How to return multiple results in Postgres but I need to use a variable

Question

我有 Microsoft SQL 背景。所以我想做的是我知道我可以在 MSSql 中做的事情，但我非常想找出 Postgres 中的等价物（截至本文发布时的最新 docker 版本）。

这是我的 MSSql.

DECLARE @id INTEGER

SELECT @id = Id FROM SomeTable where UniqueName = 'foo' // UniqueName is unique (bet you didn't guess that)

SELECT * FROM SomeTable WHERE Id = @id

SELECT * FROM AnotherTable WHERE Id = @id

SELECT * FROM FinalTable WHERE Id = @id

这可能吗？

最初，我尝试使用 CTE（等效）：

WITH vars AS(
    SELECT Id FROM SomeTable where UniqueName = 'foo' 
)

SELECT * FROM SomeTable WHERE Id = vars.Id;

SELECT * FROM AnotherTable WHERE Id = vars.Id;

SELECT * FROM FinalTable WHERE Id = vars.Id;

但是我在第二个查询中得到一个错误：

SELECT * FROM AnotherTable

说它不知道

vars

了。

那么，在 Postgres 中可以做到这一点吗？

---

编辑：所以我认为这可能可以通过 PL/pgSQL（PostgreSQL 的扩展）实现。我试过这个：

DO $$
DECLARE
    @id integer
BEGIN

    SELECT Id INTO id FROM SomeTable WHERE UniqueName = 'foo'

    SELECT * FROM SomeTable WHERE Id = id ;

    SELECT * FROM AnotherTable WHERE Id = id ;

    SELECT * FROM FinalTable WHERE Id = id ;

END $$;

但这无法返回任何数据（是的，我可以确认此查询有一些数据）：

ERROR:  query has no destination for result data
HINT:  If you want to discard the results of a SELECT, use PERFORM instead.
CONTEXT:  PL/pgSQL function inline_code_block line 8 at SQL statement 

SQL state: 42601

所以我真的迷路了，现在。

Answer 1

不幸的是，似乎无法使用 WITH 语句，您可以通过类似的方式修复它：

SELECT set_config('my.vars.id', Id::char, False) FROM SomeTable WHERE UniqueName = 'foo';

SELECT * FROM SomeTable WHERE Id = current_setting('my.vars.id')::int;

SELECT * FROM AnotherTable WHERE Id = current_setting('my.vars.id')::int;

SELECT * FROM FinalTable WHERE Id = current_setting('my.vars.id')::int;

Answer 2

WITH…AS

构造将查询结果放入一种临时表中，其结果可用于对另一个表的查询。不过，这都是一个语句，所以你不能像那样使用结果三次。

WITH vars AS(
    SELECT Id FROM SomeTable where UniqueName = 'foo' 
)
SELECT * FROM SomeTable WHERE Id in (select id from vars);

https://www.postgresql.org/docs/current/queries-with.html

Answer 3

这种功能是可以的：

CREATE OR REPLACE FUNCTION fun_test(inputName text) RETURNS setof refcursor AS
$$
DECLARE foundid INTEGER;
DECLARE result1 refcursor;
DECLARE result2 refcursor;
DECLARE result3 refcursor;
BEGIN
    SELECT id INTO foundid FROM SomeTable WHERE UniqueName = inputName;
    result1 = 'First resultset';
    result2 = 'Second resultset';
    result3 = 'Third resultset';
    OPEN result1 FOR SELECT * FROM Table1 WHERE id = foundid;
    RETURN NEXT result1;
    OPEN result2 FOR SELECT * FROM Table2 WHERE id = foundid;
    RETURN NEXT result2;
    OPEN result3 FOR SELECT * FROM Table3 WHERE id = foundid;
    RETURN NEXT result3;
END;
$$ LANGUAGE plpgsql;

检索结果是这样完成的：

• 在像 pgAdmin 这样的 SQL 客户端上，在事务中一次执行 1

  FETCH

  （如果您一次执行整个事情，pgAdmin 将只显示最后一个结果集）：

  BEGIN;
  SELECT fun_test('foo');
  FETCH ALL IN "First resultset";
  FETCH ALL IN "Second resultset";
  FETCH ALL IN "Third resultset";
  COMMIT;
  

• 如果您从事软件开发，您可以一次发送所有查询并获得返回的多个结果集，其方法类似于我使用 C++/Qt 编写的以下示例：

  if (database.open()) {
      QSqlQuery query(database);
      query.exec(
          "SELECT fun_test('foo');" \
          "FETCH ALL IN \"First resultset\";" \
          "FETCH ALL IN \"Second resultset\";" \
          "FETCH ALL IN \"Third resultset\";"
      );
      for (auto rsi = 0; rsi < 3; ++rsi) {
          //By calling nextResult() first, we ignore the first resultset
          //that contains the identifiers of the next resultsets.
          //We may sometimes want to read it though,
          //in which case the for loop must be altered.
          query.nextResult(); 
          if (query.first()) {
              do {
                  //Do something such as:
                  qDebug() << query.record();
              } while (query.next());
          }
      }
  }
  

  如您所见，您甚至不需要事务，因为所有内容都作为单个查询执行。
  在实际情况下，您可能需要 3 个独立的

  do { ... } while(...);

  循环，每个循环负责自己的记录集。
  示例：

  query.exec(
      "SELECT fun_test('foo');" \
      "FETCH ALL IN \"First resultset\";" \
      "FETCH ALL IN \"Second resultset\";" \
      "FETCH ALL IN \"Third resultset\";"
  );
  query.nextResult();
  if (query.first()) {
      do {
          //Do something with records from First resultset
      } while (query.next());
  }
  query.nextResult();
  if (query.first()) {
      do {
          //Do something with records from Second resultset
      } while (query.next());
  }
  query.nextResult();
  if (query.first()) {
      do {
          //Do something  with records from Third resultset
      } while (query.next());
  }
  query.finish();