Question

我尝试为postgresql 8.3创建一个plpgsql触发器，它在插入之前自动对表进行分区由id列

如果目标表不存在，将创建它，插入到那里

所以我创建了带有新表名的insert语句，如此

exec_insert := 'INSERT INTO '||TG_TABLE_SCHEMA||'.'||TG_RELNAME||'_'||destinationid||' VALUES('||NEW.*||')';
EXECUTE exec_insert;

导致错误：

ERROR:  NEW used in query that is not in a rule

我有两个问题：

甚至可以在EXECUTE中使用NEW，还是在声明中有错误？
如果它不可能，任何人都知道如何从NEW中获取值，所以我可以在声明中使用它们？我唯一想到的是使用information_schema检索主表的列名，然后尝试动态访问NEW的值 - 我也不知道如何：（

THX

Answer 1

您可以使用

...
_sql varchar(100)= 'insert into some_table values ($1.*)';
...
execute _sql using new;
...

我发现这隐藏在一些偏远地区：D。我正在使用9.1

Answer 2

此设置在这里工作正常（版本8.4和9.0）：

CREATE TABLE customer
(
  id bigserial NOT NULL,
  name text,
  datecreated timestamp with time zone DEFAULT now(),
  CONSTRAINT customer_pkey PRIMARY KEY (id) USING INDEX TABLESPACE pg_default
);

CREATE OR REPLACE FUNCTION test_trigger()
  RETURNS trigger AS
$BODY$
declare
    query   text;
begin
    query := 'INSERT INTO customer_' || substring(NEW.name,1,1) || ' VALUES(' || NEW.id || ','''|| NEW.name||''')';
    EXECUTE query;

    RETURN NULL;

    EXCEPTION
        WHEN undefined_table THEN
            query := 'CREATE TABLE "customer_' || substring(NEW.name,1,1) || 
            '"( CONSTRAINT "customer_' || substring(NEW.name,1,1) ||'_name_check" CHECK (lower(substring(name, 1, 1)) = ''' || substring(NEW.name,1,1) || '''::text)
            ) INHERITS (customer); CREATE INDEX "i_customer_' || substring(NEW.name,1,1) ||'_name" ON customer_' || substring(NEW.name,1,1) ||' USING btree(name);';
            EXECUTE query;

            query := 'INSERT INTO customer_' || substring(NEW.name,1,1) || ' VALUES(' || NEW.id || ','''|| NEW.name||''')';
            EXECUTE query;

            query := 'ANALYZE "customer_' || substring(NEW.name,1,1) || '"';
            EXECUTE query;

            RETURN NULL;

end;
$BODY$
  LANGUAGE plpgsql VOLATILE STRICT
  COST 100;

CREATE TRIGGER t_customer
  BEFORE INSERT OR UPDATE OR DELETE
  ON customer
  FOR EACH ROW
  EXECUTE PROCEDURE test_trigger();

CREATE TABLE customer
(
  id bigserial NOT NULL,
  name text,
  datecreated timestamp with time zone DEFAULT now(),
  CONSTRAINT customer_pkey PRIMARY KEY (id) USING INDEX TABLESPACE pg_default
);

CREATE OR REPLACE FUNCTION test_trigger()
  RETURNS trigger AS
$BODY$
declare
    query   text;
begin
    query := 'INSERT INTO customer_' || substring(NEW.name,1,1) || ' VALUES(' || NEW.id || ','''|| NEW.name||''')';
    EXECUTE query;

    RETURN NULL;

    EXCEPTION
        WHEN undefined_table THEN
            query := 'CREATE TABLE "customer_' || substring(NEW.name,1,1) || 
            '"( CONSTRAINT "customer_' || substring(NEW.name,1,1) ||'_name_check" CHECK (lower(substring(name, 1, 1)) = ''' || substring(NEW.name,1,1) || '''::text)
            ) INHERITS (customer); CREATE INDEX "i_customer_' || substring(NEW.name,1,1) ||'_name" ON customer_' || substring(NEW.name,1,1) ||' USING btree(name);';
            EXECUTE query;

            query := 'INSERT INTO customer_' || substring(NEW.name,1,1) || ' VALUES(' || NEW.id || ','''|| NEW.name||''')';
            EXECUTE query;

            query := 'ANALYZE "customer_' || substring(NEW.name,1,1) || '"';
            EXECUTE query;

            RETURN NULL;

end;
$BODY$
  LANGUAGE plpgsql VOLATILE STRICT
  COST 100;

CREATE TRIGGER t_customer
  BEFORE INSERT OR UPDATE OR DELETE
  ON customer
  FOR EACH ROW
  EXECUTE PROCEDURE test_trigger();

INSERT INTO customer(name) VALUES ('john'), ('peter');

postgresql：在EXECUTE的动态命令中使用NEW。*

2 个答案: