realizations not in plan are executed anyway?

4 views
Skip to first unread message

Nathan Schneider

unread,
Dec 14, 2013, 4:27:09 PM12/14/13
to ducttap...@googlegroups.com

Hi,

I’m using ducttape-0.3 with a modified version of the cdec tutorial, and am noticing unexpected behavior with plans.

Right now I have a plan like

plan P {
  reach extract_gra via (ExtractSection: dev)
}

and am invoking it via the command line: ducttape workflow.tape -p P. Strangely, it says it will run extract_gra with ExtractSection:dev (and its prerequisites), as well as a realization which I have not requested: ExtractSection:devtest (the other branch of ExtractSection). Am I misunderstanding how plans are supposed to work, or is that behavior unexpected?

The inputs to extract_gra are:

 < corpus=(DataSeg: plain=(ExtractSection: dev=$out@preproc_corpus[DataSection:dev] devtest=$out@preproc_corpus[DataSection:devtest]) mwe=(ExtractSection: dev=$mwetokenized@mwe_tokenize[DataSection:dev] devtest=$mwetokenized@mwe_tokenize[DataSection:devtest]))
 < ini=$ini@index_training

I ran ducttape mtlab.tape -p P viz to generate the attached graph; essentially I am wondering why the bottom left realization is ever reached under this plan.

Any insights?

Thanks,

Nathan

mtlab-viz.pdf

Greg Hanneman

unread,
Dec 16, 2013, 9:33:01 AM12/16/13
to Nathan Schneider, ducttap...@googlegroups.com
I'm afraid I'm no help, but the same thing is happening to me and others at Safaba in one of our pipelines.  We're similarly confused as to why the non-requested realization ever gets run when the plan doesn't select for it.  This is also with version 0.3.

Greg.



--
You received this message because you are subscribed to the Google Groups "ducttape-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ducttape-user...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

Jon Clark

unread,
Dec 16, 2013, 11:32:11 AM12/16/13
to Greg Hanneman, Nathan Schneider, ducttap...@googlegroups.com
Sounds like a bug.

Nathan: Can you open an issue on github and attach the description and full workflow? Bonus points if you can find a minimal workflow that repros.

Sent from my Windows Phone

From: Greg Hanneman
Sent: ‎12/‎16/‎2013 6:33 AM
To: Nathan Schneider
Cc: ducttap...@googlegroups.com
Subject: Re: realizations not in plan are executed anyway?

Nathan Schneider

unread,
Dec 16, 2013, 5:39:04 PM12/16/13
to Jon Clark, Greg Hanneman, ducttap...@googlegroups.com
Voilà—the behavior is pretty simple to describe: https://github.com/jhclark/ducttape/issues/162

Cheers,
Nathan


On Mon, Dec 16, 2013 at 11:32 AM, Jon Clark <jon.h...@gmail.com> wrote:
Sounds like a bug.

Nathan: Can you open an issue on github and attach the description and full workflow? Bonus points if you can find a minimal workflow that repros.

Sent from my Windows Phone

From: Greg Hanneman
Sent: 12/16/2013 6:33 AM

Jon Clark

unread,
Dec 16, 2013, 7:06:14 PM12/16/13
to Nathan Schneider, Greg Hanneman, ducttap...@googlegroups.com
Thanks Nathan. This should shave some time off my debug cycle. No ETA on a fix right now -- I want to wrap up a thesis chapter before diving into this.


Sent from my Windows Phone

From: Nathan Schneider
Sent: ‎12/‎16/‎2013 2:39 PM
To: Jon Clark
Cc: Greg Hanneman; ducttap...@googlegroups.com

Chris Dyer

unread,
Dec 16, 2013, 7:27:47 PM12/16/13
to Jon Clark, Nathan Schneider, Greg Hanneman, ducttap...@googlegroups.com
Like.

On Mon, Dec 16, 2013 at 7:06 PM, Jon Clark <jon.h...@gmail.com> wrote:
> Thanks Nathan. This should shave some time off my debug cycle. No ETA on a
> fix right now -- I want to wrap up a thesis chapter before diving into this.
>
>
> Sent from my Windows Phone
> ________________________________
> From: Nathan Schneider
> Sent: 12/16/2013 2:39 PM

Jonathan Clark

unread,
Feb 1, 2014, 8:16:46 PM2/1/14
to Chris Dyer, Nathan Schneider, Greg Hanneman, ducttap...@googlegroups.com
TACL paper done = Bug fixed: https://github.com/jhclark/ducttape/issues/162

Any volunteers to do 0.3.1 testing & release?
Reply all
Reply to author
Forward
0 new messages