您的事务，然后它将自动回滚。

Answer 2

最重要的是，您需要使用早期 CTE 中的 SELECT 来获取 RETURNING 子句产生的值：

WITH userins AS (
   INSERT INTO public."user"
          ("FirstName", "LastName", "Email", "PasswordHash", "Address", "City", "State", "Zip", "Phone", "IsEnabled", "IsVerified", "IsDeleted", "CreatedDate") --  no dangling ,
   VALUES (@FirstName , @LastName , @Email , @PasswordHash , @Address , @City , @State , @Zip , @Phone , @IsEnabled , @IsVerified , @IsDeleted , @CreatedDate)  --  no ;
   RETURNING id AS userid
   )
 , clientins AS (
   INSERT INTO public.client
         ("Description", "PlanCustomerId", "IsActive", "IsDeleted", "CreatedBy", "CreatedDate", userid)
   SELECT @Description , @PlanCustomerId , @IsActive , @IsDeleted , userid     , @CreatedDate , userid
   FROM   userins
   RETURNING id AS clientid, userid  -- no ;
   )
INSERT INTO public.client_subscription
      ("PlanId", "IsActive","StartDate", "EndDate", "IsDeleted", "CreatedBy", "CreatedDate", clientid , subscriptionid)
SELECT @PlanId , @IsActive , @StartDate, @EndDate , @IsDeleted , userid     , @CreatedDate , clientid,  @subscriptionid
FROM   clientins
RETURNING clientid;

您还遇到了一些语法错误。并且我统一了clientid等的拼写。

每个 CTE（以及外部的 INSERT）取决于前一个 CTE 的结果行。这样，如果第一个 INSERT 没有插入（并返回）任何内容，则不会插入任何内容。

由于这是一个单个语句，它会自动在单个事务中运行。任何错误都会取消整个操作。

甚至可以在过时的 Postgres 9.3 中使用。

见：

• Insert data in 3 tables at a time using Postgres

Answer 3

您收到的答案很好，但是我会给出我的建议。由于您正在处理事务块，因此最好将其作为函数或存储过程的一部分包含在内。例如，我更喜欢将返回的 id 分配到一个变量中，这样我就可以在整个执行过程中使用它。如果一切正常，我提交并返回 user_id，或者如果需要，我也可以将返回类型更改为布尔值。如果有异常，我捕获它并返回-1，表示不成功，这里也可以执行其他选项，比如在一个表中注册异常的原因。如果一个表失败，则全部回滚：

CREATE OR REPLACE FUNCTION test_function(p_name CHAR(50), p_other_params CHAR(50)) 
RETURNS INT AS
$$
DECLARE
  new_user_id INT;
  new_purchase_id INT;
BEGIN
    INSERT INTO public.user(name)
    VALUES(p_name) 
    RETURNING public.user.id INTO new_user_id;
    
    INSERT INTO public.purchase(user_id)
    VALUES(new_user_id) 
    RETURNING public.purchase.id INTO new_purchase_id;
    
     INSERT INTO public.purchase_detail(purchase_id)
    VALUES(new_purchase_id);
    
    RAISE EXCEPTION 'Exception when inserting user: %', p_name; --remove to test when there is no issue
    --commited if reaches this point
    RETURN new_user_id;
    
    EXCEPTION WHEN OTHERS THEN
    --ROLLED back if exception
    RETURN -1; -- this value indicates the app it failed
END;
$$
LANGUAGE plpgsql;
          
 select * from test_function('SomeName', 'OtherParameter');
 select id, name from public.user;
 select id, user_id from public.purchase;
 select id, purchase_id from public.purchase_detail;

这是架构：

CREATE TABLE public.user(  name CHAR(50)); 
CREATE TABLE public.purchase( user_id int);
CREATE TABLE public.purchase_detail( purchase_id INT);
ALTER TABLE public.user ADD COLUMN id SERIAL PRIMARY KEY;
ALTER TABLE public.purchase ADD COLUMN id SERIAL PRIMARY KEY;
ALTER TABLE public.purchase_detail ADD COLUMN id SERIAL PRIMARY KEY;

检查this fiddle in action, 并记住删除 RAISE Exception 行以在没有异常时进行测试。 注意：由于某种原因，小提琴中的 $$ 应该用引号代替，因此里面的单引号是双引号。

更多关于exception handling（40.6.6. 捕获错误）。

Answer 4

如果您能说明为什么要在一个事务中插入多个表的总体意图，将会有所帮助。

我不想听起来很无聊，但是要充分利用某些东西并有效地使用它，您首先应该对正在使用的组件有一个合理到高级的理解。可悲的是，在当今世界，很多人只是去“做事”，然后不得不寻求帮助，因为他们使用的技术没有做他们想做的事，而是没有配置它来做到这一点。我不能也不会判断这里是否是这种情况。

无论如何。如果您想确保数据更改是否应用于所有相关表，对我来说第一件显而易见的事情就是查看dapper transactions。这是否进行原子事务（所以应用所有更改或不应用）？我不知道，希望如此。

您可以像其他发帖人所建议的那样，将 DML（数据操作语言）语句包装在“事务块”中，如果发生错误，这将回滚事务块中的所有更改，或者将其包装在一个函数中，并自己处理错误。