Reputation: 655
I have a logging table collecting values from many probes:
CREATE TABLE [Log]
(
[LogID] int IDENTITY (1, 1) NOT NULL,
[Minute] datetime NOT NULL,
[ProbeID] int NOT NULL DEFAULT 0,
[Value] FLOAT(24) NOT NULL DEFAULT 0.0,
CONSTRAINT Log_PK PRIMARY KEY([LogID])
)
GO
CREATE INDEX [Minute_ProbeID_Value] ON [Log]([Minute], [ProbeID], [Value])
GO
Typically, each probe generates a value every minute or so. Some example output:
LogID Minute ProbeID Value
====== ================ ======= =====
873875 2014-07-27 09:36 1972 24.4
873876 2014-07-27 09:36 2001 29.7
873877 2014-07-27 09:36 3781 19.8
873878 2014-07-27 09:36 1963 25.6
873879 2014-07-27 09:36 2002 22.9
873880 2014-07-27 09:36 1959 -30.1
873881 2014-07-27 09:36 2005 20.7
873882 2014-07-27 09:36 1234 23.8
873883 2014-07-27 09:36 1970 19.9
873884 2014-07-27 09:36 1991 22.4
873885 2014-07-27 09:37 1958 1.7
873886 2014-07-27 09:37 1962 21.3
873887 2014-07-27 09:37 1020 23.1
873888 2014-07-27 09:38 1972 24.1
873889 2014-07-27 09:38 3781 20.1
873890 2014-07-27 09:38 2001 30
873891 2014-07-27 09:38 2002 23.4
873892 2014-07-27 09:38 1963 26
873893 2014-07-27 09:38 2005 20.8
873894 2014-07-27 09:38 1234 23.7
873895 2014-07-27 09:38 1970 19.8
873896 2014-07-27 09:38 1991 22.7
873897 2014-07-27 09:39 1958 1.4
873898 2014-07-27 09:39 1962 22.1
873899 2014-07-27 09:39 1020 23.1
What is the most efficient way to get just the latest reading for each Probe?
e.g.of desired output (note: the "Value" is not e.g. a Max() or an Avg()):
LogID Minute ProbeID Value
====== ================= ======= =====
873899 27-Jul-2014 09:39 1020 3.1
873894 27-Jul-2014 09:38 1234 23.7
873897 27-Jul-2014 09:39 1958 1.4
873880 27-Jul-2014 09:36 1959 -30.1
873898 27-Jul-2014 09:39 1962 22.1
873892 27-Jul-2014 09:38 1963 26
873895 27-Jul-2014 09:38 1970 19.8
873888 27-Jul-2014 09:38 1972 24.1
873896 27-Jul-2014 09:38 1991 22.7
873890 27-Jul-2014 09:38 2001 30
873891 27-Jul-2014 09:38 2002 23.4
873893 27-Jul-2014 09:38 2005 20.8
873889 27-Jul-2014 09:38 3781 20.1
Upvotes: 3
Views: 84
Reputation: 13248
This is another approach
select *
from log l
where minute =
(select max(x.minute) from log x where x.probeid = l.probeid)
You can compare the execution plan w/ a fiddle - http://sqlfiddle.com/#!3/1d3ff/3/0
Upvotes: 2
Reputation: 1271141
Your question is: "What is the most efficient way to get just the latest reading for each Probe?"
To really answer this question, you test to test different solutions. I would generally go with the row_number()
method suggested by @jyparask. However, the following might have better performance:
select l.*
from log l
where not exists (select 1
from log l2
where l2.probeid = l.probeid and
l2.minute > l.minute
);
For performance, you want an index on log(probeid, minute)
.
Although not exactly your problem, here is an example of where not exists
performs better than other methods on SQL Server.
Upvotes: 1
Reputation: 4137
Try this:
SELECT T1.*
FROM Log T1
INNER JOIN (SELECT Max(Minute) Minute,
ProbeID
FROM Log
GROUP BY ProbeID)T2
ON T1.ProbeID = T2.ProbeID
AND T1.Minute = T2.Minute
You can play around with it on SQL Fiddle
Upvotes: 1
Reputation: 18431
;WITH MyCTE AS
(
SELECT LogID,
Minute,
ProbeID,
Value,
ROW_NUMBER() OVER(PARTITION BY ProbeID ORDER BY Minute DESC) AS rn
FROM LOG
)
SELECT LogID,
Minute,
ProbeID,
Value
FROM MyCTE
WHERE rn = 1
Upvotes: 0