SQL How to combine Timestamps to show a duration? - sql

Question

I have the following query where the data is the location of an ITEM at any given time in a STORAGE_BOX, where the location of the item can be specified further with a delimiter so you could have STORAGE_BOX/SLOT/TRAY/ROW_ID. This is why I do the SUBSTRING for that column.
What I'm trying to do: Create a view where to show what START_DATE and END_DATE a particular item was in a certain storage box.
select at.primary_key as 'ITEM_ID', f.name as 'INSTITUTION', SUBSTRING(sc.location +'/',0, CHARINDEX('/', sc.location + '/'))
as 'STORAGE_BOX', at.db_timestamp as 'TIMESTAMP' from [dbo].AUDIT_TRAIL at
RIGHT JOIN [dbo].BIOMATERIAL bio on at.primary_key = bio.id
LEFT JOIN [dbo].FACILITY f on bio.at_facility_id = f.id
LEFT JOIN [dbo].STORAGE_CONTAINER sc on at.primary_key = sc.id
where at.table_name = 'Biomaterial' AND sc.location IS NOT NULL
So the following output from the above query
+---------+-------------+-------------+------------+
| ITEM_ID | INSTITUTION | STORAGE_BOX | TIMESTAMP |
+---------+-------------+-------------+------------+
| 1 | Building#1 | STORAGE_0 | 2012-03-25 |
| 1 | Building#1 | STORAGE_0 | 2013-12-25 |
| 1 | Building#1 | STORAGE_1 | 2015-03-25 |
| 2 | Building#2 | STORAGE_3 | 2012-03-25 |
| 2 | Building#2 | STORAGE_4 | 2013-03-25 |
| 2 | Building#2 | STORAGE_5 | 2015-03-25 |
+---------+-------------+-------------+------------+
And change it into the below result where the START_DATE is the first timestamp of the new STORAGE_BOX and the END_DATE is the next timestamp of whatever STORAGE_BOX is next or the current timestamp if it is still there.
I have no idea how to compute these fields in the above query to get it to show
+---------+-------------+-------------+------------+---------------------+
| ITEM_ID | INSTITUTION | STORAGE_BOX | START_DATE | END_DATE |
+---------+-------------+-------------+------------+---------------------+
| 1 | Building#1 | STORAGE_0 | 2012-03-25 | 2015-03-25 |
| 1 | Building#1 | STORAGE_1 | 2015-03-25 | {Current_TimeStamp} |
| 2 | Building#2 | STORAGE_3 | 2012-03-25 | 2013-03-25 |
| 2 | Building#2 | STORAGE_4 | 2013-03-25 | 2015-03-25 |
| 2 | Building#2 | STORAGE_5 | 2015-03-25 | {Current_TimeStamp} |
+---------+-------------+-------------+------------+---------------------+
EDIT
I used the answer provided by Gordon Linoff to create the following query with sql server 2008 limitations
with t as (
select at.transaction_uid,at.primary_key as BIOMATERIAL_ID, f.name as INSTITUTION,
at.new_value as FREEZER,
at.db_timestamp as TIMESTAMP
from [dbo].AUDIT_TRAIL at RIGHT JOIN
[dbo].BIOMATERIAL bio
on at.primary_key = bio.id LEFT JOIN
[dbo].FACILITY f
on bio.at_facility_id = f.id
where at.table_name = 'Biomaterial' AND at.column_name = 'container_id.location' AND at.new_value IS NOT NULL
),
t1 as (
select t.*,
row_number() over (partition by BIOMATERIAL_ID, INSTITUTION, FREEZER
order by timestamp) as seqnum
from t
),
t2 as(
select t1.*,
ROW_NUMBER() over (partition by BIOMATERIAL_ID order by seqnum) as seqnum_b
from t1
where t1.seqnum = 1
)
SELECT a.BIOMATERIAL_ID, a.INSTITUTION, a.FREEZER, a.TIMESTAMP as START_DATE,coalesce(b.TIMESTAMP, getdate()) as END_DATE
FROM t2 a left join t2 b on a.BIOMATERIAL_ID = b.BIOMATERIAL_ID AND a.seqnum_b = (b.seqnum_b + 1) order by a.BIOMATERIAL_ID

Answer 1

You can do this using window functions. First use row_number() to get just the first row for each group:
with t as (
select at.primary_key as ITEM_ID, f.name as INSTITUTION,
SUBSTRING(sc.location +'/',0, CHARINDEX('/', sc.location + '/')) as STORAGE_BOX,
at.db_timestamp as TIMESTAMP
from [dbo].AUDIT_TRAIL at RIGHT JOIN
[dbo].BIOMATERIAL bio
on at.primary_key = bio.id LEFT JOIN
[dbo].FACILITY f
on bio.at_facility_id = f.id LEFT JOIN
[dbo].STORAGE_CONTAINER sc
on at.primary_key = sc.id
where at.table_name = 'Biomaterial' AND sc.location IS NOT NULL
),
t1 as (
select t.*,
row_number() over (partition by ITEM_ID, INSTITUTION, STORAGE_BOX
order by timestamp) as seqnum
from t
),
t2 as (
select t1.*, lead(timestamp) over (partition by item_id, institution order by timestamp) as next_timestamp
from t1
where seqnum = 1
)
select t2.ITEM_ID, t2.INSTITUTION, t2.STORAGE_BOX,
t2.timstamp as START_DATE,
coalesce(t2.next_timestamp, getdate()) as END_DATE
from t2 ;
The first CTE is your query. The second enumerates the rows for each item, institution, and storage box to eliminate duplicates. This appears to be the logic for your query, although if a storage box is used twice for the same item/location somewhat more complicated logic may be necessary.
The third CTE, t2 gets the next timestamp. And the final query applies the logic.
This assumes SQL Server 2012+ (based on your syntax I'm assuming SQL Server). You can do something similar with outer apply in earlier versions.