Oracle Regex - Remove duplicates including chinese - sql

Question

I'm trying to remove the duplicates in the results of a query involving listagg.
I'm using this syntax:
REGEXP_REPLACE (LISTAGG (PR.NAME, ',' ) WITHIN GROUP (ORDER BY 1),
'([^,]+)(,\1)+',
'\1') AS PRODUCERS
However, occurrences including chinese characters are not removed:
Any idea ?

Answer 1

Your regular expression does not work. If the LISTAGG output is A,A,AA then the regular expression ([^,]+)(,\1)+ does not check that it has matched a complete element of your list and will match A,A,A which is 2½ elements of the list and will give the output AA instead of the expected A,AA. Worse, if you have the string BA,BABAB,BABD then the regular expression will replace BA,BA with BA and then replace BAB,BAB with BAB and you end up with the string BABABD which does not match any of the elements of the original list.
An example demonstrating this is:
SQL Fiddle
Oracle 11g R2 Schema Setup:
CREATE TABLE names ( id, name ) AS
SELECT 1, 'A' FROM DUAL UNION ALL
SELECT 2, 'A' FROM DUAL UNION ALL
SELECT 3, 'B' FROM DUAL UNION ALL
SELECT 4, 'C' FROM DUAL UNION ALL
SELECT 5, 'A' FROM DUAL UNION ALL
SELECT 6, 'AA' FROM DUAL UNION ALL
SELECT 7, 'A' FROM DUAL UNION ALL
SELECT 8, 'BA' FROM DUAL UNION ALL
SELECT 9, 'A' FROM DUAL
/
Query 1:
SELECT REGEXP_REPLACE (
LISTAGG (NAME, ',' ) WITHIN GROUP (ORDER BY 1),
'([^,]+)(,\1)+',
'\1'
) AS constant_sort
FROM names
Results:
| CONSTANT_SORT |
|---------------|
| AA,BA,C |
If you want to get the distinct elements then you can use DISTINCT (as per Littlefoot's answer) or you can COLLECT the values into a user-defined collection and then use the SET function to remove duplicates. You can then pass this de-duplicated collection to a table collection expression and use LISTAGG to get your output:
Oracle 11g R2 Schema Setup:
CREATE TYPE StringList IS TABLE OF VARCHAR2(4000)
/
Query 2:
SELECT (
SELECT LISTAGG( column_value, ',' )
WITHIN GROUP ( ORDER BY ROWNUM )
FROM TABLE( n.unique_names )
) AS agg_names
FROM (
SELECT SET( CAST( COLLECT( name ORDER BY NAME ) AS StringList ) )
AS unique_names
FROM names
) n
Results:
| AGG_NAMES |
|-------------|
| A,AA,B,BA,C |
Regarding your comment:
in the context of a bigger query involving a lot of join and given my begginers skills I would have no idea how to implement this model
For example, if your query was:
SELECT REGEXP_REPLACE(
LISTAGG (PR.NAME, ',' ) WITHIN GROUP (ORDER BY 1),
'([^,]+)(,\1)+',
'\1'
) AS PRODUCERS,
other_column1,
other_column2
FROM table1 pr
INNER JOIN table2 t2
ON (pr.some_condition = t2.some_condition )
WHERE t2.some_other_condition = 'TRUE'
GROUP BY other_column1, other_column2
Then you can change it to:
SELECT (
SELECT LISTAGG( COLUMN_VALUE, ',' ) WITHIN GROUP ( ORDER BY ROWNUM )
FROM TABLE( t.PRODUCERS )
) AS producers,
other_column1,
other_column2
FROM (
SELECT SET( CAST( COLLECT( PR.name ORDER BY PR.NAME ) AS StringList ) )
AS PRODUCERS,
other_column1,
other_column2
FROM table1 pr
INNER JOIN table2 t2
ON (pr.some_condition = t2.some_condition )
WHERE t2.some_other_condition = 'TRUE'
GROUP BY other_column1, other_column2
) t

Answer 2

(I can't see images; company policy).
Why wouldn't you remove duplicates before applying LISTAGG? Something like
select listagg(x.distinct_name, ',') within group (order by 1) producers
from (select DISTINCT name distinct_name
from some_table
) x

Answer 3

Another way to remove duplicates is to use window functions and case:
select listagg(case when seqnum = 1 then name end, ',') within group (order by 1) as producers
from (select . . .,
row_number() over (partition by name order by name) as seqnum
from . . .
) t
This does require modifications to the rest of the query, but you should still be able to do the rest of the aggregations and computations.