ReducedFunctional doesn't like InitialConditionParameter("name")

[Patrick Farrell]

[pef@aislinn:~/src/dolfin-adjoint/tests/optimization]$ bzr diff
=== modified file 'tests/optimization/optimization.py'
--- tests/optimization/optimization.py 2012-08-19 19:55:02 +0000
+++ tests/optimization/optimization.py 2012-10-08 14:52:25 +0000
@@ -54,7 +54,7 @@
     lb = project(Expression("-1"), V)

     # Define the reduced funtional
- reduced_functional = ReducedFunctional(J, InitialConditionParameter(u))
+ reduced_functional = ReducedFunctional(J, InitialConditionParameter("Velocity"))

     # Run the optimisation problem with gradient tests and L-BFGS-B
     u_opt = minimize(reduced_functional, algorithm = 'scipy.l_bfgs_b', pgtol=1e-6, factr=1e5, bounds = (lb, 1), iprint = 1)

[pef@aislinn:~/src/dolfin-adjoint/tests/optimization]$ python optimization.py
Traceback (most recent call last):
  File "optimization.py", line 60, in <module>
    u_opt = minimize(reduced_functional, algorithm = 'scipy.l_bfgs_b', pgtol=1e-6, factr=1e5, bounds = (lb, 1), iprint = 1)
  File "/home/pef/src/dolfin-adjoint/dolfin_adjoint/optimization.py", line 236, in minimize
    return optimization_algorithms_dict[algorithm][1](reduced_func_array, reduced_func_deriv_array, [p.data() for p in reduced_func.parameter], **kwargs)
  File "/home/pef/src/dolfin-adjoint/dolfin_adjoint/optimization.py", line 106, in minimize_scipy_fmin_l_bfgs_b
    m_global = get_global(m)
  File "/home/pef/src/dolfin-adjoint/dolfin_adjoint/optimization.py", line 35, in get_global
    raise TypeError, 'Unknown parameter type %s.' % str(type(m))
TypeError: Unknown parameter type <type 'str'>.

[Patrick Farrell (pefarrell)]

Making the change
=== modified file 'dolfin_adjoint/parameter.py'
--- dolfin_adjoint/parameter.py 2012-08-01 14:43:02 +0000
+++ dolfin_adjoint/parameter.py 2012-10-08 14:54:16 +0000
@@ -60,7 +60,10 @@
       return None

   def data(self):
- return self.coeff
+ if isinstance(self.coeff, str):
+ return adjglobals.adjointer.get_variable_value(self.var).data
+ else:
+ return self.coeff

 class ScalarParameter(DolfinAdjointParameter):
   '''This Parameter is used as input to the tangent linear model (TLM)

helps -- it now gets a little further:

RUNNING THE L-BFGS-B CODE

           * * *

Machine precision = 2.220D-16
 N = 21 M = 10

At X0 1 variables are exactly at the bounds
Traceback (most recent call last):
Traceback (most recent call last):
  File "optimization.py", line 60, in <module>
    u_opt = minimize(reduced_functional, algorithm = 'scipy.l_bfgs_b', pgtol=1e-6, factr=1e5, bounds = (lb, 1), iprint = 1)
  File "/home/pef/src/dolfin-adjoint/dolfin_adjoint/optimization.py", line 236, in minimize
    return optimization_algorithms_dict[algorithm][1](reduced_func_array, reduced_func_deriv_array, [p.data() for p in reduced_func.parameter], **kwargs)
  File "/home/pef/src/dolfin-adjoint/dolfin_adjoint/optimization.py", line 115, in minimize_scipy_fmin_l_bfgs_b
    mopt, f, d = fmin_l_bfgs_b(J, m_global, fprime = dJ, bounds = bounds, **kwargs)
  File "/usr/lib/python2.7/dist-packages/scipy/optimize/lbfgsb.py", line 196, in fmin_l_bfgs_b
    f, g = func_and_grad(x)
  File "/usr/lib/python2.7/dist-packages/scipy/optimize/lbfgsb.py", line 152, in func_and_grad
    g = fprime(x, *args)
  File "/home/pef/src/dolfin-adjoint/dolfin_adjoint/optimization.py", line 213, in reduced_func_deriv_array
    seed = dolfin.parameters["optimization"]["test_gradient_seed"])
  File "/home/pef/src/dolfin-adjoint/dolfin_adjoint/utils.py", line 455, in test_gradient_array
    j_direct = J(x)
  File "/home/pef/src/dolfin-adjoint/dolfin_adjoint/optimization.py", line 229, in reduced_func_array
    m = [p.data() for p in reduced_func.parameter]
  File "/home/pef/src/dolfin-adjoint/dolfin_adjoint/parameter.py", line 64, in data
    return adjglobals.adjointer.get_variable_value(self.var).data
  File "/usr/lib/python2.7/dist-packages/libadjoint/libadjoint.py", line 1429, in get_variable_value
    clib.adj_get_variable_value(self.adjointer, var.var, vec)
  File "/usr/lib/python2.7/dist-packages/libadjoint/libadjoint.py", line 18, in handle_error
    raise exception, errstr
libadjoint.exceptions.LibadjointErrorNeedValue: Need a value for Velocity:0:0:Forward, but don't have one recorded.

If you apply the following patch (which you shouldn't), this fixes it:
=== modified file 'dolfin_adjoint/optimization.py'
--- dolfin_adjoint/optimization.py 2012-09-20 15:40:13 +0000
+++ dolfin_adjoint/optimization.py 2012-10-08 14:55:59 +0000
@@ -204,7 +204,7 @@
         if (m_array != get_global(m)).any():
             reduced_func_array(m_array)

- dJdm = utils.compute_gradient(reduced_func.functional, reduced_func.parameter)
+ dJdm = utils.compute_gradient(reduced_func.functional, reduced_func.parameter, forget=False...