Return latest transaction data, based upon transaction dates.

Question

I appreciate I'm being a little dense here, and I have searched, read, and tried out a few different solutions I've seen give around the place. However; I think I'm struggling more with the naming conventions and logic of other people queries than you might in understanding mine (here's hoping!)
I have a huge table, which contains a record for every transaction which has an effect on our inventory (yup - BIG table!)
For a given transaction type 'CHC' (CHange Costs) I want to return the Part code, Transaction Date and Transaction cost of the LAST TWO changes.
Because its to be used for tracking updates to the cost of materials, and further for calculating the ongoing effect of these, I just need the two for now.
So,
Table is I50F
Columns required are
I50PART - [Part Code|http://forums.oracle.com/forums/]
I50TDAT - [Transaction Date|http://forums.oracle.com/forums/]
I50UCOST - [Price changed to|http://forums.oracle.com/forums/]
I50TRNS - [Transaction Type - we just want CHC]
Sample Data (Including just columns we are interested in)
I50PART              I50TDAT             I50UCOST         I50TRNS
BACCA001             08/03/2006 07:34:51 0.08829          CHC    
BACCA001             25/07/2007 08:26:30 0.10329          CHC    
BACCA001             10/04/2008 16:29:02 0.10639          CHC    
BACCA003             20/06/2006 12:22:30 0.16814          CHC    
BACCA003             25/07/2007 08:26:54 0.17024          CHC    
BACCA003             10/04/2008 13:30:12 0.17535          CHC    
BACCA004             28/08/2007 15:46:03 0.06486          CHC    
BACCA004             28/08/2007 15:49:15 0.06328          CHC    
BACCA004             30/10/2008 09:22:40 0.06952          CHC    
BACCA004             13/01/2009 09:09:07 0.06867          CHC    
BACCA005             25/07/2007 08:27:24 0.06715          CHC    
BACCA005             10/04/2008 15:45:14 0.06916          CHC    
BACCA005             30/10/2008 09:05:17 0.07453          CHC    
BACCA005             13/01/2009 09:06:49 0.07275          CHC     To take a part in isolation, BACCA005;
I'm interested in the last two records.
It makes sense for there to be two records output per part at this stage, as it may be that the powers that be decide that they want the last 3, or 4, or whatever (I'm sure everybody has similar experiences with beancouters)
Is it A) Easy, and B) relatively efficient. There are 2.4m records in the table.
If I've been stupid and not included enough info, please do [metaphorically] poke me in the eye, and I'll pad out a bit.
Thanks ever so much for reading - and even more so if you can help!
Cheers
J

Answer 1

Analytic functions FTW!
with I50F as (select 'BACCA001' I50PART, to_date('08/03/2006 07:34:51', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.08829 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA001' I50PART, to_date('25/07/2007 08:26:30', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.10329 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA001' I50PART, to_date('10/04/2008 16:29:02', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.10639 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA003' I50PART, to_date('20/06/2006 12:22:30', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.16814 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA003' I50PART, to_date('25/07/2007 08:26:54', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.17024 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA003' I50PART, to_date('10/04/2008 13:30:12', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.17535 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA004' I50PART, to_date('28/08/2007 15:46:03', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.06486 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA004' I50PART, to_date('28/08/2007 15:49:15', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.06328 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA004' I50PART, to_date('30/10/2008 09:22:40', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.06952 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA004' I50PART, to_date('13/01/2009 09:09:07', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.06867 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA005' I50PART, to_date('25/07/2007 08:27:24', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.06715 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA005' I50PART, to_date('10/04/2008 15:45:14', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.06916 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA005' I50PART, to_date('30/10/2008 09:05:17', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.07453 I50UCOST, 'CHC' I50TRNS from dual union all
              select 'BACCA005' I50PART, to_date('13/01/2009 09:06:49', 'dd/mm/yyyy hh24:mi:ss') I50TDAT, 0.07275 I50UCOST, 'CHC' I50TRNS from dual)
select I50PART, I50TDAT, I50UCOST, I50TRNS
from   (select I50PART, I50TDAT, I50UCOST, I50TRNS, row_number() over (partition by I50PART order by I50TDAT desc) rn
        from   I50F
        where  I50TRNS = 'CHC')
where  rn <= 2
order by I50PART, I50TDAT desc;
I50PART  I50TDAT               I50UCOST I50
BACCA001 10/04/2008 16:29:02     .10639 CHC
BACCA001 25/07/2007 08:26:30     .10329 CHC
BACCA003 10/04/2008 13:30:12     .17535 CHC
BACCA003 25/07/2007 08:26:54     .17024 CHC
BACCA004 13/01/2009 09:09:07     .06867 CHC
BACCA004 30/10/2008 09:22:40     .06952 CHC
BACCA005 13/01/2009 09:06:49     .07275 CHC
BACCA005 30/10/2008 09:05:17     .07453 CHC