Parameterizing a view to use with IN list queries..

Question

Brief bit of intro:
I'm writing an app that lets the user create a stored procedure of varying number of arguments. One or more of these arguments will be of the form:
1,2,3
'a','b','c'
Supplied as strings, and without using dynamic sql, it becomes difficult to use them as IN-list arguments because we essentially have the IN operator searching for a record that contains the literal text 1,2,3 rather than looking for records that match on 1 or 2 or 3
Having done some reading, it would appear that I can pull this trick:
SELECT
   TRIM(SUBSTR(src, INSTR(src, ',', 1, LEVEL)+1, INSTR(src, ',', 1, LEVEL+1) - INSTR(src, ',', 1, LEVEL)-1)) output
FROM
   t
CONNECT BY ROWNUM<=LENGTH(src) - LENGTH(replace(src,','))-1Where src is the string list
Now thats quite cumbersome, so I'd like to hide it. Tom gives us a way that I dont fully understand, but it seems to basically be of setting an Oracle environment variable of name XYZ to value 'a','b','c' - i.e. our string, and then writing a view that targets the variable.
My first question is, can someone help me out with a bit of code that does that?
Is there a problem with using the variable, because what happens if someone else calls this procedure while I am still calling it - is it possible that their call will set my env var to something else before my sql is run? How can I localise the environment variable to me/my session only?
Also could you provide an example PL/SQL of how it would be used?
I guess something like:
PROCEDURE xyz(in_list varchar2)
IS BEGIN
  set_environment_var(XYZ, in_list);
  SELECT 1 FROM table WHERE inlistcolumn IN (SELECT * FROM in_list_view)
ENDin_list_view is the view that targets env variable XYZ. Other bits are mainly pseudocode because I dont know how/where to set the environment variable.
The second question I have is:
Once I have this in place, am I correct in thinking that I will have to create two views if I wish to write a query that has 2 in-list clauses?
i.e.
PROCEDURE xyz(in_list_one varchar2, in_list_two varchar2)
IS BEGIN
  set_environment_var(XYZ, in_list_one);
  set_environment_var(ABC, in_list_two);
  SELECT 1 FROM table WHERE in_list_column IN (SELECT * FROM in_list_view_XYZ) and other_in_list_column IN (SELECT * FROM in_list_view_ABC)
ENDFor N in list ops, i must have N queries and N environment variables, yes?

Answer 1

I've never used Tom's environment variable approach, but here's a different technique you may wish to consider (assuming you're on 10g).
SQL*Plus: Release 10.2.0.1.0 - Production on Fri Jun 29 15:20:55 2007
create table t ( n number, v varchar2(10) );
Table created.
insert into t values ( 1, 'a' );
1 row created.
insert into t values ( 2, 'b' );
1 row created.
insert into t values ( 3, 'c' );
1 row created.
insert into t values ( 4, 'd' );
1 row created.
insert into t values ( 5, 'e' );
1 row created.
commit;
Commit complete.
-- for numeric in-list
variable d varchar2(1)
variable p varchar2(10)
execute :d := ',' ;
PL/SQL procedure successfully completed.
execute :p := '1,2,3' ;
PL/SQL procedure successfully completed.
select *
from t
where n in
  ( -- this query converts string '1,2,3' into
    -- three rows containing 1, 2, and 3
    select val
    from   dual
    model
      return updated rows
      dimension by ( 0 as position )
      measures     ( :p || :d as val )
      rules upsert
      ( val
        [ for position
            from 1
            to   length( regexp_replace( val[0], '[^'||:d||']', null ) )
            increment 1
        ] = rtrim( regexp_substr( val[0], '.*?' || :d, 1, cv(position) ), :d )
order by 1 ;
         N V
         1 a
         2 b
         3 c
3 rows selected.
-- for string in-list
execute :d := ',' ;
PL/SQL procedure successfully completed.
execute :p := '''a'',''b''' ;
PL/SQL procedure successfully completed.
select *
from t
where v in
  ( -- this query converts string 'a','b' into
    -- two rows containing 'a' and 'b'
    select val
    from   dual
    model
      return updated rows
      dimension by ( 0 as position )
      measures     ( replace( :p || :d, '''', null ) as val )
      rules upsert
      ( val
        [ for position
            from 1
            to   length( regexp_replace( val[0], '[^'||:d||']', null ) )
            increment 1
        ] = rtrim( regexp_substr( val[0], '.*?' || :d, 1, cv(position) ), :d )
order by 1 ;
         N V
         1 a
         2 b
2 rows selected.There are alternatives to using MODEL as well. Check out SQL Snippets: SQL Techniques Tutorials - String to Rows for some other approaches.
Joe Fuda
SQL Snippets