On Sun, 05 May 2013 12:27:34 +0200
Johan Hake <[email protected]> wrote:
> On 05/04/2013 02:08 AM, Jan Blechta wrote:
> > I don't anymore expect that mkfifo will work but could you sugest
> > somehow similiar solution?
>
> Solution to what?
Well, I need to do
> _________________
> File('fifo.xml') << some_dolfin_object
> ...
> File('fifo.xml') >> other_dolfin_object
> _________________
but I'd like to avoid drive access, so I hoped it could be piped.
The reason why to do this is that some_dolfin_object is serial and
other_dolfin_object is parallel (and vice versa). This is way how to
transfer objects between serial and parallel computations in
serial-meshes DOFLIN branch - yes, whole implementation is a major
hack, but for now it enables user to perform serial projections on
non-matching meshes while doing rest in parallel, etc..
>
>
> > I mean the way to deal with << and >>
> > without drive access. I feel it is not obvious why to do it but
> > there is the reason.
>
> Think of these as read and write methods. Not as pipe methods. They
> should probably once and for all be renamed to avoid confusion.
I think this not necessary - I guess this happens few times you hear
such a question...
Jan
>
> Johan
>
> > Jan
> >
> >
> > On Fri, 03 May 2013 16:41:45 +0200
> > Johan Hake <[email protected]> wrote:
> >> Disclaimer: I have not used named pipes...
> >>
> >> That said, dolfin Files probably does not communicate with a bash
> >> pipe. Is there anything that indicates that a file sink should be
> >> able to communicate with a bash pipe?
> >>
> >> Johan
> >>
> >> On 05/03/2013 03:26 PM, Jan Blechta wrote:
> >>> Does anybody have a clue how to do something like
> >>> _________________
> >>> File('fifo.xml') << some_dolfin_object
> >>> ...
> >>> File('fifo.xml') >> other_dolfin_object
> >>> _________________
> >>>
> >>> using named pipe? If you do
> >>> mkfifo fifo.xml
> >>> in bash prior to calling python it deadlocks at <<.
> >>>
> >>> Jan
> >>> _______________________________________________
> >>> fenics mailing list
> >>> [email protected]
> >>> http://fenicsproject.org/mailman/listinfo/fenics
> >>>
> >>
> >> _______________________________________________
> >> fenics mailing list
> >> [email protected]
> >> http://fenicsproject.org/mailman/listinfo/fenics
> >
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics
