Script parsing fails when list/dictionary literals are not valid JSON

[jeremyadamsfisher]

Describe the bug
When importing a pipeline with {import,update}_pipe_project.py, some script constants that are valid python literals cannot be parsed. Although the overall script parsing succeeds, this leads to errors downstream:

• When ENVS are invalid JSON, the script hangs on execution
• When the ARGUMENTS are invalid JSON, the script can be run, but the arguments cannot be changed from the frontend

To Reproduce: script hangs on execution

1. In any script, make the constant a valid python literal, but an invalid JSON. For example:

ENVS = ["custom-worker"]  # comment about custom worker

2. This results in a json.decoder.JSONDecodeError
3. Script hangs indefinitely
   

To Reproduce: Arguments cannot be changed from the frontend

1. In any script, make ARGUMENTS a valid python literal, but an invalid JSON. For example:

ARGUMENTS = {
    # comments are not allowed in JSON
    "file_name": {
        "value": "annos.csv",
        "help": "Name of the file with exported bbox annotations."
    }, # <- also, notice the trailing comma
}

2. This results in a json.decoder.JSONDecodeError
3. Front-end no longer displays "file_name" as an argument

Expected behavior

• Front-end should display arguments
• ARGUMENTS and ENVS hould be parsed like python literals, raise a more descriptive error or fail outright.

Additional context
It seems to me that this line could simply be changed to arg_dict = ast.literal_eval(args) to fix both issues. I've tested this minimally and it seems to work.

I would be happy to write some unit tests and make a PR, if there is interest.

[jaeger-j]

Hi @jeremyadamsfisher,
thank you for your contribution. Your proposed solution will be released with version 1.3.0.
It would be great if you write some unit test.
Best,
Jonas