Closed varvara-l closed 9 years ago
for now we should use the conf.yaml file for this. That's already what
we're doing with the workers:
param in yaml.
I think we want to avoid globals because that's a code smell (especially in python). If a component needs a parameter from the config, like the tmp/ directory path, then it needs to be initialized with that parameter, probably with a sensible default in most cases.
So, for the alignment model, it should be initialized with a tmp/ dir location, and it should probably delete the alignment files by default once we're finished using them.
In the future, the user should be able to name the tmp/ dir after the experiment, so that different experiment configurations can be stored in different directories.
On Thu, Jan 15, 2015 at 3:26 PM, varvara-l notifications@github.com wrote:
There are some configuration parameters in the system that different functions and classes can require any time (for example, a special directory for temporary files, the number of processors the system can use). It's sometimes inconvenient to pass them to functions, so there should be a list of global variables which are seen from everywhere. But I don't know what is a good way to do it in python.
— Reply to this email directly or view it on GitHub https://github.com/qe-team/marmot/issues/5.
All global parameters should be defined in the config and passed to functions that need them. There won't be any global parameters seen from everywhere.
There are some configuration parameters in the system that different functions and classes can require any time (for example, a special directory for temporary files, the number of processors the system can use). It's sometimes inconvenient to pass them to functions, so there should be a list of global variables which are seen from everywhere. But I don't know what is a good way to do it in python.