Sql query optimization for calculating ratio
-
30-04-2021 - |
Question
I have 2 tables (it is only part of table columns that are needed for this query):
Items: Id int, OwnerId int
ItemsRelations: RelationId int, ItemId Int (FK Items(Id))
For each OwnerId i need to calculate result of ratio function: Count of user elements that have links to ItemsRelationstable / count of all user elements.
I wrote this query, but it is slow and not optimal. Is there easier and faster method to do that?
SELECT OwnerId , (100 * SUM(HasRelation))/COUNT(Id) AS Ratio
FROM (
SELECT
oi.OwnerId,
oi.Id,
(CASE WHEN SUM(ir.Id) > 0 THEN 1 ELSE 0 END) HasRelation
FROM Items AS oi
LEFT JOIN ItemsRelations AS ir ON ir.ItemId = oi.Id
GROUP BY oi.Id, oi.OwnerId) tempTab
GROUP BY OwnerId
Solution
SELECT oi.OwnerId, Ratio = 1.0 * Count(ir.ItemId) / Count(*)
FROM Items AS oi
LEFT JOIN ItemsRelations AS ir ON ir.ItemId = oi.Id
GROUP BY oi.OwnerId
Notes:
- 1.0 is added to convert the expression into a float result
- Count(ir.ItemId) skips NULLS, so it counts only those that have a relationship
If Items
to ItemsRelations
is one to many, then you may need to DISTINCT it
SELECT oi.OwnerId, Ratio = 1.0 * Count(DISTINCT ir.ItemId) / Count(DISTINCT oi.Id)
FROM Items AS oi
LEFT JOIN ItemsRelations AS ir ON ir.ItemId = oi.Id
GROUP BY oi.OwnerId
OTHER TIPS
I think this should do the trick:
select i.OwnerId,
case when count(i.Id) > 0 then count(r.Id)/count(i.Id) else 0 end as Ratio
from Items i
left join ItemsRelations r on r.ItemId = i.Id
group by i.Id, i.OwnerId
Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow