Be more pythonic

[obo]

SLTev is not written in a nice python way. Aside from worse readability, it brings needless platform dependence. For instance making os.system calls like this one: https://github.com/ELITR/SLTev/blob/ddc77fff7101d58a8b5d0d8286658480dc518330/SLTev-scripts/SLTev#L326 is:

• slower (a new process is started)
• risky (filenames get passed to command line and back)
• likely platform-dependent
• very likely also insecure (someone can plug in a different "evaluator.py"

Make SLTev a module which can be either imported or directly called (it should contain the standard test at the end and call its "main" if it is run directly).

If people only import SLTev, they should have access to the evaluation routines, to be directly called from within the python. Avoid external tools as much as possible! (with the exception of mwerSegmenter or GIZA).

[pyRis]

A more suitable way will be to use subprocess library. In general, subprocess is almost always preferred over os.system() calls.

[obo]

The point here is not to achieve any parallelization (for which subprocess would be the solution). The way os.system() is used in current SLTev is just to make a function call. :-//

[obo]

Rishu, could you please test the current code and close this issue if SLTev can be now used both:

• as module
• as a script

• From the code, it looks good, but I have not succeeded with import SLTev due to the fact that SLTev lives in the subdirectory SLTev-scripts, see also #21.

[obo]

Matusi, this is also something you might immediately see and close. If SLTev is easy to use both as a script and as a module directly from python, mark this issue as closed. Thanks, O.

[mzilinec]

@obo I think that it is usable as both script and a module, but still, the title "Be more pythonic" - is not the case yet. There are still things such as calling python evaluator.py with os.system() - this is really not good.

[obo]

@mzilinec I fully agree. Please rename the issue (or fix the call if it is very easy -- which I do not really expect).