Check missing columns in SQL table before copying into another table - sql-server

Question

I have 1000s of tables like ABC_0001, ABC_0002, ABC_0003 and so on.. and I would like to do 2 things:
From each table, I am looking to copy columns x, y and z in new table.
However not all tables have all 3 columns e.g. ABC_0001 is missing column x, ABC_0002 is missing y and z or any other combinations.
How can I check if the columns is present in selected table and if not assign default value say "0".
There are actually many variables instead of x, y, z which needs to be copied from all the tables into new table. I have this list in a separate table (say TABLE_COL_NAME with variable name, var type e.g. a int, b varchar(200) and so on) and this will change in future, hence I would like to create NEW_TABLE and assign variables in #VarNames dynamically using TABLE_COL_NAME.
Thanks Annamalai for helping with below base code. cheers.
Edit: This is SSMS v18.7
DECLARE
#TableName VARCHAR(500)
,#Sql NVARCHAR(MAX) = ''
,#Id INT
,#VarNames VARCHAR(8000)
DROP TABLE IF EXISTS TABLE_NEW
CREATE TABLE TABLE_NEW (x int, y varchar(200), z varchar(500))
SET #VarNames = 'x, y, z'
DECLARE Table_Cursor CURSOR
FOR
SELECT
ROW_NUMBER() OVER (ORDER BY TABLE_NAME ASC) Id
,TABLE_NAME
FROM INFORMATION_SCHEMA.TABLES
WHERE TABLE_TYPE = 'BASE TABLE'
AND TABLE_NAME LIKE 'ABC%'
OPEN Table_Cursor
FETCH NEXT FROM Table_Cursor INTO #Id,#TableName
WHILE ##FETCH_STATUS = 0
BEGIN
IF(#Id = 1)
BEGIN
SET #Sql = #Sql + 'SELECT ' + #VarNames + 'FROM '+#TableName ----Modify the columns based on your column names
SELECT #SQL
END
ELSE
BEGIN
SET #Sql = #Sql + ' UNION ALL SELECT '+ #VarNames + 'FROM '+#TableName --Modify the columns based on your column names
END
FETCH NEXT FROM Table_Cursor INTO #Id,#TableName
END
CLOSE Table_Cursor
DEALLOCATE Table_Cursor
INSERT INTO TABLE_NEW
EXEC (#Sql)

Answer 1

You have a lousy data model. It looks like you need some code to help fix it.
You can construct a union all statement as follows. First, construct the appropriate logic for the column names for each table:
select 'select ' +
coalesce(max(case when c.column_name = 'x' then c.column_name end), 'null as x') + ', ' +
coalesce(max(case when c.column_name = 'y' then c.column_name end), 'null as y') + ', ' +
coalesce(max(case when c.column_name = 'z' then c.column_name end), 'null as z') + ' ' +
'from ' + c.table_name
from information_schema.columns c
where table_name like 'abc_%';
group by c.table_name;
Then construct the SQL statement:
select string_agg(sql,
'
union all
'
)
from (select 'select ' +
coalesce(max(case when c.column_name = 'x' then c.column_name end), 'null as x') + ', ' +
coalesce(max(case when c.column_name = 'y' then c.column_name end), 'null as y') + ', ' +
coalesce(max(case when c.column_name = 'z' then c.column_name end), 'null as z') + ' ' +
'from ' + c.table_name as sql
from information_schema.columns c
where table_name like 'abc_%'
group by c.table_name
) t;
You can then assign this to a variable and run it. A SQL Fiddle illustrates this.
Note that the above is intentionally simplified to focus on the steps, not taking into account:
The _ in the table name is a wildcard for like.
The schema name as well as the table name.
Column names that need to be escaped.
You can also use the constructed SQL to create a view rather than execute it directly.