run_discovery_script_file accepts only non list data, flat dict

[mcMunich]

🔍 Before submitting the issue

• I have searched among the existing issues
• I am using a Python virtual environment

🐞 Description of the bug

if I try to use the run_discovery_script_file to send anything other than individual string parameters, it fails. While this is slightly documented https://geometry.docs.pyansys.com/version/dev/api/ansys/geometry/core/modeler/Modeler.html#Modeler.run_discovery_script_file() it doesn't say that the dict can be exclusively flat dict. I have tried all variations (nested dictionaries, tuples, lists of strings etc) and only a key:string combination works for the script_args sent to a discovery file.
And, to make it useful in the future, it would be nice to be able to send more data to the disco launch.

📝 Steps to reproduce

run the following script to try to send a dict with two keys, the second key being a list of a single string, but it will fail with more than one also.

import tempfile
from ansys.geometry.core import launch_modeler_with_discovery
import os

version = 242

file_name = 'dummy.py'
path2dummy = os.path.join(tempfile.gettempdir(),file_name)

with open(path2dummy,'w') as f:
    f.write('#nothing')

input_dict = {'a':'1', 'b':['2']}

modeler_discovery = launch_modeler_with_discovery(product_version = version)

modeler = modeler_discovery

values = modeler.run_discovery_script_file(path2dummy, script_args = input_dict)

modeler.close()

💻 Which operating system are you using?

Windows

📀 Which ANSYS version are you using?

242 and/or 251

🐍 Which Python version are you using?

3.12

📦 Installed packages

ni

[RobPasMue]

I completely agree - let me see if we can do something here to have it working.

[Alberto-DM]

The limitation is known.
I'm using this workaround to overcome the issue.

First I serialize every value in the args dict.
If args is the dictionary that I want to pass to the script, I will actually pass new_args:

new_args = copy.deepcopy(args)
for k, v in new_args.items():
    new_args[k] = json.dumps(v)

Then, inside the script that I will run, I add this lines at the beginning of the file to deserialize the arguments:

import json
keys = list(argsDict.Keys)
for k in keys:
    argsDict[k] = json.loads(argsDict[k])

In this way, the args values can be any structure that can be sterilized by json.dumps

I am actually doing this programmatically in a wrapper function of modeler.run_discovery_script_file so that I can directly pass any arg without worrying of its convertion.
You can also do the same for the result dictionary returned by the method (serialize inside the script and de-serialize outside).

@RobPasMue maybe a similar approach can be implemented natively in the method itself.