Call ocropy scripts from other python code

[bugsyb]

Hi All,

What is suggested way to run ocropy scripts from other python code?
The most recommended way is to us "import " but:

• that searches for .py file,
• expects to process module and not script (as is ocropy built)

Could scripts be wrapped into standard:
if name == "main":
main()
sys.exit(0)

Some further minor teaks to map args.* to other variables in main code.

The idea is to avoid duplication of code included in scripts, i.e threshold calculation, etc. and have it re-usable.

The other methods considered for today is to use:

• execfile(),
• putting init.py file in folder (not sure if that would work at all correctly),

Other options I've looked at are listed here:
https://stackoverflow.com/questions/2349991/python-how-to-import-other-python-files/20749411#20749411

Another option could be one of demonstrated approaches (0,1 or 2):
https://m8051.blogspot.com/2014/02/python-tricks-running-python-scripts.html

        if bDirect == 0:
            sRunPath = sPath
            with open(path.join(sPath,sFile)) as f:
                exec(compile(f.read(), "temp.p", 'exec'), globals())
        elif bDirect == 1:
            tmp = os.getcwd()
            os.chdir(sPath)
            os.system("python " + sFile)
            os.chdir(tmp)
        elif bDirect == 2:
            sys.path.insert(0,sPath)
            tmp = importlib.import_module(sFile.split(".py")[0])
            tmp.run()
sys.path.remove(sPath)

[mittagessen]

You can't really use ocropy as a library right now because data loading and actual functionality is so intermeshed in the main scripts and most of them also implicitely assume a certain layout of the data on the file system. The last time I tried I just called them using the subprocess module but then went on to refactor everything just because ensuring that all the pseg,nrm,.... files are in the correct location is a pain.

f you want a clean-ish API with mostly the same functionality you might want to look at kraken. The documentation is admittedly lacking but just throwing a PIL image at most functions is a lot nicer than working around each script's peculiarities.

[ChillarAnand]

I have written a simple wrapper function to run ocoropy pipeline. If you are working with lots of images, something like this might be useful.