Joining tables containing comma delimited values - sql

Question

I have three excel sheet I push them into tables in SQL server and I need to join these table. However, I believe - as I have tried already - normal join wouldn't work. I have programming background but not that much with SQL.
Table1
ID Data_column reference_number
1 some data 1528,ss-456
2 some data 9523
3 some data ss-952
4 some data null
Table2
ID Data_column
ss-456 some data
ss-952 some data
Table3
ID Data_column
1528 some data
9523 some data
In the case below How I will be able to join this raw on both table.
Table1
ID Data_column reference_number
1 some data 1528,ss-456

Answer 1

declare #t1 as table(
id int
,data_column varchar(20)
,reference_number varchar(20)
)
declare #t2 as table(
id varchar(20)
,data_column varchar(20)
)
declare #t3 as table(
id varchar(20)
,data_column varchar(20)
)
insert into #t1 values(1,'some data','1528,ss-456'),(2,'some data','9523'),(3,'some data','ss-952'),(4,'some data',null);
insert into #t2 values('ss-456','some data'),('ss-952','some data');
insert into #t3 values(1528,'some data'),(9523,'some data');
Quick solution
select * from #t1 t1
left outer join #t2 t2 on t1.reference_number like '%'+t2.id or t1.reference_number like t2.id+'%'
left outer join #t3 t3 on t1.reference_number like '%'+t3.id or t1.reference_number like t3.id+'%'
Result (left join):
id data_column reference_number id data_column id data_column
1 some data 1528,ss-456 ss-456 some data 1528 some data
2 some data 9523 NULL NULL 9523 some data
3 some data ss-952 ss-952 some data NULL NULL
4 some data NULL NULL NULL NULL NULL
You can change 'left outer join' to 'inner join' for exact match.

Answer 2

Clumsy design, clumsy solution:
SELECT *
FROM Table1
INNER JOIN Table2 ON ',' + Table1.reference_number + ',' LIKE '%,' + Table2.ID + ',%'
INNER JOIN Table3 ON ',' + Table1.reference_number + ',' LIKE '%,' + Table3.ID + ',%'
You must append leading and trailing commas to make sure that, for example, 1528,ss-456asdf does not match %ss-456%.

Answer 3

I see two problems here. First is the inconsistent type of ID in table 2 and 3 and aggregation of referenced keys in table 1. Here is an example how to solve both problems. To split REFERENCE_NUMBER column I used STRING_SPLIT function.
Update:
I added the solution which should work with SQL Server 2012.
I assumed that you wish to join data from table 1 with 2 or 3 depending in existence of this data. This is just my idea what you wanted to achive.
-- data preparing
declare #t1 as table(
id int
,data_column varchar(20)
,reference_number varchar(20)
)
declare #t2 as table(
id varchar(20)
,data_column varchar(20)
)
declare #t3 as table(
id int
,data_column varchar(20)
)
insert into #t1 values(1,'some data','1528,ss-456'),(2,'some data','9523'),(3,'some data','ss-952'),(4,'some data',null);
insert into #t2 values('ss-456','some data'),('ss-952','some data');
insert into #t3 values(1528,'some data'),(9523,'some data');
-- Solution example version >= 2016
with base as (
select t1.id,t1.data_column,f1.value from #t1 t1 outer apply string_split(t1.reference_number,',') f1)
select b.id,b.data_column,b.value,t2.data_column from base b join #t2 t2 on b.value = t2.id
union all
select b.id,b.data_column,b.value,t3.data_column from base b join #t3 t3 on try_cast(b.value as int ) = t3.id
union all
select b.id,b.data_column,b.value,null from base b where b.value is null;
-- Solution for SQL Version < 2016
with base as (
select t1.id,t1.data_column,f1.value from #t1 t1 outer apply(
SELECT Split.a.value('.', 'NVARCHAR(MAX)') value
FROM
(
SELECT CAST('<X>'+REPLACE(t1.reference_number, ',', '</X><X>')+'</X>' AS XML) AS String
) AS A
CROSS APPLY String.nodes('/X') AS Split(a)
) f1)
select b.id,b.data_column,b.value,t2.data_column from base b join #t2 t2 on b.value = t2.id
union all
select b.id,b.data_column,b.value,t3.data_column from base b join #t3 t3 on try_cast(b.value as int ) = t3.id
union all
select b.id,b.data_column,b.value,null from base b where b.value is null;

Answer 4

You will require a function to divide the comma separated sting into rows. If you don't have access to thr inbuilt string_split() function (as of mssql 2017 with compatibility of 130) there are several to choose from here
CREATE TABLE table1(
ID INTEGER NOT NULL PRIMARY KEY
,Data_column VARCHAR(10) NOT NULL
,reference_number VARCHAR(11)
);
INSERT INTO table1(ID,Data_column,reference_number) VALUES
(1,'t1somedata','1528,ss-456')
, (2,'t1somedata','9523')
, (3,'t1somedata','ss-952')
, (4,'t1somedata',NULL);
CREATE TABLE table2(
ID VARCHAR(6) NOT NULL PRIMARY KEY
,Data_column VARCHAR(10) NOT NULL
);
INSERT INTO table2(ID,Data_column) VALUES
('ss-456','t2somedata'),
('ss-952','t2somedata');
CREATE TABLE table3(
ID VARCHAR(6) NOT NULL PRIMARY KEY
,Data_column VARCHAR(10) NOT NULL
);
INSERT INTO table3(ID,Data_column) VALUES
('1528','t3somedata'),
('9523','t3somedata');
I have used this splistring function, but you can use almost any of the many freely available.
CREATE FUNCTION dbo.SplitStrings_Moden
(
#List NVARCHAR(MAX),
#Delimiter NVARCHAR(255)
)
RETURNS TABLE
WITH SCHEMABINDING AS
RETURN
WITH E1(N) AS ( SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1
UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1
UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1),
E2(N) AS (SELECT 1 FROM E1 a, E1 b),
E4(N) AS (SELECT 1 FROM E2 a, E2 b),
E42(N) AS (SELECT 1 FROM E4 a, E2 b),
cteTally(N) AS (SELECT 0 UNION ALL SELECT TOP (DATALENGTH(ISNULL(#List,1)))
ROW_NUMBER() OVER (ORDER BY (SELECT NULL)) FROM E42),
cteStart(N1) AS (SELECT t.N+1 FROM cteTally t
WHERE (SUBSTRING(#List,t.N,1) = #Delimiter OR t.N = 0))
SELECT Item = SUBSTRING(#List, s.N1, ISNULL(NULLIF(CHARINDEX(#Delimiter,#List,s.N1),0)-s.N1,8000))
FROM cteStart s;
This is what the data looks like using the splitstring function:
select *
from table1
cross apply SplitStrings_Moden(reference_number,',')
ID | Data_column | reference_number | Item
-: | :---------- | :--------------- | :-----
1 | t1somedata | 1528,ss-456 | 1528
1 | t1somedata | 1528,ss-456 | ss-456
2 | t1somedata | 9523 | 9523
3 | t1somedata | ss-952 | ss-952
4 | t1somedata | null | null
and now joining to the other tables:
select
*
from (
select *
from table1
cross apply SplitStrings_Moden(reference_number,',')
) t1
left join table2 on t1.item = table2.id
left join table3 on t1.item = table3.id
where t1.item is not null
GO
ID | Data_column | reference_number | Item | ID | Data_column | ID | Data_column
-: | :---------- | :--------------- | :----- | :----- | :---------- | :--- | :----------
1 | t1somedata | 1528,ss-456 | 1528 | null | null | 1528 | t3somedata
1 | t1somedata | 1528,ss-456 | ss-456 | ss-456 | t2somedata | null | null
2 | t1somedata | 9523 | 9523 | null | null | 9523 | t3somedata
3 | t1somedata | ss-952 | ss-952 | ss-952 | t2somedata | null | null
db<>fiddle here

Answer 5

TRY THIS: If your reference_number is fixed and always stored IDs upto 2 only then you can go with the below approach
SELECT *
FROM(
SELECT ID,
data_column,
CASE WHEN PATINDEX ( '%,%', reference_number) > 0 THEN
SUBSTRING(reference_number, PATINDEX ( '%,%', reference_number)+1, LEN(reference_number))
ELSE reference_number END AS ref_col
FROM #table1
UNION
SELECT ID,
data_column,
CASE WHEN PATINDEX ( '%,%', reference_number) > 0 THEN
SUBSTRING(reference_number, 0, PATINDEX ( '%,%', reference_number))
END
FROM #table1) t1
LEFT JOIN #table2 t2 ON t2.id = t1.ref_col
LEFT JOIN #table3 t3 ON t3.id = t1.ref_col
WHERE t1.ref_col IS NOT NULL
OUTPUT:
ID data_column ref_col ID Data_column ID Data_column
1 some data 1528 NULL NULL 1528 some data
1 some data ss-456 ss-456 some data NULL NULL
2 some data 9523 NULL NULL 9523 some data
3 some data ss-952 ss-952 some data NULL NULL
4 some data null NULL NULL NULL NULL

Answer 6

You can Implement and get desired result using Substring and charIndex functions on the reference_number.
I upvoted an answer of 'is_oz' since i used his ready made schema to test and build a query for you.
below is the final query i build after several tries i made here:
select * from abc
left join abc2 on abc2.id = case when charindex(',',abc.reference_number) > 0
then substring(abc.reference_number
,charindex(',',abc.reference_number)+1
,len(abc.reference_number)-(charindex(',',abc.reference_number)-1)
)
else abc.reference_number
end
left join abc3 on abc3.id = case when charindex(',',abc.reference_number) > 0
then substring(abc.reference_number
,0
,(charindex(',',abc.reference_number))
)
else abc.reference_number
end
As per your requirement as much as i understand, it is returning all the matched rows from 2 other tables but still i hope this fulfills all the requirements you seek in your question. :)