是干不支持SQL的周期(或时,使用plpgsql)?

Today,I defined a function,using the insert statements in a loop. But HAWQ returned an error:

今天，我使用循环中的insert语句定义了一个函数。但是HAWQ犯了一个错误:

ERROR:  could not serialize unrecognized node type: 43983632 (outfast.c:4742)
CONTEXT:  SQL statement "insert into t(id,value) values(1,0.1)"
PL/pgSQL function "test_function" line 6 at SQL statement

I did some testing and found that when I use the 'insert statements' in the loop,it will be reported as a mistake. If I delete the relevant 'insert statements',It can run properly.

我做了一些测试，发现当我在循环中使用“insert语句”时，它会被报告为一个错误。如果我删除相关的“插入语句”，它可以正常运行。

Here is an example of a test:

这里有一个测试的例子:

 CREATE OR REPLACE FUNCTION test_function()
  RETURNS int AS
$BODY$
declare 
	number int;
begin		
	number := 1;
	while number <= 10 loop
		insert into t(id,value) values(1,0.1);
		number := number+1;
	end loop;
	return number;
end
$BODY$
  LANGUAGE plpgsql ;

Then I use 'select test_function();' to call the function.It will returned an error mentioned above.

然后使用'select test_function();'调用函数。它将返回上面提到的错误。

Does this mean that I can not use the SQL statements in a loop with plpgsql ?

这是否意味着我不能使用plpgsql循环中的SQL语句?

Thanks. Best regards.

谢谢。致以最亲切的问候。

2 个解决方案

#1

You'll want to avoid singleton statements with HAWQ but I am a little surprised it doesn't work. You'll want to use set based operations instead.

您将希望避免使用HAWQ中的单例语句，但我有点惊讶它不起作用。您将希望使用基于集合的操作。

    CREATE TABLE t (id int, value numeric);

    CREATE OR REPLACE FUNCTION test_function() RETURNS int AS
    $BODY$
    declare 
            number int;
    begin       
            insert into t (id, value) select 1, 0.1 from generate_series(1, 10);
            return 10;
    end
    $BODY$
    LANGUAGE plpgsql;

For such a simple example, you could use a sql function instead which has less overhead than plpgsql.

对于这样一个简单的示例，您可以使用一个sql函数，它的开销比plpgsql小。

    DROP FUNCTION test_function();

    CREATE OR REPLACE FUNCTION test_function() RETURNS void AS
    $BODY$
            insert into t (id, value) select 1, 0.1 from generate_series(1, 10);
    $BODY$
    LANGUAGE sql;

Both of these functions do all of the work in a single statement rather than executing 10 separate ones. I tested both in HAWQ and both work.

这两个函数在一个语句中完成所有的工作，而不是执行10个单独的语句。我在HAWQ和这两项工作中都进行了测试。

And here is a workaround if you must do singleton insert statements in a loop with HAWQ.

如果您必须使用HAWQ在循环中执行单例插入语句，这里有一个解决方案。

    CREATE OR REPLACE FUNCTION test_function()
      RETURNS int AS
    $BODY$
    declare 
            number int;
            v_sql text;
    begin       
            number := 1;
            while number <= 10 loop
                    v_sql := 'insert into t(id,value) values(1,0.1)';
                    execute v_sql;
                    number := number+1;
            end loop;
            return number;
    end
    $BODY$
      LANGUAGE plpgsql ;

#2

This is actually an issue that has been addressed in hawq 2.0. You may refer to latest https://github.com/apache/incubator-hawq for reference.

这实际上是hawq 2.0中已经解决的问题。您可以参考最新的https://github.com/apache/incubator-hawq。

Here are the result:

这里是结果:

CREATE TABLE t (id INT, value DOUBLE PRECISION);
CREATE TABLE

CREATE OR REPLACE FUNCTION test_function()
RETURNS int AS
$BODY$
declare
    number int;
begin
    number := 1;
    while number <= 10 loop
        insert into t(id, value) values(1, 0.1);
        number := number+1;
    end loop;

    return number;
end
$BODY$
LANGUAGE plpgsql;
CREATE FUNCTION



SELECT test_function();
 test_function
---------------
            11
(1 row)

SELECT * FROM t;
 id | value
----+-------
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
(10 rows)



SELECT * FROM test_function();
 test_function
---------------
            11
(1 row)

SELECT * FROM t;
 id | value
----+-------
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
  1 |   0.1
(10 rows)

#1