Apply conditions in projection level - sql

Question

Is there a simple way to apply conditions in projection level? I mean In a series of fields(columns) how can we choose only the fields those have specific conditions?
Example:
Assume we have a query with below result:
Id Col1 Col2 Col3 Col4 Col5 Col6 Col7 Col8 Col9 Col10 Col11
- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ----
1 V1_1 NULL NULL V4_1 NULL NULL NULL NULL NULL V10_1 NULL
2 V1_2 NULL NULL NULL NULL NULL NULL NULL NULL V10_2 NULL
3 V1_3 NULL NULL V4_3 NULL NULL NULL NULL NULL V10_3 NULL
how can we select only the fields which contains none null values while we don't know which of them contains none null values? so result would be this:
Id Col1 Col4 Col10
- ---- ---- ----
1 V1_1 V4_1 V10_1
2 V1_2 NULL V10_2
3 V1_3 V4_3 V10_3
Usually I do such these queries with hardcoding which I may query sys.all_obejcts, sys.all_columns in MsSQL Server and all_tab_cols, all_tables , etc in Oracle. and use none, one or more cursors , case statements, pivot, multiple ifs or dynamic sql in it.
in example: for a table with above data in T-SQL I can do:
DECLARE #sql VARCHAR(500)
SET #sql='SELECT Id,'
IF EXISTS (SELECT Col1 FROM tblTEST WHERE Col1 IS NOT NULL) SET #sql=#sql+'Col1,'
/* 9 other IF statements here */
IF EXISTS (SELECT Col11 FROM tblTEST WHERE Col11 IS NOT NULL) SET #sql=#sql+'Col11,'
SET #sql=SUBSTRING(#sql,1,LEN(#sql)-1)+' FROM tblTEST'
EXEC(#sql)
but as you see the idea is not common at all and it's a little hard to do such coding for such these scenarios
Do you know a simple and better way or a pattern for such this queries?
The RDBMS does not matter, a good solution for any RDBMS is really useful.
Please I want answers to be a good idea in general not only for the example I mentioned.

Answer 1

This is a T-SQL solution. All you should have to do is change #TargetTable to your desired table. Hope this helps! I commented my code explaining everything, but if you have any questions, let me know!
IF OBJECT_ID('myTable') IS NOT NULL
DROP TABLE dbo.myTable;
CREATE TABLE myTable
(
ID INT IDENTITY(1,1),
Col1 VARCHAR(25),
Col2 VARCHAR(25),
Col3 VARCHAR(25)
)
INSERT INTO dbo.myTable(Col1,Col2,Col3)
VALUES ('v1_1',NULL,NULL),
('v1_2','v2_2',NULL),
('v1_3',NULL,NULL);
/* Now that I've set up the problem, here's for the actual code*/
--Set your desired table
DECLARE #TargetTable VARCHAR(100) = 'myTable'
--If this global temp table exists, drop it
IF OBJECT_ID('##tempTable') IS NOT NULL
EXEC('DROP TABLE ##tempTable');
--This is going to take one character and cast it to NCHAR(1)
--If there are no characters, it will simply return NULL
DECLARE #AllColumns VARCHAR(MAX);
SELECT #AllColumns = COALESCE(#AllColumns + ',','') + 'CAST(LEFT(' + QUOTENAME(COLUMN_NAME) + ',1) AS NCHAR(1)) AS ' + QUOTENAME(COLUMN_NAME)
FROM INFORMATION_SCHEMA.COLUMNS A
WHERE TABLE_NAME = #TargetTable
/*Load your char results into a global temp table*/
EXEC('SELECT ' + #AllColumns + ' INTO ##tempTable FROM ' + #TargetTable);
/*Here's where the fun beings*/
/* #DesiredCol holds the name of columns that are not null.
UNPIVOT the data.
Where clause gets rid of columns with only NULL values.
Group by only allows distinct columns.
You the select from your ORIGINAL table, not my funky nchar(1) global temp table.
*/
EXEC(
'DECLARE #DesiredCol VARCHAR(MAX);
SELECT #DesiredCol = COALESCE(#DesiredCol + '','','''') + QUOTENAME(col)
FROM ##temptable
UNPIVOT
(
val FOR col IN ([ID],[Col1],[Col2],[Col3])
) pvt
WHERE pvt.val IS NOT NULL
GROUP BY col
EXEC (''SELECT '' + #DesiredCol + '' FROM ##' + #TargetTable + ''')'
)
--Cleanup
IF OBJECT_ID('##tempTable') IS NOT NULL
EXEC('DROP TABLE ##tempTable');