从子流程重构或捕获python错误

Question

I'm currently trying to get attributes of a python stack trace and error from either purely the stderr or intercepting a script and catching its Exception. Either way is fine. This may be an XY problem so alternatives are welcome.

I am currently writing a program that kind of, redirects the arguments to call another python script:

import sys, subprocess
pyargs = sys.argv[1:] # Suppose pyargs is something like ["a.py", "123"]
result = subprocess.run(["python"] + pyargs, stdout=subprocess.PIPE, 
    stderr=subprocess.PIPE)
error = result.stderr # do something with this

Approach 1: Parsing dump

I'm not sure whether there is any existing method that supports either of the two plans. I know that the python error dump is displayed in a very formulaic way:

import sys, traceback
exc_type, exc_value, exc_traceback = sys.exc_info() # shows formatting
traceback.print_exc() # prints the stderr of what we wanted.

And so I thought perhaps we could parse the stderr and see what happens.

Approach 2: Catching exception

I know that we can run subprocess with subprocess.run(..., check=True) to get the subprocess.SubprocessError , but since the subprocess can be anything(not just a python process), we can't actually intercept the actual exception that occurred.

One issue with loading in the python file as a module to this file, is that we cannot supply the necessary arguments, unless we can somehow spoof the sys.argv[] or argparse that's running from the underlying script. Scripts that may have if __name__ == "__main__": will not execute either unless we spoof that too.

Any ideas?

Answer 1

I prefer approach #2. Whenever possible, I avoid using subprocess or system to call Python from Python.

Changing the value of sys.argv is easy - you just overwrite it using regular assignment.

Tricking the __name__ == "__main__" conditional is harder. Unfortunately, it probably means that you can't simply import the module in the usual way. According to How to import a module as __main__ ? , you can use the runpy module to execute your module and supply whatever value of __name__ you like.

import runpy
import sys
sys.argv = ["C:/", "foo", "bar"]
try:
    runpy.run_module("yourmodulename", {}, "__main__")
except Exception as e:
    print("caught exception:", e)

Now you can catch any uncaught exceptions that your module throws, and handle them however you like within the except .