Prioritise ANUGA integration

[JamesRamm]

ANUGA Hydro has very loose integration with TuCluster.
I believe Anuga should be the first class citizen in TuCluster with best support because:

• It is FOSS and therefore both matching the spirit of TuCluster and easier to test & integrate
• It already runs natively on Linux

Basic integration has been provided by #15.

Further steps/ideas to ensure the integration is robust:

• Documentation and preferably setup scripts for installing anuga in a conda environment.

• Integration tests - run a real model. (This is one that should probably be tackled by qflow https://github.com/JamesRamm/qflow)

• Could Security and validation be provided by introspection of the user-provided ANUGA script to check:

  • Input data exists
  • Outputs are not written anywhere but the configured data dir.
  • Any obvious malicious code?
  • Verify the script is expected to terminate.

• Could we potentially wrap any module...perhaps define some coding standards (the module should be wrapped in a function) and enforce them via introspection?

[JamesRamm]

Anuga validation

Validation of the anuga script is tricky as since it is just a python module it could potentially do anything.

Due to running on a remote server, we need to be able to control where data files are output and possibly prevent other actions.
We can't stop everything, but I think scanning the script for ANUGA functions we want to ban/restrict and some obvious python functions should be possible:

• Scan the script for any call to open. This isnt required for anuga since it controls outputting files. If open is called, we terminate running the script

• Check the script for these restricted anuga functions:

  • set_datadir. If specified we should rewrite this line