Обсуждение: increasing FUNC_MAX_ARGS insufficient?

I followed Neil Conway's helpful instructions on changing FUNC_MAX_ARGS:

1. change FUNC_MAX_ARGS in src/include/pg_config.h
2. make clean
3. recompile
4. reinstall
5. initdb

Having done so, I should now be able to create functions with up to 64
arguments. I created the following functions to test it:

create function sumit (int, int, int, int, int, int, int, int, int, int, int, int, int, int, int, int, int, int, int,
int)returns int as ' 
        declare runsum int;
        begin
        runsum := $1 + $2 + $3 + $4 + $5 + $6 + $7 + $8 + $9 + $10
                + $11 + $12 + $13 + $14 + $15 + $16 + $17 + $18 + $19 + $20;
        return runsum;
        end;
' language 'plpgsql';

create function sumit (int, int, int, int, int, int, int, int, int, int) returns int as '
        declare runsum int;
        begin
        runsum := $1 + $2 + $3 + $4 + $5 + $6 + $7 + $8 + $9 + $10;
        return runsum;
        end;
' language 'plpgsql';

To save you counting, the first sumit takes 20 integers, which is more than
unmodified PostgreSQL can handle, and the second takes 10. There are several
strange things about them. First off, psql gives strange results with \df:

template1=> \df sumit

                                                                          List of functions 
 Result data type | Name  |
                                                                                      Argument data types

                                             

------------------+-------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 integer          | sumit | integer, integer, integer, integer, integer, integer, integer, integer, integer, integer,
-,-, -, -, -, -, ???, -, -, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???,
???,???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, -, -, ???, ???, oidvector, ???, ???, ???, ??? 
 integer          | sumit | integer, integer, integer, integer, integer, integer, integer, integer, integer, integer,
integer,integer, integer, integer, integer, integer, ???, -, -, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???,
???,???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???, ???,
???,???, ???, ???, ???, ???, ???, ???, ???, ??? 
(2 rows)

Second, while the 10 argument form works perfectly, the 20 argument form
gives the following:

template1=> select sumit(1,2,3,4,5,6,7,8,9,0,1,2,3,4,5,6,7,8,9,0);
NOTICE:  plpgsql: ERROR during compile of sumit near line 0
ERROR:  cache lookup for argument type 100 failed

Any thoughts on this? I am sure I am using the modified PostgreSQL
installation for everything. This isn't an issue of incompatible binaries.

--Greg

Gregory Seidman <gss+pg@cs.brown.edu> writes:
> I followed Neil Conway's helpful instructions on changing FUNC_MAX_ARGS:
> 1. change FUNC_MAX_ARGS in src/include/pg_config.h

Er ... I trust what you changed was actually the INDEX_MAX_KEYS line?

/*
 * Maximum number of columns in an index and maximum number of arguments
 * to a function. They must be the same value.
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^
 */
#define INDEX_MAX_KEYS        16
#define FUNC_MAX_ARGS        INDEX_MAX_KEYS


Other than that, your procedure sounds dead-on ...

            regards, tom lane

Tom Lane sez:
} Gregory Seidman <gss+pg@cs.brown.edu> writes:
} > I followed Neil Conway's helpful instructions on changing FUNC_MAX_ARGS:
} > 1. change FUNC_MAX_ARGS in src/include/pg_config.h
}
} Er ... I trust what you changed was actually the INDEX_MAX_KEYS line?
}
} /*
}  * Maximum number of columns in an index and maximum number of arguments
}  * to a function. They must be the same value.
}                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^
}  */
} #define INDEX_MAX_KEYS        16
} #define FUNC_MAX_ARGS        INDEX_MAX_KEYS

Framitz! Nope, I'm an idiot. When there are comments in the code, I really
need to read them. This would be the second time in as many days that I've
asked a question answered by that very block of comments. Thanks for your
help! It all works properly now.

} Other than that, your procedure sounds dead-on ...
}             regards, tom lane
--Greg