SQL code trying to concatenate two tables - sql

Question

I have table 1 and table 2 which have the same columns. I want to concatenate the tables where if ID1 and ID2 (not ‘value) in Table 1 match with Table 2, than only include Table 1 and not Table 2. If they do not match, than include.
FYI all ‘values’ in Table 2 are NULL. After doing the above step, if the output table has value ‘NULL’ and fort his row, Table 2 has a matching ID1 with Table 1, than apply this value from Table 1.
Both steps can be rolled up into one. I have broken it out to make sure its easy to understand.
I attempted this using The following code, I just have to somehow get rid of the duplicated entries and do the second part of the question. I don’t think my way is very effective. Would love some guidance…
SELECT ID1, ID2, value
FROM table1
UNION
SELECT ID1, ID2, value
FROM table2
Table 1:
ID1
ID2
value
1
1
0.1
1
2
0.2
2
2
0.2
Table 2:
ID1
ID2
value
1
1
NULL
2
1
NULL
For task #1, this is the output I want:
ID1
ID2
value
1
1
0.1
1
2
0.2
2
2
0.2
2
1
NULL
For task #2, this is the output I want:
ID1
ID2
value
1
1
0.1
1
2
0.2
2
2
0.2
2
1
0.2

Answer 1

As I mentioned in the comments, a FULL OUTER JOIN might be more appropriate here. Based on the data you've given, it'll likely look something like this:
SELECT ISNULL(T1.ID1,T2.ID1) AS ID1,
ISNULL(T1.ID2,T2.ID2) AS ID2,
ISNULL(T1.Value,T2.Value) AS Value
FROM dbo.Table1 T1
FULL OUTER JOIN dbo.Table2 T2 ON T1.ID1 = T2.ID1
AND T1.ID2 = T2.ID2;

Answer 2

So extract rows from table1 and ignore ID matching rows from table2. table1 is the first SELECT, table2 is the second SELECT in this answer.
UNION is the right way to do this academically.
It works and is fast on a few test rows.
Then there are tables with millions/billions of rows and the UNION approach is suddenly slow. UNION gets slow when:
more rows than the memory of the server can handle
indexes are too wide or include too much
tables/selection are wide
one source comes across the network
UNION works by checking each row of table1 against all rows of table2. This is not a real set operation.
If UNION gets slow then following strategies help:
check query plans (or use tools like SQL Sentry Explorer)
build an index lookup between both tables first (CTE or materialised)
proper indexing
use local tables instead of networked data
split table2 into multiple chunks
The replace UNION with UNION ALL and WHERE clauses in both parts that ensure the result is never overlapping. In essence the WHERE clause against table1 is WHERE NOT for table2.
My preferred approach for comparing medium sized sets of data or larger is besides indexing the CTE approach / UNION ALL / WHERE + WHERE NOT. Simply it was fast enough for me with the least overhead so far.