TimeSeries problem

It looks like you repeatedly store values at the same points in time.

TimeSeries assumes that the times given are strictly increasing. I'll
add a check for this.

--
Anders

On Sat, May 05, 2012 at 08:35:47AM -0000, Praveen C wrote:
> New question #196105 on DOLFIN:
> https://answers.launchpad.net/dolfin/+question/196105
>
> I am trying to store data into a TimeSeries, retrieve it, modify and then store it back. But this is not working correctly. Below is a simple example.
>
> from dolfin import *
>
> mesh = Interval(3, 0, pi)
>
> # Create Function Space
>
> Q = FunctionSpace(mesh, "CG", 1)
> u = Function(Q)
>
> # Parameters
> T = 2
> nt = T
> dt = T/nt
>
> # Time Stepping
> time_u = TimeSeries("U")
>
> uex = Expression("sin(t)*sin(x[0])+cos(t)*sin(x[0])",t=0)
>
> t = 0
> while t <= T:
> uex.t = t
> u = project(uex, Q)
> time_u.store(u.vector(),t)
> t += dt
>
> it = 0
> while it < 5:
> t = 0
> while t <= T:
> print "Iter=", it, ", t = ", t
> time_u.retrieve(u.vector(),t)
> # Do something to u and store it back
> time_u.store(u.vector(),t)
> print u.vector().array()
> t += dt
> it += 1
>
> I repeatedly retrieve and store u back to the TimeSeries. The output shows that there is something wrong
>
>
> Iter= 0 , t = 0
> [ 7.95138670e-17 8.66025404e-01 8.66025404e-01 0.00000000e+00]
> Iter= 0 , t = 1
> [ 7.95138670e-17 1.19665077e+00 1.19665077e+00 0.00000000e+00]
> Iter= 0 , t = 2
> [ 0. 0.42708094 0.42708094 0. ]
> Iter= 1 , t = 0
> [ 7.95138670e-17 8.66025404e-01 8.66025404e-01 0.00000000e+00]
> Iter= 1 , t = 1
> [ 7.95138670e-17 1.19665077e+00 1.19665077e+00 0.00000000e+00]
> Iter= 1 , t = 2
> [ 7.95138670e-17 1.19665077e+00 1.19665077e+00 0.00000000e+00]
> Iter= 2 , t = 0
> [ 7.95138670e-17 8.66025404e-01 8.66025404e-01 0.00000000e+00]
> Iter= 2 , t = 1
> [ 7.95138670e-17 1.19665077e+00 1.19665077e+00 0.00000000e+00]
> Iter= 2 , t = 2
> [ 0. 0.42708094 0.42708094 0. ]
> Iter= 3 , t = 0
> [ 7.95138670e-17 8.66025404e-01 8.66025404e-01 0.00000000e+00]
> Iter= 3 , t = 1
> [ 7.95138670e-17 1.19665077e+00 1.19665077e+00 0.00000000e+00]
> Iter= 3 , t = 2
> [ 0. 0.42708094 0.42708094 0. ]
> Iter= 4 , t = 0
> [ 7.95138670e-17 8.66025404e-01 8.66025404e-01 0.00000000e+00]
> Iter= 4 , t = 1
> [ 7.95138670e-17 1.19665077e+00 1.19665077e+00 0.00000000e+00]
> Iter= 4 , t = 2
> [ 7.95138670e-17 1.19665077e+00 1.19665077e+00 0.00000000e+00]
>
>
> For example
>
> "Iter=0, t=2" is not same as "Iter=1, t=2"
>
> Also "Iter=1, t=1" and "Iter=1, t=2" are same which is not correct.
>
> However during Iter=2, things are correct. Also correct for Iter=3, but wrong for Iter=4.
>
> praveen
>

Yes, I want to update u(t) and store it back into the same TimeSeries at the same time t. The problem involves optimization of a time dependent pde, and hence the need to do this.

The explanation here

http://fenicsproject.org/documentation/dolfin/dev/python/programmers-reference/cpp/TimeSeries.html

"A new time series will check if values have been stored to file before (for a series with the same name) and in that case reuse those values. If new values are stored, old values will be cleared."

made me think that if I store at a time which already has data, it will overwrite it.

praveen

Anders,
             Can you please tell me how to use the time series with the above modification while storing and retrieving values?
                                                                                                                                                                                                Souvik

You cannot store values at the same timestep twice. Then TimeSeries interprete this as adding a new value at the same timepoint, issuing an error. You need to store the new values in seperate timeseries:

while it < 5:
    t = 0
    time_u_updated = TimeSeries("U_updated_%d" % it)
    while t <= T:
       print "Iter=", it, ", t = ", t
       time_u.retrieve(u.vector(),t)
       # Do something to u and store it back
       time_u_updated.store(u.vector(),t)
       print u.vector().array()
       t += dt
    it += 1