| Author |
Message
|
| sriram.e |
 Posted: Wed May 12, 2010 1:53 pm Post subject: Collector node help in MB7 |
 |
|
Novice
Joined: 14 Nov 2006
Posts: 12
|
Hi,
MB7 on AIX 5.3
Here is sample input file with fixed format of 8500 bytes,
| Code: |
R1 NAME1 000101 --|
R2 DEPART XXX |
R3 PROJECT1 010 | 1 inventory
R3 PROJECT2 020 |
R3 PROJECT3 030 --|
R1 NAME2 000101 --|
R2 DEPART YYY |
R3 PROJECT1 010 | 2 inventory
R3 PROJECT2 020 --|
R1 NAME3 000101 --|
R2 DEPART ZZZ |
R3 PROJECT1 040 | 3 inventory
R3 PROJECT2 020 --|
|
Expected output should be :
| Code: |
R1 NAME1 000101 --|
R2 DEPART XXX |
R3 PROJECT1 010 | 1 inventory message
R3 PROJECT2 020 |
R3 PROJECT3 030 --|
R1 NAME2 000101 --|
R2 DEPART YYY |
R3 PROJECT1 010 | 2 inventory message
R3 PROJECT2 020 --|
R1 NAME3 000101 --|
R2 DEPART ZZZ |
R3 PROJECT1 040 | 3 inventory message
R3 PROJECT2 020 --|
|
Basically, I am planning to read file record by record, accumulating in collector node till next R1 record found, propagating previous saved records in one unit of work as a message.
Continue the above process until end of file
I wanted to know,
Is it possible to accomplish with collector node?
Are there any other option? (not using DB)
Is it doable using share variables?
Any help will be appreciated!! |
|
| Back to top |
|
 |
| sridhsri |
| Posted: Wed May 12, 2010 2:25 pm Post subject: |
|
|
Master
Joined: 19 Jun 2008
Posts: 297
|
| I don't see a need for collector nodes (and I don't think it makes sense to use one either). This is a message transformation problem - you'll need two message types (one for source format and one for destination format). Use a compute node and transform it. No collector nodes, shared variables or DB. |
|
| Back to top |
|
|
| mqjeff |
| Posted: Thu May 13, 2010 4:21 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
The reason that sriram.e thinks that a collector node is used is because sriram.e thinks that each *line* of the file will be produced as output from FileInput.
This is indeed one way to do it.
Using a proper message model that identifies the whole "message" as a "record" and telling FileInput to output each message as a "record" is a much easier solution. |
|
| Back to top |
|
|
| sridhsri |
| Posted: Thu May 13, 2010 5:32 am Post subject: |
|
|
Master
Joined: 19 Jun 2008
Posts: 297
|
| I'm a little rusty with collector nodes. But don't you have to know how many events you want to "collect" at development time. I don't think there is a way to know that.... |
|
| Back to top |
|
|
| mqjeff |
| Posted: Thu May 13, 2010 5:39 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
| You'd have to have some way to recognize the "last" record, then you could forward it to a separate input terminal that would close the collection. |
|
| Back to top |
|
|
| sriram.e |
| Posted: Mon May 17, 2010 9:11 am Post subject: |
|
|
Novice
Joined: 14 Nov 2006
Posts: 12
|
| Thanks for all your quick responses, i think with a combination of fileInput node and file output node, message model, i can accomplish this, but i'm just giving a try. But though i want to know whether batch accumalation is possible in collector node. |
|
| Back to top |
|
|
|
|
