Remove coma separated string from another coma separated string in oracle - oracle

Question

Column1 =A,B,C,D,E,F
Column2 =C,D,A,F,C,B (It can have duplicates)
I need to remove column2 values from column1 and get the missing value.
Desired output
(Column1)-(Column2) = E

Answer 1

Split columns' contents into rows, use MINUS set operator. Sample data in lines #1 - 3; query begins at line #4.
SQL> with test (col1, col2) as
2 (select 'A,B,C,D,E,F', 'C,D,A,F,C,B' from dual
3 )
4 select regexp_substr(col1, '[^,]+', 1, level) val
5 from test
6 connect by level <= regexp_count(col1, ',') + 1
7 minus
8 select regexp_substr(col2, '[^,]+', 1, level) val
9 from test
10 connect by level <= regexp_count(col2, ',') + 1
11 /
VAL
--------------------------------------------
E
SQL>
If you're comparing columns in a multi-row table, the above approach won't work OK as it'll retrieve duplicates and will be slow. In that case, rewrite it to
SQL> with test (id, col1, col2) as
2 (select 1, 'A,B,C,D,E,F', 'C,D,A,F,C,B' from dual union all
3 select 2, 'A,B,C,D,E,F', 'A,B,B,B' from dual
4 )
5 select id, listagg(val, ',') within group (order by val) missing_letters
6 from
7 (
8 select id,
9 regexp_substr(col1, '[^,]+', 1, column_value) val
10 from test cross join
11 table(cast(multiset(select level from dual
12 connect by level <= regexp_count(col1, ',') + 1
13 ) as sys.odcinumberlist))
14 minus
15 select id,
16 regexp_substr(col2, '[^,]+', 1, column_value) val
17 from test cross join
18 table(cast(multiset(select level from dual
19 connect by level <= regexp_count(col2, ',') + 1
20 ) as sys.odcinumberlist))
21 )
22 group by id;
ID MISSING_LETTERS
---------- --------------------
1 E
2 C,D,E,F
SQL>

Answer 2

You may use translate function with additional cleanup logic to remove all remaining commas. This will work only for single character replacement (one character between commas), but doesn't require to split string into tokens and uses simple string functions.
with a(col1, col2) as (
select 'A,B,C,D,E,F', 'C,D,A,F,C,B' from dual
)
select
/*Then remove leading and trailing commas*/
trim(',' from
/*Then condense all intermediate commas and spaces*/
regexp_replace(
/*Do actual replacement*/
translate(col1, replace(col2, ','), ' '),
'[, ]+', ','
)
) as res
from a
| RES |
| :-- |
| E |
db<>fiddle here

Answer 3

You do not need to split the string.
If your delimited values do not have any characters with special meaning in regular expressions then you can double-up the delimiters in col1 and then convert col2 to a regular expression and replace matches with an empty string and then remove the excess delimiters:
SELECT col1,
col2,
TRIM(
BOTH ',' FROM
REPLACE(
REGEXP_REPLACE(
',' || REPLACE(col1, ',', ',,') || ',',
',(' || REPLACE(col2, ',', '|') || '),'
),
',,',
','
)
) AS missing
FROM table_name;
Which, for the sample data:
CREATE TABLE table_name ( col1, col2 ) AS
SELECT 'A,B,C,D,E,F', 'C,D,A,F,C,B' FROM DUAL UNION ALL
SELECT 'A,AB,BA,B,', 'A,B' FROM DUAL;
Outputs:
COL1
COL2
MISSING
A,B,C,D,E,F
C,D,A,F,C,B
E
A,AB,BA,B,
A,B
AB,BA
If you do have characters with special meaning then you can do a similar replacement using a recursive sub-query:
WITH replacements ( col1, col2 ) AS (
SELECT ',' || REPLACE( col1, ',', ',,') || ',',
col2 || ','
FROM table_name
UNION ALL
SELECT REPLACE(col1, ',' || SUBSTR(col2, 1, INSTR(col2, ','))),
SUBSTR(col2, INSTR(col2, ',') + 1)
FROM replacements
WHERE col2 IS NOT NULL
)
SELECT TRIM(BOTH ',' FROM REPLACE(col1, ',,', ',')) AS missing
FROM replacements
WHERE col2 IS NULL
Which outputs:
MISSING
AB,BA
E
Note: both of these queries only require a single table scan.
db<>fiddle here

Answer 4

Using ora:tokenize you could do something like this (including a few test cases in the with clause; you should remove it, and use your actual table and column names in the main query):
with
inputs (col1, col2) as (
select 'A,B,C,D,E,F', 'C,D,A,F,C,B' from dual union all
select 'D,,F' , 'F,A' from dual union all
select 'A,B,E,F' , 'E' from dual union all
select 'ABC' , 'A,B,ABC' from dual
)
-- END OF TEST DATA; QUERY BEGINS **BELOW THIS LINE**
select i.col1, i.col2, l.diff
from inputs i cross join lateral
( select listagg(token, ',') within group (order by null) as diff
from xmltable('ora:tokenize(.,",")' passing i.col1 || ','
columns token varchar2(10) path '.')
where not ',' || col2 || ',' like '%,' || token || ',%' ) l
;
COL1 COL2 DIFF
----------- ----------- --------------------
A,B,C,D,E,F C,D,A,F,C,B E
D,,F F,A D
A,B,E,F E A,B,F
ABC A,B,ABC