How to split a string value based on a delimiter in DB2

How do you split a string value in DB2?

For example, given the value:

CHG-FFH.

I want to split on the dash (-), which would result in two values:

CHG 
FFH. 

I tried using split function but it is not a function in DB2.

Any help will be appreciated.


Solution 1:

Short answer:

You need to find the position of the delimiter, and then substring using it as the starting point, and a calculated length.

SELECT 
    SUBSTR('CHG-FFH', 1, LOCATE('-','CHG-FFH')-1) as FIRST_PART
  , SUBSTR('CHG-FFH', LOCATE('-','CHG-FFH')+1)   as SECOND_PART
FROM SYSIBM.SYSDUMMY1;

BONUS! If you do this often, create a user defined function to do it dynamically. Here's an example in a DB fiddle.

CREATE FUNCTION SPLITTER (input VARCHAR(4000), delimiter CHAR, part_number INTEGER)
      RETURNS VARCHAR(4000)
      LANGUAGE SQL
      READS SQL DATA
      NO EXTERNAL ACTION
      DETERMINISTIC
      RETURN 

with pos_info (first_pos, length) as (
select 
  case when part_number = 1 then 0
       else LOCATE_IN_STRING(input, delimiter,1, part_number-1, OCTETS)  
  end as first_pos, 

  case when part_number = 1 then LOCATE_IN_STRING(input, delimiter,1, part_number, OCTETS) - 1 
       when LOCATE_IN_STRING(input, delimiter,1, part_number, OCTETS) = 0
        and  LOCATE_IN_STRING(input, delimiter,1, part_number-1, OCTETS) = 0
       then 0
       when LOCATE_IN_STRING(input, delimiter,1, part_number, OCTETS) = 0
       then length(input) -  LOCATE_IN_STRING(input, '-',1, part_number - 1,    OCTETS)
       else LOCATE_IN_STRING(input, delimiter,1, part_number, OCTETS) -    LOCATE_IN_STRING(input, delimiter,1, part_number-1, OCTETS) - 1
  end as length
from sysibm.sysdummy1
)

select    
    substr(input, first_pos+1,length) as part
from pos_info;

Alternatively, you can see a different approach here at this answer: Split a VARCHAR in DB2 to retrieve a value inside.


Long answer:

DB2, along with other relational databases do not provide a single function to accomplish this.

The reason is likely that it's not an implicitly scalar function. If your string had more than one dash in it, would you want to split it into three parts? Four? So the first step is to note if your data is determinate - if it has a specific number of components that you want to split apart. In your example, you have two, so I'll start with that assumption, and then afterwards comment on how you would deal other situations.

Scenario: A string value with two components separated by a delimiter

With only two parts, you need to find the position of the delimiter, and then substring before and after it by using the position before and after it in a substring function.

  1. LOCATE the index of your delimiter.
LOCATE('-','CHG-FFH')

NOTE: DB2 provides two functions which can be used for this: POSITION (or POSSTR), and LOCATE (or LOCATE_IN_STRING). LOCATE is a bit more powerful because it allows you to specify a start position, which would be helpful if you had more than one delimiter.

  1. SUBSTR using the delimiter index.

For the first part, start your substring at position 1, up to the character before the delimiter (delimiter position - 1):

SUBSTR('CHG-FFH', 1,LOCATE('-','CHG-FFH')-1) as FIRST_PART

For the second part, start your substring at the position after delimiter index (delimiter position + 1), and get the rest of the String:

 SUBSTR('CHG-FFH', LOCATE('-','CHG-FFH')+1) as SECOND_PART

Final Result:

SELECT 
    SUBSTR('CHG-FFH', 1,LOCATE('-','CHG-FFH')-1) as FIRST_PART
  , SUBSTR('CHG-FFH', LOCATE('-','CHG-FFH')+1) as SECOND_PART
FROM SYSIBM.SYSDUMMY1;

Scenario: A string value with three components separated by a delimiter

Use the same concepts as the first scenario, but you have to determine the index of the second delimiter. Use the index of the first delimiter to specify a starting point: Note that LOCATE allows specifying a start position:

>>-LOCATE(search-string,source-string-+--------+-+--------------------+-)-><
                                      '-,start-' '-,--+-CODEUNITS16-+-'     
                                                      +-CODEUNITS32-+       
                                                      '-OCTETS------' 

Finding the second delimiter:

Use the position of the first delimiter as the starting point for finding the second delimiter.

LOCATE('-','CHG-FFH-EEE', LOCATE('-','CHG-FFH-EEE')+1)

Use that as a SUBSTR point for the second and third values, and you're all set. Note: For the second value, you have to use both of the delimiter locations to substring the value.

Final Result:

SELECT 
    SUBSTR('CHG-FFH-EEE', 1,LOCATE('-','CHG-FFH-EEE')-1) as FIRST_PART
  , SUBSTR('CHG-FFH-EEE', LOCATE('-','CHG-FFH-EEE')+1, LOCATE('-','CHG-FFH-EEE', LOCATE('-','CHG-FFH-EEE'))-1) as SECOND_PART
  , SUBSTR('CHG-FFH-EEE', LOCATE('-','CHG-FFH-EEE', LOCATE('-','CHG-FFH-EEE')+1)+1) as THIRD_PART
FROM SYSIBM.SYSDUMMY1;

You can see this strategy would get out of hand for a greater number of delimiters in your String.

Scenario: Indeterminate number of delimiters

This is a tricky problem that is best approached with a Stored Procedure. Think through things like: How do you want the parsed data to come out of the algorithm, how will you access the data? Arrays are not a native type in SQL, but they are in Stored Procedures, so what will you do with the array when you've parsed all the pieces out of your String?

One way to approach this scenario is answered here:

Split a VARCHAR in DB2 to retrieve a value inside

Solution 2:

This is what i tried and it fetched me effective result. Hence sharing with all.

select column_name, substr(column_name,1,locate('-',column_name)-1), 
substr(column_name,locate('-',column_name)+1,
length(substr(column_name,locate('-',column_name)+1))) from 
table_name where column_name is not null and column_name!='' 
and column_name like '%-%'

Solution 3:

The reason of so late answer is to show much more simple and universal way to achieve the goal. It's based on functions using regular expressions available even at the time, when the question was asked.

To get N-th token (2-nd in the example):

SELECT 
  COL
-- since 9.7 (2009)
, xmlcast(xmlquery('fn:tokenize($s, "-")[2]' passing COL as "s") as varchar(20)) as one
-- since 11.1
, REGEXP_SUBSTR(COL, '([^-]*)-?', 1, 2, '', 1) as two
FROM (VALUES 'CHG-FFH.', 'ABC-DEF-GH') TAB (COL);

The result is:

|COL       |ONE                 |TWO       |
|----------|--------------------|----------|
|CHG-FFH.  |FFH.                |FFH.      |
|ABC-DEF-GH|DEF                 |DEF       |

To tokenize a string:

-- since 9.7 (2009)
SELECT TAB.COL, TOK.SEQ, TOK.TOKEN
FROM 
  (VALUES 'CHG-FFH.', 'ABC-DEF-GH') TAB (COL)
, XMLTABLE
(
  'for $id in tokenize($s, "-") return <i>{string($id)}</i>' PASSING TAB.COL AS "s"
  COLUMNS
    SEQ   FOR ORDINALITY
  , TOKEN VARCHAR(20) PATH '.'
) TOK
ORDER BY TAB.COL, TOK.SEQ;

The result is:

|COL       |SEQ                 |TOKEN               |
|----------|--------------------|--------------------|
|ABC-DEF-GH|1                   |ABC                 |
|ABC-DEF-GH|2                   |DEF                 |
|ABC-DEF-GH|3                   |GH                  |
|CHG-FFH.  |1                   |CHG                 |
|CHG-FFH.  |2                   |FFH.                |