Python API?

[dennisbrookner]

I think it is worthwhile to consider a python API for rs-booster analogous to its command-line functionalities. I know that this would be useful and preferable for me, and I have discussed this with Doeke as well. Given that the spirit of reciprocalspaceship is essentially "you can do crystallography in python!" it feels entirely natural for rs-booster to work this way as well. I just wanted to open an issue to gauge thoughts on this and get folks' opinions.

I am happy to take this on. I don't think it would require the writing of much new code, just some refactoring of existing code. The end result can and certainly should be entirely reverse compatible with existing rs-booster usage.

[JBGreisman]

My plan is to restructure things (as part of #24) to provide a bit more of a python interface. Right now, many of the tools are written as simple Python scripts, so although it is possible to import things they are not structured in a useful way.

My plan though is to have a limited scope for the task -- for example, I plan to write a DifferenceMap class to back many of the CLI implementations.

Do you have any specific things in mind? My hesitation with going to broad in terms of this library, is that I don't want it to be confusing where rs ends and rsbooster begins -- I still very much see rsbooster as providing useful implementations of certain tasks.

[dennisbrookner]

I'm going to make a PR soon with regards to this, because I'm building a larger python-based thing that uses rs.scaleit and I would like to be able to access scaleit from within python. My goal will be basically to have the main() function do only the following:

def main():
    args = parse_arguments()

    mtz1 = rs.read_mtz(args.mtz1) # etc. for other input files

    scaled = do_the_stuff(mtz1, ..., *args, etc.)

    scaled.write_mtz(args.mtzout)

    return

Such that everything except the io is accessible from inside python.

Does this make sense as a general structure? I like it, and I like the idea of keeping the python and command-line APIs as similar as possible. I would also be happy to go through and reorganize other functions in this manner.

[JBGreisman]

yup -- I think this plan makes sense to me, and is a good framework for organize comparable methods/tools