I have a table in a Postgres that stores 10x10 matrices, where each row has it's own entry, defined as:
id, matrix_id, row_id, col1, col2, col3...
I'd like to compute the trace (sum of main diagonal) for every matrix identified by its matrix_id, that is, for every matrix_id, I would like to get (col1 where row_id=1) + (col2 where row_id=2) + (col3 where row_id=3)...
I've tried grouping it by matrix_id but then I cannot use subqueries, something like:
select matrix_id, (select col1 where row_id=1) + (col2 where row_id=2) +
(col3 where row_id=3) ... from matrix group by matrix_id;
but it doesn't work this way.
How could I do that?
So long as they are all 10x10 matrices, use a case statement like so:
select matrix_id,
sum(
case row_id
when 1 then col1
when 2 then col2
when 3 then col3
when 4 then col4
when 5 then col5
when 6 then col6
when 7 then col7
when 8 then col8
when 9 then col9
when 10 then col10
end
) as trace
from matrix
group by matrix_id;
Had variable-sized matrices been allowed, you could transpose columns to rows via to_jsonb() and then sum where row_id = <column suffix>.
EDIT TO ADD
Based on your comment, you really should update your version of PostgreSQL. That said, try a CTE to filter on the new trace column:
with traces as (
select matrix_id,
sum(
case row_id
when 1 then col1
when 2 then col2
when 3 then col3
when 4 then col4
when 5 then col5
when 6 then col6
when 7 then col7
when 8 then col8
when 9 then col9
when 10 then col10
end
) as trace
from matrix
group by matrix_id
)
select *
from traces
where trace > 100;
I've put together a reconciliation tool in SQL Server which identifies the number of record breaks by field (col 2 - col 4) between two identical (data types/structure) sources. The output returned is in the format below, grouped on col 1.
Col1 Col2 Col3 Col4
X 0 0 1
Y 0 1 1
Z 1 0 1
I am trying to manipulate the output so that it provides a list of the Col 1 identifier and the name of any column names (col 2 - col 4) which have breaks (value > 0).
The expected output based on the above data would look like this.
Col1 FieldBreak
X Col2
Y Col3
Y Col4
Z Col2
Z Col4
I'm newer to SQL (6 months of professional experience) and am stuck. Any help would be much appreciated!
In any database, you can use:
select col1, 'col2' as col
from t
where col2 = 1
union all
select col1, 'col3' as col
from t
where col3 = 1
union all
select col1, 'col4' as col
from t
where col4 = 1;
There are probably more efficient methods, but those depend on the database. And for a small table efficiency may not be a concern.
In SQL Server, you would unpivot using apply:
select t.col1, v.*
from t cross apply
(values ('col2', t.col2), ('col3', t.col3) . . .
) v(col, val)
where v.val is not null;
If you have a lot of columns, you can construct the expression using a SQL statement (from INFORMATION_SCHEMA.COLUMNS) and/or using a spreadsheet.
I have a table like this one
ID Col1 Col2 Col3
-- ---- ---- ----
1 7 NULL 12
2 2 46 NULL
3 NULL NULL NULL
4 245 1 792
I wanted a query that yields the following result
ID Col1 Col2 Col3 MIN
-- ---- ---- ---- ---
1 7 NULL 12 7
2 2 46 NULL 2
3 NULL NULL NULL NULL
4 245 1 792 1
I mean, I wanted a column containing the minimum values out of Col1, Col2, and Col 3 for each row ignoring NULL values. In a previous question (What's the best way to select the minimum value from multiple columns?) there is an answer for non NULL values. I need a query as efficient as possible for a huge table.
Select Id,
Case When Col1 < Col2 And Col1 < Col3 Then Col1
When Col2 < Col1 And Col2 < Col3 Then Col2
Else Col3
End As MIN
From YourTableNameHere
Assuming you can define some "max" value (I'll use 9999 here) that your real values will never exceed:
Select Id,
Case When Col1 < COALESCE(Col2, 9999)
And Col1 < COALESCE(Col3, 9999) Then Col1
When Col2 < COALESCE(Col1, 9999)
And Col2 < COALESCE(Col3, 9999) Then Col2
Else Col3
End As MIN
From YourTableNameHere;
You didn't specify which version of Teradata you're using. If you're using version 14+ then you can use least.
Unfortunately least will return null if any of its arguments are null. From the docs:
LEAST supports 1-10 numeric values.
If numeric_value is the data type of the first argument, the return
data type is numeric. The remaining arguments in the input list must
be the same or compatible types. If either input parameter is NULL,
NULL is returned.
But you can get around that by using coalesce as Joe did in his answer.
select id,
least(coalesce(col1,9999),coalesce(col2,9999),coalesce(col3,9999))
from mytable
This might work:
Select id, Col1, Col2, Col3, least(Col1, Col2, Col3) as MIN From YourTableNameHere
in this way you don't need to check for nulls, just use min and a subquery
select tbl.id,tbl.col1,tbl.col2,tbl.col3,
(select min(t.col)
from (
select col1 as col from tbl_name t where t.id=tbl.id
union all
select col2 as col from tbl_name t where t.id=tbl.id
union all
select col3 as col from tbl_name t where t.id=tbl.id
)t)
from tbl_name tbl
Output:
1 7 NULL 12 7
2 2 46 NULL 2
3 NULL NULL NULL NULL
4 245 1 792 1
Just modify your query with coalesce():
Select Id,
(Case When Col1 <= coalesce(Col2, col3, col1) And
Col1 <= coalesce(Col3, col2, col1)
Then Col1
When Col2 <= coalesce(Col1, col3, col2) And
Col2 <= coalesce(Col3, col1, col2)
Then Col2
Else Col3
End) As MIN
From YourTableNameHere;
This doesn't require inventing a "magic" number or over-complicating the logic.
I found this solution to be more efficient than using multiple case statement clauses, which can get extremely lengthy when evaluating data from several columns across one row.
Also, I can't take credit for this solution as I found it on some website a year or so ago. Today I needed a refresh on this logic, and I couldn't find it anywhere. I found my old code and decided to share it in this forum now.
Creating your test table:
create table #testTable(ID int, Col1 int, Col2 int, Col3 int)
Insert into #testTable values(1,7,null,12)
Insert into #testTable values(2,2,46,null)
Insert into #testTable values(3,null,null,null)
Insert into #testTable values(4,245,1,792)
Finding min value in row data:
Select ID, Col1, Col2, Col3 ,(SELECT Min(v) FROM ( VALUES (Col1), (Col2), (Col3) ) AS value(v)) [MIN] from #testTable order by ID
I want in SQL : if a field is equal to a specific value, I want this record if not I want the record with this field equal to null.
I can try to that:
SELECT TOP 1 COL1, COL2, COL3
FROM TABLE1 WHERE (COL2 = MY_SPECIFIC_VALUE OR COL2 IS NULL) AND COL3 = '42'
AND COL1 = 3
But, what is the result returned? The smallest id? Or it is not specified?
Assuming that this is the real question:
If a field is equal to a specific value, I want this record if not I want the record with this field equal to null.
You can do this as:
SELECT TOP 1 COL1, COL2, COL3
FROM TABLE1
WHERE (COL2 = MY_SPECIFIC_VALUE or COL2 IS NULL) AND
COL3 = '42' AND COL1 = 3
ORDER BY (CASE WHEN COL2 = MY_SPECIFIC_VALUE THEN 1 ELSE 2 END);
I have a table as
COl1____COL2____COL3
1_________1_____val1
1_________2_____val2
1_________3_____val3
2_________1_____val1
2_________2_____val2
2_________3_____val3
3_________1_____val1
3_________2_____val2
3_________4_____val4
No I want to have unique values from COL1 such that it should only show me COL1 value that does not have a value "3" under COL2
i.e. I want a query to return me
Result
3
Any help is much appreciated
SELECT col1
FROM YourTable
EXCEPT
SELECT col1
FROM YourTable
WHERE col2 = 3
Or
SELECT col1
FROM YourTable
GROUP BY col1
HAVING MAX(CASE
WHEN Col2 = 3 THEN 1
ELSE 0
END) = 0
select col1
from Tab
where col1 not in (select col1 from tab where col2 = 3)