fiat team mailing list archive

["Garth N. Wells" <g.n.wells@xxxxxxxxxx>]

This is related to numpy (making me responsible for introducing the
bug). In P0.py, on line 44,

    bary = numpy.average( numpy.array( map( numpy.array , \
                                                    vs.values() ) ) )

should be replaced by

    bary = numpy.average( numpy.array( map( numpy.array , \
                                                   vs.values() ) ) , 0 )

Some basic verification tests in FIAT would be really useful.

Garth

Anders Logg wrote:
> Rob, vector-valued DG(0) is broken:
> 
>>> from FIAT.DiscontinuousLagrange import *
>>> from FIAT.shapes import *
>>> P0 = DiscontinuousVectorLagrange(TRIANGLE, 0, 2)
> ---------------------------------------------------------------------------
> exceptions.TypeError
> 
> /usr/lib/python2.4/site-packages/FIAT/DiscontinuousLagrange.py in
> DiscontinuousVectorLagrange(shape, n, nc)
>      89     else:      return DiscLagrange( shape , n )
>      90 
>      91 def DiscontinuousVectorLagrange( shape , n , nc ):
> ---> 92     if n == 0: return P0.VecP0( shape , nc )
>      93     else:      return DiscVecLagrange( shape , n ,nc )
> 
> /usr/lib/python2.4/site-packages/FIAT/P0.py in __init__(self, shape, nc)
>      61     def __init__( self , shape , nc ):
>      62         U = polynomial.OrthogonalPolynomialArraySet( shape , 0
>      , nc )
> ---> 63         Udual = VecP0Dual( shape , U )
>      64         polynomial.FiniteElement.__init__( self , Udual , U )
>      65 
> 
> /usr/lib/python2.4/site-packages/FIAT/P0.py in __init__(self, shape, U)
>      44         bary = numpy.average( numpy.array( map( numpy.array ,
>      \
>      45
>      vs.values() ) ) )
> ---> 46         self.pts = ( tuple(bary) , )
>      47         ls = [ functional.ComponentPointEvaluation( U , c ,
>      bary ) \
>      48                for c in range( nc ) ]
> 
> TypeError: iteration over non-sequence
> 
> /Anders
> _______________________________________________
> FIAT-dev mailing list
> FIAT-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/fiat-dev
>