Counting rows until where clause condition is satisfied - sql

Question

I have a table of data which contains attributes like body, offer_id and created_at. When in chronological order I need to find the count of rows until 'body' satisfies my 'where' clause for a particular offer_id, i.e.
created at | offer id | body
---------------------------------------------
Jan | 12 | does not satisfy
Feb | 12 | does not satisfy
Mar | 12 | satisfies
Jan | 13 | does not satisfy
Feb | 13 | satisfies
Jan | 14 | does not satisfy
Feb | 14 | satisfies
Mar | 14 | does not satisfy
Apr | 14 | does not satisfy
Expected output:
offer_id | count
---------|------
12 | 3
13 | 2
14 | 2

Answer 1

First - you need to generate a number for every record inside its offer window:
select t.*, rownumber() over (partition by t.offer_ID order by t.created_at) as rn
from t
it will result in something like:
created at | offer id | body | rn
---------------------------------------------
Jan | 12 | does not satisfy | 1
Feb | 12 | does not satisfy | 2
Mar | 12 | satisfies | 3
Jan | 13 | does not satisfy | 1
Feb | 13 | satisfies | 2
Jan | 14 | does not satisfy | 1
Feb | 14 | satisfies | 2
Mar | 14 | does not satisfy | 3
Apr | 14 | does not satisfy | 4
from this subquery you can get a minimal rn (first record that satisfies the condition):
with sub as (
select t.*, rownumber() over (partition by t.offer_ID order by t.created_at) as rn
from t)
select offer_ID, min(rn)
from sub
where (satisfies)
group by offer_ID

Answer 2

straight as an arrow
select t.offer_id, count(*)
from mytable t
where not exists
(
select 1 from mytable tt
where tt.offer_id = t.offer_id
and tt.created_at < t.created_at
and tt.body = 'satisfies'
)
group by t.offer_id

Answer 3

DO you have tried something like this?
select count(*)
from mytable
where "satisfies"
Or, if you want to count only the different offer_id:
select count(distinct offer_id)
from mytable
where "satisfies"
Or, finally:
select count(offer_id)
from mytable
where "satisfies"
group by offer_id
Does is this what you need? If not, give me more details! ;)

Answer 4

One way to count the number that don't satisfy the condition is to use a cumulative sum:
select offer_id, count(*)
from (select t.*,
sum(case when <condition> then 1 else 0 end) over
(partition by offer_id order by created_at) as num
from t
) t
where num = 0;
However, this is one less than the number you have. So, instead:
select offer_id,
(sum(case when num = 0 then 1 else 0 end) +
max(case when num = 1 then 1 else 0 end)
)
from (select t.*,
sum(case when <condition> then 1 else 0 end) over
(partition by offer_id order by created_at) as num
from t
) t
where num in (0, 1)

Answer 5

If you just want the count of offer_id , you can use the below
select offer_id, count(*) as count_1 from table_name
where <<your condition>>
group by offer_id
If my understanding is wrong, please share a detailed description on what exactly you require.

Answer 6

You can break the task in two parts:
For each offer ID find the record/date that first satisfies the condition.
Count all records per offer ID until that found record/date.
With a subquery in SELECT:
select
offer_id,
(
select count(*)
from mytable m
where m.offer_id = mfit.offer_id
and m.created_at <= min(mfit.created_at)
) as cnt
from mytable mfit
where <condition>
group by offer_id
or a subquery in FROM:
select
mfit.offer_id,
count(*) as cnt
from
(
select offer_id, min(created_at) as min_date
from mytable
where <condition>
group by offer_id
) mfit
join mytable m on m.offer_id = mfit.offer_id and m.created_at <= mfit.created_at
group by mfit.offer_id;

Answer 7

Here is another query using an analytic function. Analytic functions have the advantage that you read the table just once and get different aggregations on-the-fly. The idea is to have a running total per offer_id with a one for a record matching your condition plus a count per offer_id. This looks as follows:
created at | offer id | body | s | c
---------------------------------------------------
Jan | 13 | does not satisfy | 0 | 1
Feb | 13 | satisfies | 1 | 2
Jan | 14 | does not satisfy | 0 | 1
Feb | 14 | satisfies | 1 | 2
Mar | 14 | does not satisfy | 1 | 3
Apr | 14 | does not satisfy | 1 | 4
May | 14 | satisfies | 2 | 5
Jun | 14 | does not satisfy | 2 | 6
Apr | 14 | does not satisfy | 2 | 7
May | 14 | satisfies | 3 | 8
So we are simply looking for the min(c) for s = 1.
select offer_id, min(c) as cnt
from
(
select
offer_id,
sum(case when <condition> then 1 else 0 end)
over (partition by offer_id order by created_at) as s,
count(*) over (partition by offer_id order by created_at) as c
from mytable
) data
where s = 1
group by offer_id
order by offer_id;