Converting an old Tom Kyte function to straight SQL

Question

I am on Oracle 10.2
There is an old tom kyte function (10+ years old) floating around that will remove constants so I can look for queries that don't use bind variables. It has to loop through each character i a sql so its slow. I was able to convert part of it to using a regular expression. There are alot of new SQL features, so I am trying to replace it with sql.
The old REMOVE_CONSTRAINTS function is at the bottom.
I have already converted the really slow part of the function which loops through every character in a query (first loop). I did it with a regular expression. See test case below.
I am having trouble with the second part that replaces numbers.
The translate is there because I was having trouble passing a single quote to a regular expression and got tired of looking at it. So I converted all single quotes to '@''
Basically the translate turns all single quotes to @
So anything surrounded by an @ and an @ gets turned into a single @
This is a bit flawed. If I have a literal that has a quote in it, this will get screwed up. It is not something I have to worry about right now though.
select mytest,regexp_replace(translate(to_char(mytest),chr(39),'@'),'@[^@]+@','@') from test
create table test (mytest clob);
insert into test values ('select 1 from mytable where x = '||chr(39)||'v'||chr(39)||' and y = '||chr(39)||'q'||chr(39));
insert into test values ('select 1 from mytabe where a = '||chr(39)||'xyz'||chr(39));
commit;
SQLt>  select mytest,regexp_replace(translate(mytest,chr(39),'@'),'@[^@]+@','@') from test
  2  /
MYTEST
REGEXP_REPLACE(TRANSLATE(MYTEST,CHR(39),'@'),'@[^@]+@','@')
select 1 from mytable where x = 'v' and y = 'q'
select 1 from mytable where x = @ and y = @
select 1 from mytabe where a = 'xyz'
select 1 from mytabe where a = @I can put this in a pipeline function and it will run fast enough. I think I can probably make this straight sql. anyone know how? also, I have some column names with numbers in it. anyone know a good way to ignore numbers on the left side of the where clause?
    l_query := translate( l_query, '0123456789', '@@@@@@@@@@' );
--   dbms_output.put_line(l_query);  
    for i in 0 .. 8 loop
        l_query := replace( l_query, lpad('@',10-i,'@'), '@' );
        l_query := replace( l_query, lpad(' ',10-i,' '), ' ' );
    end loop;
  function remove_constraints ( p_query in clob ) return clob AS
    l_query clob;
    l_char  varchar2(1);
    l_in_quotes boolean default FALSE;
    l_length number := 0;
    vmodule varchar2(30);
    vaction varchar2(30);
begin
    dbms_application_info.read_module(vaction,vmodule);
    dbms_application_info.set_module(vaction,nvl(vmodule,0)+1);
    for i in 1 .. length( p_query )
    loop
        l_char := substr(p_query,i,1);
        if ( l_char = '''' and l_in_quotes )
        then
            l_in_quotes := FALSE;
        elsif ( l_char = '''' and NOT l_in_quotes )
        then
            l_in_quotes := TRUE;
            l_query := l_query || '''#';
            dbms_output.put_line('a '||l_char);
            dbms_output.put_line( '''#');
        end if;
        if ( NOT l_in_quotes ) then
            l_query := l_query || l_char;
        end if;
    end loop;
--dbms_output.put_line(l_query); 
    l_query := translate( l_query, '0123456789', '@@@@@@@@@@' );
--   dbms_output.put_line(l_query);  
    for i in 0 .. 8 loop
        l_query := replace( l_query, lpad('@',10-i,'@'), '@' );
        l_query := replace( l_query, lpad(' ',10-i,' '), ' ' );
    end loop;
    return upper(l_query);
end;
   

Answer 1

In 10g you can use
FORCE_MATCHING_SIGNATURE     The signature used when the CURSOR_SHARING parameter is set to FORCE
http://docs.oracle.com/cd/B19306_01/server.102/b14237/dynviews_2113.htm#REFRN30246