Refactoring sampling

[vfisikop]

Currently dingo has many functions for sampling and with different names that might confuse the user. For example, there are:

• generate_steady_states
• generate_steady_states_no_multiphase
• sample_from_polytope
• sample_from_polytope_no_multiphase
• sample_from_fva_output
  All of the are performing sampling from a polytope (but with some differences).

With the current interface if the user want to sample from a metabolic model then:

input_file_json = os.getcwd() + "/ext_data/e_coli_core.json"
model = MetabolicNetwork.from_json( input_file_json )
sampler = PolytopeSampler(model)
sampler.generate_steady_states()

but what happens if the user has the matrices that define a polytope? (or they want to sample a polytope not related to a metabolic model)

I propose the following change in the interface. Rename class PolytopeSampler to Polytope that can be initialized either with a metabolic model or with a set of matrices. Then there is a single function sample (or sampling or sample_points) that has parameters the ess, psrf, the distribution, the walk algorithm, rounding or not etc. In the case of metabolic models, the Polytope class will have a generate_stead_states function that for all the other cases (the polytope does not correspond to a metabolic model) will return a non applicable warning.