I am using dblink to move certain data between databases. Everything is save and sound but I am wondering if there is a more convenient way to define the column definition list of a dblink query result. I can do something like this:

SELECT *
FROM dblink('dbname=remote', 'select * from test')
AS t1(id integer, data text); 

The tables I'm interacting with have the same schema definition in both databases (remote & local). I was thinking of something like:

SELECT *
FROM dblink('dbname=remote', 'select * from test')
AS t1 LIKE public.test; 

Or:

SELECT *
FROM dblink('dbname=remote', 'select * from test')
AS t1::public.test;

The column definition list tends to become quite long. Is there something I may have overlooked?

EDIT:

As this has been a problem for me before I created a small function as a work-around.

CREATE OR REPLACE FUNCTION dblink_star_func(_conn text, _schema_name text, _table_name text)
RETURNS text
LANGUAGE PLPGSQL
VOLATILE STRICT
AS $function$
    DECLARE       
        _dblink_schema text;
        _cols          text; 
        _q             text;
        _func_name     text := format('star_%s', $3);
        _func          text;        
    BEGIN
        SELECT nspname INTO _dblink_schema
        FROM pg_namespace n, pg_extension e
        WHERE e.extname = 'dblink' AND e.extnamespace = n.oid;

        SELECT array_to_string(array_agg(column_name || ' ' || udt_name), ', ') INTO _cols
        FROM information_schema.columns
        WHERE table_schema = $2 AND table_name = $3;

        _q := format('SELECT * FROM %I.dblink(%L, %L) AS remote (%s)',
            _dblink_schema,
            _conn,
            format('SELECT * FROM %I.%I', $2, $3),
            _cols
        );

        _func := $_func$
            CREATE OR REPLACE FUNCTION %s()
            RETURNS SETOF %I.%I
            LANGUAGE SQL
            VOLATILE STRICT
            AS $$ %s; $$
        $_func$;

        EXECUTE format(_func, _func_name, $2, $3, _q);

        RETURN _func_name;
    END;
$function$;

This function creates and yields a function that wraps the dblink call. It's certainly not meant for heavy lifting but convenience.It would be nice if it turns out it's not necessary at all.

> select dblink_star_func('dbname=ben', 'public', 'test');
┌──────────────────┐
│ dblink_star_func │
├──────────────────┤
│ star_test        │
└──────────────────┘
(1 row)

> select * from star_test() where data = 'success';
┌────┬─────────┐
│ id │  data   │
├────┼─────────┤
│  1 │ success │
└────┴─────────┘
(1 row)
有帮助吗?

解决方案 2

You might need to make sure that your types are always in sync but this should work:

SELECT (t1::test).* 
  FROM dblink('dbname=remote', 'select * from test') AS t1;

The key is that often you need parentheses to ensure that the parser knows you are dealing with tuples.

For example this works for me:

 CREATE TABLE test (id int, test bool);
 select (t1::test).* from (select 1, true) t1;

But this throws a syntax error:

 select t1::test.* from (select 1, true) t1;

其他提示

Try something like this :

select (rec).* from dblink('dbname=...','select myalias from foreign_table
  myalias') t1 (rec local_type)

Example (get tables stats from other database) :

select (rec).* from dblink('dbname=foreignDb','select t1 from
  pg_stat_all_tables t1') t2 (rec pg_stat_all_tables)
许可以下: CC-BY-SA归因
不隶属于 StackOverflow
scroll top