On Wed, 11 Feb 2015, Dimitri Maziuk wrote:
Sounds like you might have missed the point: job myjob1 myjob.sub vars myjob1 outfile=myjob1.out stdout=myjob1.stdout job myjob2 myjob.sub vars myjob2 outfile=myjob2.out stdout=myjob2.stdout is the condor_submit_dag file format. When I need to run a batch and give the output files sensible names -- which is what the original complaint was IIRC -- I submit is as a DAG specifically because of that. There may be downsides to it, but the basic functionality already exists: you can specify command line arguments, filenames, whatever. It just takes a second submit file and the _sumbit_dag rather than plain _submit.
Yeah, that makes sense. I was just saying that's not what I had in mind; I was thinking more along the lines of Todd's 2nd proposal, to queue based on each line of input (from stdin or a file) and populate $(input_line) rather than a queue N and $(process).
I get why you might want to do it in a dag file format as you described, just it seemed like you could accomplish the same thing with
queue for jobname,outfile,stdout in listfile.txt and listfile.txt containing: myjob1 myjob1.out myjob1.stdout myjob2 myjob2.out myjob2.stdout ... which strikes me as easier both to generate and to parse.
And then the thread diverged into the interesting ideas like "I want a globbing syntax that would do the right thing for users who can't glob" and "condor submit doesn't scale, that's bad unless it's a dag in which case it's ok" etc...
Yeah :( (Sorry for the noise!) Carl