Oracle PLSQL : How to remove duplicate data in string - string

Question

Step 01 : I have a column A in table tab_T contains that strings :
SELECT A FROM tab_T;
((<123>+<123>+<123>)(*<213>+<213>+<213>+<354>+<354>+<354>+1)(*<985>))(+<654>+<654>+1)
(<599>*<592>*<591>)
(<10945>)
(<736>+<736>+1)
(<216>*<518>)
(<598>*<593>)(*<594>+<594>+<594>+<597>+<595>+<595>+<595>)
...
...
I want to get :
((<123>)(*<213>+<354>+1)(*<985>))(+<654>+1)
(<599>*<591>)
(<10945>)
(<736>)
(<216>*<518>)
(<598>*<593>)(*<594>+<597>+<595>)
...
...
Step 02 : Then i will replace '+' by 'AND' and '*' by 'OR' and delete the number '1' from my string
this is my query (it works good and i share it with you if you need a help)
SELECT RTRIM(RTRIM(REPLACE(REPLACE(REPLACE(REPLACE(REPLACE(REPLACE(
REPLACE(REPLACE(REPLACE(REPLACE(REPLACE(A,'+','AND'),'*','OR'),'(OR','OR('),'(AND','AND('),'(1)','')
,'OR1',''),'AND1',''),'1OR',''),'1AND',''),'ANDAND','AND'),'OROR','OR'),'AND'),'OR') AS logic
FROM tab_T
Result :
((<123>AND<123>AND<123>)OR(<213>AND<213>AND<213>AND<354>AND<354>AND<354>)OR(<985>))OR(<654>AND<654>)
(<599>OR<592>OR<591>)
(<10945>)
(<736>AND<736>)
(<216>OR<518>)
(<598>OR<593>)OR(<594>AND<594>AND<594>AND<597>AND<595>AND<595>AND<595>)
...
...
so when i apply step 01 and step 2 i will have this result
((<123>)OR(<213>AND<354>)OR(<985>))AND(<654>)
(<599>OR<591>)
(<10945>)
(<736>)
(<216>OR<518>)
(<598>OR<593>)OR(<594>AND<597>AND<595>)
...
...
I need a help or an idea for the step 01 please?
Thx

Answer 1

This will preserve the plus signs in-between the bracketed numbers:
select A original, regexp_replace(A, '(<\d+>)(\+?\1){1,}', '\1') fixed
from tab_T;
The regex can be read as: Remember a group of one or more digits inside of brackets when followed by a group of one or more of the SAME group of remembered numbers preceded by an optional plus sign. When this group is encountered, replace it with the first remembered group.
EDIT: For the sake of completeness, here's the whole thing done with successive CTE's breaking the replaces into logical groupings. This way it's a complete answer and I believe reduced the number of REPLACE() calls. You could do it as a bunch of nested REPLACE's, but I think this is arguably cleaner and easier to understand and maintain down the road.
with tab_T(A) as (
select '((<123>+<123>+<123>)(*<213>+<213>+<213>+<354>+<354>+<354>+1)(*<985>))(+<654>+<654>+1)' from dual union all
select '(<599>*<592>*<591>)' from dual union all
select '(<10945>)' from dual union all
select '(<736>+<736>+1)' from dual union all
select '(<216>*<518>)' from dual union all
select '(<598>*<593>)(*<594>+<594>+<594>+<597>+<595>+<595>+<595>)' from dual
),
-- Remove dups and '+1'
pass_1(original, fixed) as (
select A original, replace(regexp_replace(A, '(<\d+>)(\+?\1){1,}', '\1'), '+1') fixed
from tab_T
),
replace_ors(original, fixed) as (
select original, replace(replace(fixed, '(*', 'OR('), '*', 'OR')
from pass_1
),
replace_ands(original, fixed) as (
select original, replace(replace(fixed, '(+', 'AND('), '+', 'AND')
from replace_ors
)
select original, fixed
from replace_ands
;

Answer 2

I know this is not full answer for your question. But maybe it can help you:
with t as (select '((<123>+<123>+<123>)(*<213>+<213>+<213>+<354>+<354>+<354>+1)(*<985>))(+<654>+<654>+1)' as exp from dual)
, t1 as ( select distinct regexp_substr(exp, '[^+]+', 1, level) names
from t
connect by level <= length(regexp_replace(exp, '[^*+]'))+1
)
SELECT
RTrim(listagg(t1.names,'+') WITHIN GROUP (order by names desc)) string
from t1

Answer 3

I found it :)
select REGEXP_REPLACE
(A,
'(<[^>]+>)(\+|\*?\1)*',
'\1') as logic
FROM tab_T
Thank you anyway ;)