Loop through table data and compare using split - sql

Question

I have two tables and i need to compare data and update/insert one table records. What iam trying to do is I need to take each record from Table1,
use a split function then for each text in split, compare dataelement field between both these tables. We are syncing data in Table2 to similar to Table1.
Please let me know how this can be done. I am ok using cursor or merge. This is the scenario
DataTable:
dataId dataelement
1 Check
2 System
3 Balances
4 City
5 State
6 Zip
7 Other
Table1:
Id reqId dataelementValues
1 52 Check
2 52 City;State;System
3 52 Other
Table2:
elId dataId dataelement reqId Active
1 6 Zip 52 1
2 1 Check 52 1
3 4 city 52 1
4 5 State 52 1
Outcome Should be similar to after compare in table2
Table2:
elId dataId dataelement reqId Active
1 6 Zip 52 0 (Should be set to inactive as it exists in table2 but not in table1)
2 1 Check 52 1 (NO Updates as it exists in both the tables)
3 4 city 52 1 (NO Updates as it exists in both the tables)
4 5 State 52 1 (NO Updates as it exists in both the tables)
5 2 System 52 1 (Get the dataid for system from datatable and insert in table2 as it exists in table1 but not in table2)
6 7 Other 52 1 (Get the dataid for other from datatable and insert in table2 as it exists in table1 but not in table2)
This is where iam at, not sure how to set inactive on table2.
WHILE Exists(Select * from #Table1)
BEGIN
Select #currentId = Id, #dataValue = dataelementValues FROM #Table1 where rowID=(SELECT top 1 rowID from #Table1 order by rowID asc)
SET #pos = 0
SET #len = 0
WHILE CHARINDEX(';', #dataValue, #pos+1)>0
BEGIN
SET #dataValueValue = SUBSTRING(#dataValue, #pos, CHARINDEX('|', #dataValue, #pos+1) - #pos)
SET #glbaDEId = (Select DataTable.dataId from datatable where dataelement = #dataValue)
IF NOT Exists (Select * from #Table2 Where DataElement=#dataValue)
BEGIN
--Insert into table2
END
SET #pos = CHARINDEX('|', #dataValue, #pos+#len) +1
END
DELETE from #Table1 where rowID=(SELECT top 1 rowID from #Table1 order by rowID asc )
END

Answer 1

You can try using a MERGE statement with a few other tricks.
Merge Guide
-- Create a CTE that will split out the combined column and join to DataTable
-- to get the dataId
;WITH cteTable1Split AS
(
SELECT reqId, dt.* FROM
(
SELECT
[dataelement] = y.i.value('(./text())[1]', 'nvarchar(4000)'),
reqId
FROM
(
-- use xml to split column
-- http://sqlperformance.com/2012/07/t-sql-queries/split-strings
SELECT x = CONVERT(XML, '<i>'
+ REPLACE([dataelementValues], ';', '</i><i>')
+ '</i>').query('.'),
reqId
FROM Table1
) AS a CROSS APPLY x.nodes('i') AS y(i)
) a
JOIN DataTable dt ON dt.[dataelement] = a.[dataelement]
)
-- Merge Table2 with the CTE
MERGE INTO Table2 AS Target
USING cteTable1Split AS Source
ON Target.[dataelement] = Source.[dataelement]
-- If exists in Target (Table2) but not Source (CTE) then UPDATE Active flag
WHEN NOT MATCHED BY Source THEN
UPDATE SET ACTIVE = 0
-- If exists in Source (CTE) but not Target (Table2) then INSERT new record
WHEN NOT MATCHED BY TARGET THEN
INSERT ([dataId], [dataelement], [reqId], [Active])
VALUES (SOURCE.[dataId], SOURCE.[dataelement], SOURCE.[reqId], 1);
SQL Fiddle

Answer 2

You've not mentioned whether you have control over the structure of these tables, so I'm going to go ahead and suggest you redesign Table1 to normalise the dataelementValues column.
That is, instead of this:
Table1:
Id reqId dataelementValues
1 52 Check
2 52 City;State;System
3 52 Other
You should be storing this:
Table1_New:
Id reqId dataelementValues
1 52 Check
2 52 City
2 52 State
2 52 System
3 52 Other
You may also need a new, surrogate, primary key column on the table, using an IDENTITY(1,1) specification.
Storing your data like this is how relational databases are intended to be used/designed. As well as simplifying the problem at hand right now, you might find it removes potential problems in the future as well.

Answer 3

The main challenge here is creating a rowset with the dataelementValues correctly split into separate rows. The accepted answer shows clearly how this can be used as the source for a merge statement to achieve the update and insert operation.
However, there are alternative ways to create the split or normalised rowset.
One way, which I personally prefer for clarity over resorting to xml in this particular case, is a like join. This takes advantage of the fact that you already have the separate rows you need in DataTable, just not with all the columns you need.
Select DT.dataId, DT.dataelement, T1.reqid
From Table1 T1
Inner join DataTable DT
On T1.dataelementValues like '%' + DT.dataelement + '%'
I've not been able to test this right now but should give the rowset you need, because the like operator causes one T1 row to match three times to the corresponding DT rows.