Thread: highest match in group

have a data table that records entries by date(unix timestamp) and customer
number.  each custnum will have several entries showing a running ledger type
snapshot.  we have the need to get the most recent entry from not one, but all
unique customers, in the most cost effective manner.

        Table "summary"
     Attribute |  Type   | Modifier
    -----------+---------+----------
     custnum   | integer |
     date      | integer |
     amount    | float8  |
     balance   | float8  |

sample data;
     custnum  |    date    |  amount | balance
    ----------+------------+---------+---------
     12025702 | 1019151676 |   47.96 |       0
     12045401 | 1019145600 |   17.12 |  -17.12
     12040601 | 1019229292 |    26.7 |    1.02
     12045701 | 1019232000 |   16.59 |  -16.59
     12045702 | 1019232000 |   16.59 |  -16.59
     12045703 | 1019232000 |    9.87 |   -9.87
     12045704 | 1019232000 |   16.59 |  -16.59
     12045705 | 1019232000 |   16.59 |  -16.59
     12045704 | 1019408919 |   15.52 |   -1.07
     12045704 | 1019404800 |   15.52 |  -16.59

Currently we are running through all our customer numbers in one query, then
for each customer number querying the summary table to get each customers
latest entry (select order by date desc limit 1). Obviously this results in a
large number of queries and is expensive.  Looking for a more concise, less
expensive way.

thanks

Dave

Dave [Hawk-Systems] wrote:

> have a data table that records entries by date(unix timestamp) and customer
> number.  each custnum will have several entries showing a running ledger type
> snapshot.  we have the need to get the most recent entry from not one, but all
> unique customers, in the most cost effective manner.
>
> [snip]
>
> Currently we are running through all our customer numbers in one query, then
> for each customer number querying the summary table to get each customers
> latest entry (select order by date desc limit 1). Obviously this results in a
> large number of queries and is expensive.  Looking for a more concise, less
> expensive way.
>
> thanks
>
> Dave
What about something like:
SELECT so.* FROM summary so, (SELECT custnum, MAX(date) as date FROM
summary si GROUP BY custnum) as cd WHERE so.date = cd.date AND
so.custnum = cd.custnum

Best regards,

Arjen van der Meijden

"Dave [Hawk-Systems]" <dave@hawk-systems.com> writes:
> have a data table that records entries by date(unix timestamp) and customer
> number.  each custnum will have several entries showing a running ledger type
> snapshot.  we have the need to get the most recent entry from not one, but all
> unique customers, in the most cost effective manner.

If you don't mind using a Postgres-only construct, SELECT DISTINCT ON
is made for this.  See the "weather report" example in the SELECT
reference page.

            regards, tom lane

Is it possible to update multiple columns of a table using a select
statement to derive the values?

For example (trying to port from Oracle to Postgres)

-- update the image record for original size image
update WPImage set (WPImageStateID, Width, Height, ContentType,
ContentLength) = (
  select 3, Width, Height, ContentType, ContentLength
  from WPImageHeader
  where WDResourceID = pResourceID
)
where WDResourceID = pResourceID and WPSizeTypeID = 0;

I have seen mention of a Postgres (specific) feature,

   update ... set .... from {other_table} where {join_condition}


Is this the most appropriate way to do the above in postgres?

update WPImage
set WPImageStateID =  3,
Width = WPImageHeader.Width,
Height = WPImageHeader.Height,
ContentType = WPImageHeader.ContentType,
ContentLength = WPImageHeader.ContentLength
where WPImage.WDResourceID = WPImageHeader.WDResourceID
and WPImage.WDResourceID = pResourceID
and WPImage.WPSizeTypeID = 0;

{where pResourceID is a variable}

Thanks

John Sidney-Woollett

>"Dave [Hawk-Systems]" <dave@hawk-systems.com> writes:
>> have a data table that records entries by date(unix timestamp) and customer
>> number.  each custnum will have several entries showing a running
>ledger type
>> snapshot.  we have the need to get the most recent entry from not
>one, but all
>> unique customers, in the most cost effective manner.
>
>If you don't mind using a Postgres-only construct, SELECT DISTINCT ON
>is made for this.  See the "weather report" example in the SELECT
>reference page.

thanks Tom, had a forest for the trees problem there and that put me on the
right track.

Dave