can't get "enrich" to work

[julianc]

Hi,

I just started playing with this stuff and I'm trying to duplicate some legacy functionality.

We have a system which exports xml files via ftp and when a file is complete it creates another file with the same name but a .mrk extension.

So it seems to me that I could have a file consumer filtered on the mrk file and then enrich it with the xml file. Without the pollEnrich I can get the mrk file copied to the output but with the pollEnrich nothing happens. Using the latest windows binaries.

Here is the xml for my route:

    <pollEnrich timeout="0" uri="file://D:\Downloads?fileName=$.xml" inheritErrorHandler="true" id="enrich2"/>

I get a java exception saying Camel doesn't currently support pollEnrich where both consumers are files.

Any ideas about alternative approaches?

Julian

Edited by: julianc on Sep 24, 2010 3:31 AM

Edited by: julianc on Sep 24, 2010 3:36 AM

[davsclaus]

Yeah as the exception says you can use pollEnrich with files when the from was file based as well. This has just recently been implement in the next Camel 2.5 release.

Instead of doing like that you can instead just use a custom filter on the file endpoint. And then check if there is a .mrk file for the given file.

http://camel.apache.org/file2

You just implement a GenericFiler and then check if the .mrk file exists and return either true or false. Much easier.

[julianc]

Thanks, that sounds like a plan.

I'll try enrich again in 2.5. (Not necessarily for this use case, but to understand).

Julian

[julianc]

OK, I tried the filter method and it works. I also tried the processStrategy idea and that works too, after a fashion.

I can't get the processed file moved in the strategy case - its like its ignoring the "super" calls (java attached).

But both seem supremely inefficient - my legacy code polls for "ready files" and only looks at the "real" files when there is a ready file. This is more like the "pollEnrich" method.

I'm thinking a better approach is to translate the ready file presence into a claim check and route this through to the grunt process which will do the ETL (ideally on aggregated sets of files).

[davsclaus]

You can also do the reverse. Configure the file endpoint to only accept the .mrk files.

Then use a Processor to set the message body to the real file, and then add an on completion with logic to delete/move etc. the real file when the routing is done.

http://camel.apache.org/oncompletion.html

Exchange has API to add completion directly to it.

For example the Camel in Action book shows this in chapter 9, section 9.6

[davsclaus]

I added a ticket to Apache to see if we could add support for marker files out of the box. Fell free to comment and feedback.

https://issues.apache.org/activemq/browse/CAMEL-3163

[jessicaleahy]

I already tried the reverse. so far, it worked out great for me.

[julianc]

That would be ideal. I've voted for it.

Thanks

Julian

[julianc]

Cool - any chance you could post your solution? (on a  steep learning curve here).

cheers

Julian