Executing function embedded SELECT sql query to increase performance - oracle

Question

I have the following SQL query that will return the name of a column in a specific table. Let's say it return 'USER_PK' as column name when it runs.
the query:
SELECT max(COLUMN_NAME)
FROM ALL_TAB_COLUMNS
WHERE OWNER= 'DW_01'
AND table_name='D_O_USERS'
AND COLUMN_NAME<>'USER_PK';
Now I would like to run the above query as part of a function but instead of running it and storing the value it returns in a variable (using INTO or attribution like initial_sql: = '...', followed by exec ) I would need to have it run inside one line of code as below (see part in bold)... So far I have been unsuccessful as it is interpreted as a string when using quotes ...
CREATE OR REPLACE function DW_01.EXECUTE_AUTO (db_schema IN VARCHAR2, db_table IN VARCHAR2, pk_name IN VARCHAR2, id_pk IN INTEGER) RETURN VARCHAR2
IS
result VARCHAR2(4000);
begin
EXECUTE IMMEDIATE 'select STANDARD_HASH( '|| **SELECT max( COLUMN_NAME) FROM ALL_TAB_COLUMNS WHERE OWNER='' || db_schema || '' AND table_name=''||db_table ||'' AND COLUMN_NAME<>'' ||pk_name ||'** ,''SHA512'' ) from '||db_table||' where '|| pk_name ||'='||id_pk into RESULT ;
return result;
end;
Many thanks in advance for your thoughts!

Answer 1

You need to amend you r code like below -
CREATE OR REPLACE function DW_01.EXECUTE_AUTO (db_schema IN VARCHAR2,
db_table IN VARCHAR2,
pk_name IN VARCHAR2,
id_pk IN INTEGER) RETURN VARCHAR2
IS
result VARCHAR2(4000);
begin
EXECUTE IMMEDIATE 'select STANDARD_HASH( ' || pk_name || ',256 )
from '||db_table||' where '|| pk_name ||'='||id_pk into RESULT;
return result;
end;
/

Answer 2

There are only a few ways to run dynamic SQL in SQL, and they're neither pretty nor fast. The function below uses DBMS_XMLGEN.GETXML to dynamically run a SQL statement.
create or replace function execute_auto(db_schema in varchar2, db_table in varchar2, pk_name in varchar2, id_pk in integer) return varchar2
is
v_column_name varchar2(128);
v_result varchar2(4000);
begin
select standard_hash(to_number(extractValue(xml_results, '/ROWSET/ROW/' || max_column)), 'SHA512') hash_value
into v_result
from
(
--Create a single XML file with the ROWIDs that match the condition.
select max(column_name) max_column, xmltype(dbms_xmlgen.getxml
('
select '||max(column_name)||'
from '||db_schema||'.'||db_table||'
where id = '||id_pk
)) xml_results
from all_tab_columns
where owner = db_schema
and table_name = db_table
and column_name <> pk_name
);
return v_result;
end;
/
For example, let's create this sample table with 100,000 rows:
--drop table test1;
create table test1(id number, a number, b number, constraint pk_test1 primary key(id));
insert into test1
select level, level, level from dual connect by level <= 100000;
commit;
This shows how to use the function;
select execute_auto(user, 'TEST1', 'ID', id) hash
from test1 where id = 1;
HASH
----
A36753F534728ED84A463ECB13750B8E920A7E4D90244258DE77D9800A0F3DAF8CBAD49602E960A2355933C689A23C30377CE10FC4B8E1F197739FF86C791022
In addition to problems with type conversion and SQL injection, the performance is terrible. Selecting all 100,000 rows this way takes 200 seconds on my machine.
select sum(length(execute_auto(user, 'TEST1', 'ID', id))) from test1;
Normally, running everything in a single SELECT statement is a good way to improve performance. But this extreme type of dynamic SQL will never run fast. You probably want to rethink your approach. Instead of trying to optimize the SQL inside a function that is run one-row-at-a-time, try to change the process to process once-per-table