Group by T-SQL vs. MySQL (single column) - sql

Question

I am new to SQL Server/used to MySQL databases and I am running into an issue that I never ran into with MySQL. I am looking to pull all current policy numbers, the name of the company/person it belongs to, their total premium, and whether or not they have what we call 'equipment breakdown' coverage. This is all pretty simple, the issue I am having is with grouping. I want to group by one column only, aka one distinct policy number, the company name, a sum of the premium (it is possible to have several premium amounts both negative and positive so I want to sum these to see what the true total is), and a simple Yes or No column for equipment breakdown.
Here is the query I am running:
SELECT pol_num as policy_number,
insd_name as insureds_name,
SUM(amt) as 'total_premium',
(SELECT
CASE
WHEN cvg_desc = 'Equipment Breakdown'
THEN 'Y'
ELSE 'N'
END) as 'equipment_breakdown'
FROM bapu.dbo.fact_prem
WHERE '2014-05-06' between d_pol_eff and d_pol_exp
AND amt_type = 'Premium'
AND amt_desc = 'Written Premium'
GROUP BY pol_num
ORDER BY policy_number
I get the an error saying that I need to group by insd_name and cvg_desc as well, but I DON'T want that as it gives me duplicate policy numbers.
Here is an example of what I get when I group everything it tells me to:
policy_number insureds_name total_premium equipment_breakdown
001 company a 0.00 n
001 company a 25,000.00 n
001 company a -10,000.00 n
002 company b 100.00 y
002 company b 10,000.00 y
Here is an example of the results I want:
policy_number insureds_name total_premium equipment_breakdown
001 company a 15,000.00 n
002 company b 10,100.00 y
Basically, I just want to group by the policy number and sum the premium amounts. Above is how I would achieve this in MySQL, how can I achieve the results I am looking for in SQL Server?
Thanks

Answer 1

MySQL doesn't require all non-aggregate fields to be included in the GROUP BY clause, even though not doing so can yield unexpected results. SQL Server requires this, so you are forced to decide how you want to handle multiple insd_name values for a given pol_num, you can use MAX(), MIN(), or if the values are always the same, just add them to your GROUP BY:
SELECT pol_num AS policy_number
, MAX(insd_name) AS insureds_name
, SUM(amt) AS 'total_premium'
, MAX(CASE WHEN cvg_desc = 'Equipment Breakdown' THEN 'Y'
ELSE 'N'
END) AS 'equipment_breakdown'
FROM bapu.dbo.fact_prem
WHERE '2014-05-06' BETWEEN d_pol_eff AND d_pol_exp
AND amt_type = 'Premium'
AND amt_desc = 'Written Premium'
GROUP BY pol_num
ORDER BY policy_number
Or:
SELECT pol_num AS policy_number
, insd_name AS insureds_name
, SUM(amt) AS 'total_premium'
, CASE WHEN cvg_desc = 'Equipment Breakdown' THEN 'Y'
ELSE 'N'
END AS 'equipment_breakdown'
FROM bapu.dbo.fact_prem
WHERE '2014-05-06' BETWEEN d_pol_eff AND d_pol_exp
AND amt_type = 'Premium'
AND amt_desc = 'Written Premium'
GROUP BY pol_num
, insd_name
, CASE WHEN cvg_desc = 'Equipment Breakdown' THEN 'Y'
ELSE 'N'
END
ORDER BY policy_number

Answer 2

It looks like the cvg_desc column is probably what's messing you up. You want to group by the resulting Y or N from your CASE statement, but SQL server is grouping by the original cvg_desc column. You could approach this in a way that resolves the CASE statement before it groups. For example, wrap the main query in a common table expression (CTE), which is sort of like an inline-view. Then with the equipment breakdown column reduced to just a Y or an N, a subsequent query from the CTE with your SUM aggregation on premium should give you the results you desire:
WITH Policies(policy_number, insureds_name, premium, equipment_breakdown) AS
(
SELECT
pol_num
,insd_name
,amt
,(CASE WHEN cvg_desc = 'Equipment Breakdown' THEN 'Y' ELSE 'N' END)
AS 'equipment_breakdown'
FROM
bapu.dbo.fact_prem
WHERE
'2014-05-06' BETWEEN d_pol_eff AND d_pol_exp
AND
amt_type = 'Premium'
AND
amt_desc = 'Written Premium'
)
SELECT
policy_number
,insureds_name
,SUM(premium) AS total_premium
,equipment_breakdown
FROM
Policies
GROUP BY
policy_number
,insureds_name
,equipment_breakdown

Answer 3

You'll need an aggregate function on the fields you don't want to group by. A simple one to use is MAX which works with most types;
SELECT pol_num as policy_number,
MAX(insd_name) as insureds_name,
SUM(amt) as 'total_premium',
(SELECT
CASE
WHEN MAX(cvg_desc) = 'Equipment Breakdown'
THEN 'Y'
ELSE 'N'
END) as 'equipment_breakdown'
FROM bapu.dbo.fact_prem
WHERE '2014-05-06' between d_pol_eff and d_pol_exp
AND amt_type = 'Premium'
AND amt_desc = 'Written Premium'
GROUP BY pol_num
ORDER BY policy_number
The reason SQL Server wants this is that it likes to give deterministic answers, for example
column_a | column_b
1 | 1
1 | 2
...grouped by only column_a would in MySQL give either 1 or 2 as an answer for column_b, while SQL Server wants you to tell it explicitly which one to use.

Answer 4

I would probably write this as below -- did not test
SELECT pol_num as policy_number,
insd_name as insureds_name,
SUM(amt) as total_premium
CASE
WHEN cvg_desc = 'Equipment Breakdown'
THEN 'Y'
ELSE 'N'
END as equipment_breakdown
FROM bapu.dbo.fact_prem
WHERE '2014-05-06' between d_pol_eff and d_pol_exp
AND amt_type = 'Premium'
AND amt_desc = 'Written Premium'
GROUP BY
pol_num, policy_number,
CASE
WHEN cvg_desc = 'Equipment Breakdown'
THEN 'Y'
ELSE 'N'
END
ORDER BY policy_number