Trying to "solve" an integer CellVariable should raise an error

[guyer]

Declaring

var = CellVariable(..., value=0)
 :
eq.solve(var)

produces confusing error messages deep in PySparse that result from var holding integers ("value=0" instead of "value=0.").

One option is to chance the default type for CellVariable, since integer arrays are probably only useful for FiPy's internal purposes.

Perhaps better to leave the semantics of "0" vs. "0." to Python and NumPy to sort out and simply raise an error (or warning?) if an attempt is made to "solve" an integer CellVariable, since that is not meaningful.

Imported from trac ticket #143, created by guyer on 05-27-2008 at 12:53, last modified: 06-12-2013 at 14:14

[guyer]

Previously reported as #133

Trac comment by guyer on 06-26-2008 at 16:32

[guyer]

Warning raised in r2829

Trac comment by guyer on 11-20-2008 at 15:57

[guyer]

As reported in http://thread.gmane.org/gmane.comp.python.fipy/2826/focus=2838, coupled equations produce weird results if the solution variables hold integers. This should have been addressed by this ticket, and is for single equations, but coupled equations don't raise a warning, e.g.,

import fipy as fp

mesh = fp.Grid1D(nx=10)

v1 = fp.CellVariable(mesh=mesh, value=1, hasOld=True)
v2 = fp.CellVariable(mesh=mesh, value=1., hasOld=True)

eq1 = fp.TransientTerm(var=v1) h2. fp.DiffusionTerm(var=v1)
eq2 = fp.TransientTerm(var=v2) fp.DiffusionTerm(var=v2)

(eq1 & eq2).solve(dt=1.)

Trac comment by guyer on 12-12-2012 at 15:05

[fipymigrate]

In 5441732:

#CommitTicketReference repository="fipy" revision="5441732192e13c4f95dd891c5458fff10f6287a2"
Raise a warning when trying to "solve" an integer `CellVariable`.
Fixes issue #128


git-svn-id: svn+ssh://code.matforge.org/fipy/trunk@2829 d80e17d7-ff13-0410-a124-85740d801063

Trac comment by Jonathan E. Guyer guyer@nist.gov on 06-12-2013 at 14:14