Plotting of model diagrams and various other changes

[mattfidler]

Thanks for all the useful packages; both the ones you created and the ones you pointed me to. You have a great deal of experience in R.

After reviewing RxODE, I think the it has the following benefits over deSolve:

• I believe the solving is faster;
  • There is no intermediary R code call to choose the solver. The dlls include the lsoda routines.
  • To achieve the faster solving routines you don’t need to create C or FORTAN code, like you do for deSolve.
  • The only drawback is not being able to use native r code. I cannot see a clear advantage to R code for odes…
• Easier code generation
  • The parser allows you to specify parameters easily. While the overall program still requires an ordered vector, this book-keeping is taken care of internally
  • Without specifying inits, everything defaults to an initial condition of 0; This is usually true and quite handy.
  • I expanded this further to allow you to specify inits out of order and only specify an init diferent from zero, like c(eff=100)

I think that my patches can provide additional value to your RxODE package:

• It can draw model diagrams, which can be quite handy.
  • This is still a bit rudimentary, and could possibly benefit from the dparser.
  • It currently indirectly uses your parsed code by using the model variables created by the parser.
• It is less picky about syntax:
  • It supports the \ operator
  • It supports the <- R assignment paradigm
  • It supports a “.” In variable names
  • Semicolons are not required at the end of lines.
  • Differential equations can be supported with different syntax:

 d(eff) =
 d(eff)/dt =
 (eff) =
 …..(eff)…. =
 [“a”]=

• These can be taken out by specifying strict=TRUE
  • Specifying RxODE(“model.txt”) opens model file if it exists and uses it, allowing lazy specification of filename
  • Based on the model text provided,
• Models are not recompiled if already compiled
• Models can be built and loaded without a compilation directory (flat=TRUE). Produces dll and Rdata files
• By default, RxODE(x) will use the flat directory structure
• The c-code model function names imported are also based on the digest of the model. I think that this is needed so that if I name all models “RxODE” there isn’t an overlap with the dll imported functions….

[mattfidler]

This also allows duplicate d/dt(x) syntax like:

d/dt(x) = ka*depot
d/dt(x) = d/dt(x)-kel*x

Although undocumented (and it doesn't raise an error), this leads to problems and/or errors in the compiled code.

[mattfidler]

This branch is no longer the current branch.