Thread: [BUGS] ORDER BY $1 behaves inconsistently
Version: 10.0
jordan=# SELECT * FROM t ORDER BY 'foo';
ERROR: non-integer constant in ORDER BY
LINE 1: SELECT * FROM t ORDER BY 'foo';
However, using a prepared statement, this behavior can be avoided:
jordan=# PREPARE x as SELECT * FROM t ORDER BY $1;
PREPARE
jordan=# EXECUTE x('foo');
c
---
1
It seems to me that there is some missing type checking from ORDER BY.
Actually, it's even worse than I thought. It seems that placeholders in ORDER BY clauses get entirely ignored,
as running `EXECUTE x(1)` on the previous example with unsorted table data does not sort the table by the 1st
column as expected.
On Fri, Oct 27, 2017 at 12:18 PM Jordan Lewis <jordanthelewis@gmail.com> wrote:
Version: 10.0As I understand it, the only valid constant datatype in an ORDER BY is integer. That's validated by the following test:jordan=# SELECT * FROM t ORDER BY 'foo';ERROR: non-integer constant in ORDER BYLINE 1: SELECT * FROM t ORDER BY 'foo';However, using a prepared statement, this behavior can be avoided:jordan=# PREPARE x as SELECT * FROM t ORDER BY $1;PREPAREjordan=# EXECUTE x('foo');c---1It seems to me that there is some missing type checking from ORDER BY.
Jordan Lewis <jordanthelewis@gmail.com> writes: > Actually, it's even worse than I thought. It seems that placeholders in > ORDER BY clauses get entirely ignored, > as running `EXECUTE x(1)` on the previous example with unsorted table data > does not sort the table by the 1st > column as expected. "ORDER BY $1" is asking to order by some constant value (constant within any one execution of the command, anyway), which is useless, because all rows in the query will have the same sort key. The planner will throw that away as being a no-op. "ORDER BY 1" is asking to order by the first output column. This is not the same thing. You cannot get that effect with a parameter; if you could, it would probably represent a SQL-injection hazard. If the value of $1 were "foo" and you complained that it didn't order by column foo, it would be exactly the same issue... regards, tom lane -- Sent via pgsql-bugs mailing list (pgsql-bugs@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-bugs
Thanks for the explanation - that makes sense.
On Fri, Oct 27, 2017 at 2:33 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
Jordan Lewis <jordanthelewis@gmail.com> writes:
> Actually, it's even worse than I thought. It seems that placeholders in
> ORDER BY clauses get entirely ignored,
> as running `EXECUTE x(1)` on the previous example with unsorted table data
> does not sort the table by the 1st
> column as expected.
"ORDER BY $1" is asking to order by some constant value (constant within
any one execution of the command, anyway), which is useless, because all
rows in the query will have the same sort key. The planner will throw
that away as being a no-op.
"ORDER BY 1" is asking to order by the first output column. This is
not the same thing. You cannot get that effect with a parameter;
if you could, it would probably represent a SQL-injection hazard.
If the value of $1 were "foo" and you complained that it didn't order
by column foo, it would be exactly the same issue...
regards, tom lane