dolfin team mailing list archive

[logg <logg@xxxxxxxxx>]

On Fri, Sep 11, 2009 at 05:56:37AM -0000, Johan Hake wrote:
> On Thursday 10 September 2009 17:01:01 logg wrote:
> > On Thu, Sep 10, 2009 at 01:16:57PM -0000, Johan Hake wrote:
> > > > Works now!
> > > >
> > > > Now if only I could also get values from a vector directly from a
> > > > Python list or Python array of numpy.uint32... :-)
> > > >
> > > > This fails:
> > > >
> > > >   values = x[dofs]
> > > >   values = x[numpy.array(dofs)]
> > > >
> > > > But this works:
> > > >
> > > >   values = [x[dof] for dof in dofs]
> > > >
> > > > dofs is a Python list of numpy.uint32.
> > >
> > > Try now...
> > >
> > > numpy c-api is not fun...
> >
> > I understand your pain.
> >
> > > Feel free to close the bug if it works.
> >
> > Works now! Thanks! I'll close the way.
> >
> > Something I still can't do is the following. It's related but it may
> > be more of a feature request than a bug report. I'll just post it here
> > and we can open a new bug report if you think it's a bug:
> >
> > Works:
> >
> >   for i in range(n):
> >     x[cell_dofs[i]] = coefficients[i]
> >
> > Does not work:
> >
> >    x[cell_dofs] = coefficients
>
> This should work now, with any numpy array of numbers on the right side. The
> arrays are converted to double before assignment.
>
> Johan

Thanks!

--
Anders

-- 
Vector indexing in Python with NumPy indices
https://bugs.launchpad.net/bugs/427047
You received this bug notification because you are subscribed to DOLFIN.

Status in DOLFIN: Fix Committed

Bug description:
The following breaks:

import dolfin, numpy

x = Vector(10)

print x[numpy.int32(0)]
print x[numpy.int64(0)]
print x[numpy.array((0, 1), dtype='I')

The typechecks in la_post.i looks like they can be expanded. (I tried it myself but couldn't get it working correctly.)