asked on

Can this be done without a subquery?

I need to select the latest result from a table.

table person:
id----------name
1-----------a
2-----------b

table result:
id---------person_id-----------result--------date
1----------1-----------------------NULL---------2014-01-01
2----------1-----------------------1---------------2014-01-02
3----------2-----------------------2---------------2014-01-03

Desired outcome:
name-------------result------------date
a-------------------1-------------------2014-01-02
b-------------------2-------------------2014-01-03

Doing a simple GROUP BY does not work, obviously:

SELECT person.id, name, result, MAX(date)
FROM person LEFT JOIN result ON person.id = result.person_id
GROUP BY person.id

will result in the following wrong outcome
name-------------result------------date
a-------------------NULL-------------2014-01-02
b-------------------2-------------------2014-01-03

I can get the correct result using a subquery where I sort by date before using MAX:

SELECT id, name, result, MAX(date) FROM (
SELECT person.id, name, result, date
FROM person LEFT JOIN result ON person.id = result.person_id
ORDER BY person.id, date DESC
) AS z
GROUP BY id

How can I do this without using a subquery?

Thank you.

SOLUTION

Phillip Burton

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

Randy Poole

When will result be null? Do you just want to omit all records with a result of null?

Dan Craciun

ASKER

I don't have 2 results from the same day for any person. The events are once or twice a year for a person.

Yes, I only need non-NULL results, but I can filter those out in the JOIN clause.

Randy Poole

Can you give a larger amount of test data, and also let us know why you don't want to use a sub query??

Dan Craciun

ASKER

I try to avoid subqueries on large tables, because they tend to bog down MySQL. On the production db this query takes about 4 seconds.
I guess in this case it's OK, since I only do a GROUP BY on the outside, but I thought there might be a more straightforward way to return the values from the row with the max date.

Let's see if I can get a sample that better approximates the live data:

table result:
id---------person_id-----------result--------date
1----------1-----------------------NULL---------2014-01-01
2----------1-----------------------1---------------2014-01-02
3----------1-----------------------2---------------2014-01-03
4----------1-----------------------NULL---------2014-01-04

I would need the query to return this:
id------------name-------------result------------date
1-------------a-------------------2-------------------2014-01-03

ASKER CERTIFIED SOLUTION

PortletPaul

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

SOLUTION

PortletPaul

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

Dan Craciun

ASKER

Thank you Paul. I mainly use MySQL and I got used to specify only one column in the GROUP BY.
It's easier to read (by me): oh, this query will return a single row for each id.
It hasn't got me into trouble yet, but I agree it's best practice to specify all the non-aggregating fields in the GROUP BY clause.

Returning to the query, I don't need the max result and the max date. I need the latest result.

So if I have this:
table result:
id---------person_id-----------result--------date
1----------1-----------------------NULL---------2014-01-01
2----------1-----------------------3---------------2014-01-02
3----------1-----------------------2---------------2014-01-03
4----------1-----------------------NULL---------2014-01-04

Your query will return
id------------name-------------result------------date
1-------------a-------------------3-------------------2014-01-03

PortletPaul

For "latest record", in MySQL, you simply cannot avoid subquery(ies) as far as I know. And you either need a costly self-join, or an ordered subquery using variable calculations. That's because MySQL lacks some really useful windowing functions like row_number().

or, you make do with the results of max() or similar.

    | ID | NAME | MAX(R.RESULT) |                    MAX(R.DATE) |
    |----|------|---------------|--------------------------------|
    |  1 |    a |             2 | January, 03 2014 00:00:00+0000 |

Open in new window

which is it to be?

--------
do please specify all the fields of a group by
what happens if someone turns off that "extension"?
(lots of wrong queries...ooops)
it is indeed best practice and one you really should follow

PortletPaul

sorry, you have changed the sample data yet again, and that sample/expected result no longer lines-up with the concept of "latest"

could I clarify please.

Most recent sample/expected result
table result:
id---------person_id-----------result--------date
1----------1-----------------------NULL---------2014-01-01
2----------1-----------------------3---------------2014-01-02 <<-- this is a new row or at least new result value
3----------1-----------------------2---------------2014-01-03 <<lastest record, so result = 2???
4----------1-----------------------NULL---------2014-01-04

Your query will return
id------------name-------------result------------date
1-------------a-------------------3------------------2014-01-03 <<-- maximum result = 3 date = 2014-01-03

I'm now very confused.

Could we work with just one set of data please?

Dan Craciun

ASKER

:) sorry if I confused you.

I was only pointing out that
- if I use your second query and
- the latest result was not the biggest (so on 2014-01-03 the result -2- was smaller than an older result) then
- the query will no longer return the latest result

PS: yeah, I know I chose a poor name for that column; in the original data set it was lab_cd

Dan Craciun

ASKER

In short, the answer is : NO.

Thank you all.

PortletPaul

OK, got it. You do want the "latest" - apologies; I was like a rabbit in headlights for a moment

PortletPaul

@Dan

humble humble apologies.... try this

SELECT
       p.id
     , p.name
     , r.result
     , r.date
FROM person p
INNER JOIN result_1 r ON p.id = r.person_id
LEFT JOIN result_1 r2 ON r.person_id = r2.person_id
                      AND r2.id > r.id
WHERE r.result IS NOT NULL
and r2.id is null

Open in new window

It produces this outcome:

| ID | NAME | RESULT |                           DATE |
|----|------|--------|--------------------------------|
|  1 |    a |      1 | January, 02 2014 00:00:00+0000 |
|  2 |    b |      2 | January, 03 2014 00:00:00+0000 |
|  3 |  ccc |      2 |   April, 04 2014 00:00:00+0000 |

Open in new window

from this test data

CREATE TABLE Person
	(`id` int, `name` varchar(20))
;
	
INSERT INTO Person
	(`id`, `name`)
VALUES
	(1, 'a'),
	(2, 'b'),
	(3, 'ccc')
;

CREATE TABLE Result_1
	(`id` int, `person_id` int, `result` varchar(4), `date` datetime)
;
	
INSERT INTO Result_1
	(`id`, `person_id`, `result`, `date`)
VALUES
	(1, 1, NULL, '2014-01-01 00:00:00'),
	(2, 1, '1', '2014-01-02 00:00:00'),
	(3, 2, '2', '2014-01-03 00:00:00'),

    (31, 3, NULL, '2011-01-01 00:00:00'),
    (32, 3, '1', '2012-01-01 00:00:00'),
    (33, 3, '3', '2013-03-03 00:00:00'),
    (34, 3, '2', '2014-04-04 00:00:00')
;

Open in new window

So it does (I believe) meet your criteria of no subquery, but that self join may be "expensive" (but perhaps not as bad as a subquery)

{+edit}
this also works:

SELECT
       p.id
     , p.name
     , r.result
     , r.date
FROM person p
INNER JOIN result_1 r ON p.id = r.person_id
LEFT JOIN result_1 r2 ON r.person_id = r2.person_id
                      AND r2.`date` > r.`date`
WHERE r.result IS NOT NULL
and r2.id is null
;

Open in new window