ToDo.txt

Bugfix: Adding solute to multiple compartments causes segfault

Switch magnus (and precursors) to sparse matrices

Finalize and clean up matrix exponential routines

Try to get quadratic fit working in 

Secondary considerations:
  Change traces from list<Trace> to list<Trace*> in differential_equation
    restore NeuronModel::outputRecordedTraces to const

  In DiffEqFunctor:
    change addTraceTarget to have boolean option to specify if a target is
      clampable or fitable
        -non-state values should NOT be fitable
        -Injector values should NOT be clampable
        -Maybe default to false and explicitely make state values fitable
        -change interaction with TraceTarget in Trace.h
    Improve initializeDiffEq so that it's easier to use

  Move to c-interface for MPI
  
  Dynamic connection to MPI workers
    Better responding to slave exceptions in mpi_parallel
    prevent core files from being produced when program is stopped

  changes to CurrentInjector classes
    remove time from ion channel dynamics?
    Maybe keep track of constant/changing temperature (use in initialize())
    opportunity to record individual CurrentInjector currents
      -actually done, see NeuronModel::addOnePointInjectorCurrent()
    opportunity to record reversal potentials
  
  Move accumulator / reversal potential calculation live in Solute classes [?]
  
  Change to runtime compilation of NeuronModel::getDerivatives
  
  Change Minimizable::initialize() to Minimizable::initializeMinimizable()
  
  Maybe revert to scheme with two saved states and one future state
    (presently use one saved state, one half step in future, and one full step)

  Maybe improve efficiency of scheme to make separate action of Record traces vs
    Fit/Clamp
  
  Re-do temperature dependence of current injectors:
    detect if temperature will be varied and alter Q10 behavior accordingly
    put in temperature dependence of gBar_t
  
  
  when passing more than two parameters to perturbneuron.bin vary the
    parameters pairwise (in all possible pairs, with no duplicate evaluations)

  Unhack the following:
    NeuronFitInfo::dataIsLoaded
    exception messages in NeuronFitInfo::addFitTrace
    checking for compatible ParameterDescriptionLists
  
  improve flexibility of perturb_neuron.bin
    -maybe incorporate into general batch simulator, e.g. batch_neuron.bin

  outputting progress:
    change verbosity, logToTermain -> verbose
    change if(verbose) {cout << ...} -> logOut << ...
  
  Change namespace scheme
    
  update documentation for OhmicChannel, NernstChannel, GhkChannel
    -make clearer what must be done to create one
    -update comments on inherited members/functions