Support for passing custom environment variables to subprocess

[stefan6419846]

I would like to pass my own environment variables to the subprocess, for example to set OMP_THREAD_LIMIT. This custom value should be limited to a specific scope, as other parts of my program should not be affected by this.

Judging from the code, this is not possible for now.

[bozhodimitrov]

One way to workaround that is to set the environment beforehand, because if I recall correctly, forking inherits the env from the parent process.

Ofcourse, the downside is that you are setting the env vars once and it pollutes the current env forever until you change them again.

[stefan6419846]

As you already mentioned, your approach would make this setting permanent for the whole program. I could write a context manager for temporarily changing the value, but this is quite an overhead.

For now, I settled on monkey patching the pytesseract environment, but this is just an ugly hack as well:

import copy
import os
from pytesseract import pytesseract

path = 'path/to/file.jpg'

pytesseract.environ = copy.deepcopy(os.environ)
pytesseract.environ['OMP_THREAD_LIMIT'] = '1'

print(pytesseract.image_to_data(path, output_type='data.frame'))

[bozhodimitrov]

I guess that you want to dynamically change the environment variables.
Because for standalone usage of pytesseract, you can just use OMP_THREAD_LIMIT=1 pytesseract ...

The python docs refer to os.environ.copy() as a good way to make environment copy of the mapping.
Then you can pass your specific modified version, but monkey patching pytesseract is still required with this method.

I am not sure, but maybe os.putenv() might work for your use case.
It looks less hacky, but also looks very implicit and the support for it varies.