On Tue, 3 Jun 2014 14:49:49 +0200
Jan Blechta <[email protected]> wrote:

> On Tue, 3 Jun 2014 14:34:19 +0200
> Anders Logg <[email protected]> wrote:
> 
> > On Tue, Jun 03, 2014 at 02:25:02PM +0200, Jan Blechta wrote:
> > > On Tue, 3 Jun 2014 13:50:04 +0300
> > > Päivo Simson <[email protected]> wrote:
> > >
> > > > Hi!
> > > >
> > > > Is there a way to disable console output for
> > > > interpolate_nonmatching_mesh? This happens when the point is
> > > > outside the original mesh. The output looks like this:
> > > > ...
> > > > <Point x = -7 y = 17.5 z = 0>
> > > > <Point x = -7 y = 10.5 z = 0>
> > > > <Point x = 7 y = 17.5 z = 0>
> > > > <Point x = -7 y = 3.5 z = 0>
> > > > <Point x = 21 y = 17.5 z = 0>
> > > > ...
> > > > I could not find anywhere in the source code where these
> > > > messages could be disabled. Im using fenicstools-1.3.0
> > >
> > > This is probably because of
> > >
> > >    384       id =
> > > mesh.bounding_box_tree()->compute_closest_entity(point).first;
> > > 385       cout << "Extrapolating function value at x = " << point
> > > 386            << " (not inside domain)." << endl;
> > >
> > > in dolfin/function/Function.cpp which happens when extrapolating.
> > > This is really annoying and flooding stdout and should be fixed.
> > > AFAIK this
> > 
> > Agree, I can fix this.
> > 
> > Suggestions for how? Possibilities:
> > 
> > 1. Send it through the log system at a lower log level (like
> > PROGRESS). Then it can be seen if someone wants to see it and turned
> > off otherwise, but sending those messages through the log system may
> >    take a long time for many points...
> 
> If you would log less than PROGRESS you would get rid of many
> important information, like convergence of Newton.

Ok, making it TRACE would not be a bad idea. Still no printing is
faster.

Jan

> 
> > 
> > 2. Print the points only when some parameter is set.
> 
> The most flexible but floods the codebase.
> 
> > 
> > 3. Just not print anything.
> 
> Possibly. Does anybody need this information? The information printed
> is sometimes useful for debugging when extrapolation is undesirable
> but this debugging can be definitely done using more clever ways,
> like mesh entity iterators and methods from bounding box tree. So the
> printing should be avoided in intentionally extrapolating production
> code. So I pick this option.
> 
> Jan
> 
> > 
> > --
> > Anders
> > 
> > 
> > 
> > > can only be muted using
> > >
> > >   set_log_active(false);
> > >
> > > before executing some extrapolating code and afterwards reverted
> > > by
> > >
> > >   set_log_active(true);
> > >
> > > Jan
> > >
> > > >
> > > > Any help is much appreciated.
> > > >
> > > > Best regards,
> > > > Päivo
> > >
> > > _______________________________________________
> > > fenics-support mailing list
> > > [email protected]
> > > http://fenicsproject.org/mailman/listinfo/fenics-support
> > _______________________________________________
> > fenics-support mailing list
> > [email protected]
> > http://fenicsproject.org/mailman/listinfo/fenics-support
> 
> _______________________________________________
> fenics-support mailing list
> [email protected]
> http://fenicsproject.org/mailman/listinfo/fenics-support

_______________________________________________
fenics-support mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics-support

Reply via email to