model modules
henrus opened this issue · 6 comments
With this enhancement everything that a model needs will be stored in a single python module. See modelModules branch for work in progress.
I've done quite a step towards this in commit 4b07f7d
Next step is to move the model modules for the two tanks example zoo into common directory.
I think we are almost there. The only issue I see is that python must be able to find the library, ie. PYTHONPATH must be adapted. I think this is just logical not a major issue.
I pulled #32. Looks good. initModels and workflow will be need to go into a unified driver script, ie. incorporated into the main library.
examples/twoTanksChained and examples/twoTanksFortran fail for me ...
Not sure what you mean by "Note that I am still doing the path mangling. I am not sure of what would be the best way to proceed in order to get rid of it, there are several options."?
@Not working examples
The chained example I cannot explain, but Fortran did not work for me even
before I started doing the surgery.
@path mangling
If you for instance look in models/twoTank/twoTank.py you see:
script = os.path.dirname(os.path.abspath(file)) +
"/src/twoTanksMacroscopicProblem"
where it used to be:
script = "../src/twoTanksMacroscopicProblem"
The problem is that I do not know how to let cmake compile the executables
in the simulation directory. If I can do that we can go back to the “old”
style of using the relative path to the executables.
In order to solve this I somehow need to let cmake know:
- Which modules I want to compile
- Where to find them
Ideally I want to do that with minimal effort from the user.
Sigve
On 3 October 2015 at 13:50, Henrik Rusche notifications@github.com wrote:
I pulled #32 #32. Looks
good. initModels and workflow will be need to go into a unified driver
script, ie. incorporated into the main library.examples/twoTanksChained and examples/twoTanksFortran fail for me ...
Not sure what you mean by "Note that I am still doing the path mangling. I
am not sure of what would be the best way to proceed in order to get rid of
it, there are several options."?—
Reply to this email directly or view it on GitHub
#4 (comment)
.
Hi Sigve,
I don't think that the absolute paths are actually a problem. These applications could be big and there should be no need to compile them per simulation.
Henrik