Oracle regex count multiple occurrences of a string surrounded by commas - sql

Question

This question is similar to a previous question of mine. I am looking for a way to count a character string in a comma-separated list of values in a column in an Oracle (11g) SQL database. For example, suppose I have the following data:
SELECT ('SL,PK') as col1 FROM dual
UNION ALL
SELECT ('SL,CR,SL') as col1 FROM dual
UNION ALL
SELECT ('PK,SL') as col1 FROM dual
UNION ALL
SELECT ('SL,SL') as col1 FROM dual
UNION ALL
SELECT ('SL') as col1 FROM dual
UNION ALL
SELECT ('PK') as col1 FROM dual
UNION ALL
SELECT ('PI,SL,PK') as col1 FROM dual
UNION ALL
SELECT ('PI,SL,SL,PK') as col1 FROM dual
UNION ALL
SELECT ('PI,SL,SL,SL,PK') as col1 FROM dual
UNION ALL
SELECT ('PI,SL,SL,SL,SL,PK') as col1 FROM dual
UNION ALL
SELECT ('PI,OSL,SL,PK') as col1 FROM dual
UNION ALL
SELECT ('PI,SL,SLR,PK') as col1 FROM dual
COL1
-----
SL,PK
SL,CR,SL
PK,SL
SL,SL
SL
PK
PI,SL,PK
PI,SL,SL,PK
PI,SL,SL,SL,PK
PI,SL,SL,SL,SL,PK
PI,OSL,SL,PK
PI,SL,SLR,PK
I am looking to count all occurrences of the substring 'SL', strictly (i.e. not including 'OSL', 'SLR', etc). The ideal result would look like this:
COL1 COL2
----- -----
SL,PK 1
SL,CR,SL 2
PK,SL 1
SL,SL 2
SL 1
PK 0
PI,SL,PK 1
PI,SL,SL,PK 2
PI,SL,SL,SL,PK 3
PI,SL,SL,SL,SL,PK 4
PI,OSL,SL,PK 1
PI,SL,SLR,PK 1
I can accomplish this using length and regexp_replace:
SELECT
col1,
(length(col1) - NVL(length(regexp_replace(regexp_replace(col1,'(^|,)(SL)($|,)','\1' || '' || '\3',1,0,'imn'),'(^|,)(SL)($|,)','\1' || '' || '\3',1,0,'imn')),0))/length('SL') as col2
FROM (
SELECT ('SL,PK') as col1 FROM dual
UNION ALL
SELECT ('SL,CR,SL') as col1 FROM dual
UNION ALL
SELECT ('PK,SL') as col1 FROM dual
UNION ALL
SELECT ('SL,SL') as col1 FROM dual
UNION ALL
SELECT ('SL') as col1 FROM dual
UNION ALL
SELECT ('PK') as col1 FROM dual
UNION ALL
SELECT ('PI,SL,PK') as col1 FROM dual
UNION ALL
SELECT ('PI,SL,SL,PK') as col1 FROM dual
UNION ALL
SELECT ('PI,SL,SL,SL,PK') as col1 FROM dual
UNION ALL
SELECT ('PI,SL,SL,SL,SL,PK') as col1 FROM dual
UNION ALL
SELECT ('PI,OSL,SL,PK') as col1 FROM dual
UNION ALL
SELECT ('PI,SL,SLR,PK') as col1 FROM dual
)
COL1 COL2
----- -----
SL,PK 1
SL,CR,SL 2
PK,SL 1
SL,SL 2
SL 1
PK 0
PI,SL,PK 1
PI,SL,SL,PK 2
PI,SL,SL,SL,PK 3
PI,SL,SL,SL,SL,PK 4
PI,OSL,SL,PK 1
PI,SL,SLR,PK 1
but was hoping for a more elegant solution, perhaps with regexp_count. I have achieved my goal successfully in other regex implementations that have the word boundary \b construct available (with \bSL\b), but have not found a solution for Oracle's regex.

Answer 1

You can use regexp_count() if you hack the string:
select col1, regexp_count(replace(col1, ',', ',,'), '(^|\W)SL(\W|$)')
This doubles the delimiter so the first match doesn't eat it up -- getting around the underlying issue which is that Oracle regular expressions do not support look-ahead.
Here is a db<>fiddle.

Answer 2

Here's one option:
SQL> with temp as
2 (select col1,
3 regexp_substr(col1, '[^,]+', 1, column_value) val
4 from test cross join
5 table(cast(multiset(select level from dual
6 connect by level <= regexp_count(col1, ',') + 1
7 ) as sys.odcinumberlist))
8 )
9 select col1,
10 sum(case when val = 'SL' then 1 else 0 end) col2
11 From temp
12 group by col1;
COL1 COL2
----------------- ----------
PI,SL,SLR,PK 1
PK,SL 1
PK 0
SL,CR,SL 2
PI,OSL,SL,PK 1
SL,SL 2
PI,SL,SL,PK 2
PI,SL,SL,SL,PK 3
SL,PK 1
SL 1
PI,SL,PK 1
PI,SL,SL,SL,SL,PK 4
12 rows selected.
SQL>
What does it do?
temp CTE splits each column into rows (separator is comma)
the final select simply counts number of SLs for each col1

Answer 3

You can use an XMLTABLE to spilt the string and then count:
SELECT col1,
(
SELECT COUNT(*)
FROM XMLTABLE(
('"' || REPLACE( col1, ',', '","' ) || '"')
COLUMNS
value CHAR(2) PATH '.'
)
WHERE value = 'SL'
) AS col2
FROM test_data
So, for your test data:
CREATE TABLE test_data ( col1 ) AS
SELECT 'SL,PK' FROM dual UNION ALL
SELECT 'SL,CR,SL' FROM dual UNION ALL
SELECT 'PK,SL' FROM dual UNION ALL
SELECT 'SL,SL' FROM dual UNION ALL
SELECT 'SL' FROM dual UNION ALL
SELECT 'PK' FROM dual UNION ALL
SELECT 'PI,SL,PK' FROM dual UNION ALL
SELECT 'PI,SL,SL,PK' FROM dual UNION ALL
SELECT 'PI,SL,SL,SL,PK' FROM dual UNION ALL
SELECT 'PI,SL,SL,SL,SL,PK' FROM dual UNION ALL
SELECT 'PI,OSL,SL,PK' FROM dual UNION ALL
SELECT 'PI,SL,SLR,PK' FROM dual
This outputs:
COL1 | COL2
:---------------- | ---:
SL,PK | 1
SL,CR,SL | 2
PK,SL | 1
SL,SL | 2
SL | 1
PK | 0
PI,SL,PK | 1
PI,SL,SL,PK | 2
PI,SL,SL,SL,PK | 3
PI,SL,SL,SL,SL,PK | 4
PI,OSL,SL,PK | 1
PI,SL,SLR,PK | 2
db<>fiddle here