error when using DELETE USING

[jamatthews]

After upgrading from 1.3.2 to 1.4.2 DELETE USING no longer works and throws an error:

BEGIN;
DELETE FROM table_1
USING table_2
WHERE table_1.partiton_column = 1 AND table_2.partition_column = 1 AND table_1.id = table_2.id;
ERROR: variable not found in subplan target lists

[funbringer]

We've created a hotfix branch called master_hotfix_delete_using, could you try it?

[funbringer]

I've just tested the following case (master branch, PostgreSQL 9.6.4):

create table test (id int not null, val float8);
select create_range_partitions('test', 'id', 1, 10, 10);

create table ids (id int);


explain (verbose) delete from test using ids where test.id = ids.id;                 
                                   QUERY PLAN                                    
---------------------------------------------------------------------------------
 Delete on public.test  (cost=0.01..1303.34 rows=20401 width=12)
   Delete on public.test
   Delete on public.test_1
   Delete on public.test_2
   Delete on public.test_3
   Delete on public.test_4
   Delete on public.test_5
   Delete on public.test_6
   Delete on public.test_7
   Delete on public.test_8
   Delete on public.test_9
   Delete on public.test_10
   ->  Hash Join  (cost=0.01..45.09 rows=1 width=12)
         Output: test.ctid, ids.ctid
         Hash Cond: (ids.id = test.id)
         ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
         ->  Hash  (cost=0.00..0.00 rows=1 width=10)
               Output: test.ctid, test.id
               ->  Seq Scan on public.test  (cost=0.00..0.00 rows=1 width=10)
                     Output: test.ctid, test.id
   ->  Hash Join  (cost=67.38..125.83 rows=2040 width=12)
         Output: test_1.ctid, ids.ctid
         Hash Cond: (test_1.id = ids.id)
         ->  Seq Scan on public.test_1  (cost=0.00..30.40 rows=2040 width=10)
               Output: test_1.ctid, test_1.id
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
   ->  Hash Join  (cost=67.38..125.83 rows=2040 width=12)
         Output: test_2.ctid, ids.ctid
         Hash Cond: (test_2.id = ids.id)
         ->  Seq Scan on public.test_2  (cost=0.00..30.40 rows=2040 width=10)
               Output: test_2.ctid, test_2.id
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
   ->  Hash Join  (cost=67.38..125.83 rows=2040 width=12)
         Output: test_3.ctid, ids.ctid
         Hash Cond: (test_3.id = ids.id)
         ->  Seq Scan on public.test_3  (cost=0.00..30.40 rows=2040 width=10)
               Output: test_3.ctid, test_3.id
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
   ->  Hash Join  (cost=67.38..125.83 rows=2040 width=12)
         Output: test_4.ctid, ids.ctid
         Hash Cond: (test_4.id = ids.id)
         ->  Seq Scan on public.test_4  (cost=0.00..30.40 rows=2040 width=10)
               Output: test_4.ctid, test_4.id
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
   ->  Hash Join  (cost=67.38..125.83 rows=2040 width=12)
         Output: test_5.ctid, ids.ctid
         Hash Cond: (test_5.id = ids.id)
         ->  Seq Scan on public.test_5  (cost=0.00..30.40 rows=2040 width=10)
               Output: test_5.ctid, test_5.id
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
   ->  Hash Join  (cost=67.38..125.83 rows=2040 width=12)
         Output: test_6.ctid, ids.ctid
         Hash Cond: (test_6.id = ids.id)
         ->  Seq Scan on public.test_6  (cost=0.00..30.40 rows=2040 width=10)
               Output: test_6.ctid, test_6.id
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
   ->  Hash Join  (cost=67.38..125.83 rows=2040 width=12)
         Output: test_7.ctid, ids.ctid
         Hash Cond: (test_7.id = ids.id)
         ->  Seq Scan on public.test_7  (cost=0.00..30.40 rows=2040 width=10)
               Output: test_7.ctid, test_7.id
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
   ->  Hash Join  (cost=67.38..125.83 rows=2040 width=12)
         Output: test_8.ctid, ids.ctid
         Hash Cond: (test_8.id = ids.id)
         ->  Seq Scan on public.test_8  (cost=0.00..30.40 rows=2040 width=10)
               Output: test_8.ctid, test_8.id
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
   ->  Hash Join  (cost=67.38..125.83 rows=2040 width=12)
         Output: test_9.ctid, ids.ctid
         Hash Cond: (test_9.id = ids.id)
         ->  Seq Scan on public.test_9  (cost=0.00..30.40 rows=2040 width=10)
               Output: test_9.ctid, test_9.id
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
   ->  Hash Join  (cost=67.38..125.83 rows=2040 width=12)
         Output: test_10.ctid, ids.ctid
         Hash Cond: (test_10.id = ids.id)
         ->  Seq Scan on public.test_10  (cost=0.00..30.40 rows=2040 width=10)
               Output: test_10.ctid, test_10.id
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
(111 rows)


explain (verbose) delete from test using ids where test.id = ids.id and test.id < 10;
                                   QUERY PLAN
---------------------------------------------------------------------------------
 Delete on public.test_1  (cost=67.38..112.22 rows=680 width=12)
   ->  Hash Join  (cost=67.38..112.22 rows=680 width=12)
         Output: test_1.ctid, ids.ctid
         Hash Cond: (test_1.id = ids.id)
         ->  Seq Scan on public.test_1  (cost=0.00..35.50 rows=680 width=10)
               Output: test_1.ctid, test_1.id
               Filter: (test_1.id < 10)
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
(11 rows)


explain (verbose) delete from ids using test where test.id = ids.id;                        
                                     QUERY PLAN                                      
-------------------------------------------------------------------------------------
 Delete on public.ids  (cost=67.38..651.88 rows=20400 width=16)
   ->  Hash Join  (cost=67.38..651.88 rows=20400 width=16)
         Output: ids.ctid, test_1.ctid, test_1.tableoid
         Hash Cond: (test_1.id = ids.id)
         ->  Append  (cost=0.00..304.00 rows=20400 width=14)
               ->  Seq Scan on public.test_1  (cost=0.00..30.40 rows=2040 width=14)
                     Output: test_1.ctid, test_1.id, test_1.tableoid
               ->  Seq Scan on public.test_2  (cost=0.00..30.40 rows=2040 width=14)
                     Output: test_2.ctid, test_2.id, test_2.tableoid
               ->  Seq Scan on public.test_3  (cost=0.00..30.40 rows=2040 width=14)
                     Output: test_3.ctid, test_3.id, test_3.tableoid
               ->  Seq Scan on public.test_4  (cost=0.00..30.40 rows=2040 width=14)
                     Output: test_4.ctid, test_4.id, test_4.tableoid
               ->  Seq Scan on public.test_5  (cost=0.00..30.40 rows=2040 width=14)
                     Output: test_5.ctid, test_5.id, test_5.tableoid
               ->  Seq Scan on public.test_6  (cost=0.00..30.40 rows=2040 width=14)
                     Output: test_6.ctid, test_6.id, test_6.tableoid
               ->  Seq Scan on public.test_7  (cost=0.00..30.40 rows=2040 width=14)
                     Output: test_7.ctid, test_7.id, test_7.tableoid
               ->  Seq Scan on public.test_8  (cost=0.00..30.40 rows=2040 width=14)
                     Output: test_8.ctid, test_8.id, test_8.tableoid
               ->  Seq Scan on public.test_9  (cost=0.00..30.40 rows=2040 width=14)
                     Output: test_9.ctid, test_9.id, test_9.tableoid
               ->  Seq Scan on public.test_10  (cost=0.00..30.40 rows=2040 width=14)
                     Output: test_10.ctid, test_10.id, test_10.tableoid
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
(29 rows)


explain (verbose) delete from ids using test where test.id = ids.id and test.id < 10;
                                    QUERY PLAN
-----------------------------------------------------------------------------------
 Delete on public.ids  (cost=67.38..112.22 rows=680 width=16)
   ->  Hash Join  (cost=67.38..112.22 rows=680 width=16)
         Output: ids.ctid, test_1.ctid, test_1.tableoid
         Hash Cond: (test_1.id = ids.id)
         ->  Append  (cost=0.00..35.50 rows=680 width=14)
               ->  Seq Scan on public.test_1  (cost=0.00..35.50 rows=680 width=14)
                     Output: test_1.ctid, test_1.id, test_1.tableoid
                     Filter: (test_1.id < 10)
         ->  Hash  (cost=35.50..35.50 rows=2550 width=10)
               Output: ids.ctid, ids.id
               ->  Seq Scan on public.ids  (cost=0.00..35.50 rows=2550 width=10)
                     Output: ids.ctid, ids.id
(12 rows)

[funbringer]

We've just released 1.4.3.