Submitted by: @pavel-zotov

Votes: 1

DDL:
set term ^;
recreate function fx(a_n int) returns varchar(31) deterministic as
begin
return (select rt.rdb$type_name from rdb$types rt order by rt.rdb$type rows (:a_n) to (:a_n));
end
^set term ;^

Test-1: select fx(10) from rdb$formats rows 1
Trace-1:
select fx(10) from rdb$formats rows 1
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
PLAN (RDB$FORMATS NATURAL)
1 records fetched
0 ms, 519 fetch(es)

Table Natural
unge
******************************************
****
RDB$FORMATS 1

RDB$TYPES 251

Test-2: select fx(10) from rdb$formats rows 5;
Trace-2:

Statement 30:
------------------------------------------
select fx(10) from rdb$formats rows 5
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
PLAN (RDB$FORMATS NATURAL)
5 records fetched
1 ms, 2585 fetch(es)

Table Natural
unge
******************************************
****
RDB$FORMATS 5

RDB$TYPES 1255

So, stored function re-reads RDB$TYPES table for each row of source.
It will be good for performance if all input arguments will be mapped to output results of such functions.