feat: add script to generate python scripts from xml

[sanjayankur31]

This will be useful in multiple cases:

a) creating python scripts from xml models as starting templates for users
b) for llms/agents---where we want them to write python, not xml

[sanjayankur31]

@HengyeZhu: would you have time to review this please?

[HengyeZhu]

No problem! I’ll start working on it soon.

[HengyeZhu]

Hi @sanjayankur31,

Sorry for the delayed reply. This PR may be fine, except for inline-cell cases such as https://github.com/OpenSourceBrain/262670/blob/master/NeuroML2/KC.net.nml.

from pyneuroml import pynml

def test_inline_cell_validate():
    model = read_neuroml2_file("HH_example_net.nml", include_includes=True)

    with tempfile.NamedTemporaryFile("w", dir=".") as f:
        write_neuroml2_file(model, f.name)

        script = NmlPythonizer(f.name, "temp_nml2py_1").write()[-1]
        exec(Path(script).read_text(), {})

        assert pynml.validate_neuroml2(Path(f.name).name)

With the current code, this fails because the generated Python includes the extracted inline cell as:

nml_doc.add("IncludeType", href='hh_cell.cell.nml')

while the extracted inline cell file is written to:

temp_nml2py_1/hh_cell.cell.nml

After generating the extracted-component href relative to the output directory, the test passes.

[sanjayankur31]

Hi @sanjayankur31,

Sorry for the delayed reply. This PR may be fine, except for inline-cell cases such as https://github.com/OpenSourceBrain/262670/blob/master/NeuroML2/KC.net.nml.

from pyneuroml import pynml

def test_inline_cell_validate():
    model = read_neuroml2_file("HH_example_net.nml", include_includes=True)

    with tempfile.NamedTemporaryFile("w", dir=".") as f:
        write_neuroml2_file(model, f.name)

        script = NmlPythonizer(f.name, "temp_nml2py_1").write()[-1]
        exec(Path(script).read_text(), {})

Shouldn't the script be run after changing into the temp_nml2py_1 directory?

The idea is that all the new files are generated in a new folder and must be run there. A few reasons for this:

• keep the original model files completely separate from the new generated files
• do not risk overwriting original files (if the id etc. matches in your test, for example, the HH_example_net.nml file could be overwritten)
• simplifying paths: we could technically add folders/relative paths to the include hrefs, but that then means that whenever these extracted files are to be used, the same folder structure will have to be followed. That's not a bad thing, necessarily, but i thought it's better to use a flat directory structure, and then users can move files around and update paths as they wish.

What do you think?

[HengyeZhu]

The idea is that all the new files are generated in a new folder and must be run there.

I think this is completely fine! I realize why I could not run https://github.com/OpenSourceBrain/262670/blob/master/NeuroML2/KC.net.nml: I was launching it directly from the CLI without setting include_includes=True, so the include paths were wrong in that case.

[sanjayankur31]

Cool, do you think it's OK to merge the PR?

[HengyeZhu]

No problem!

[sanjayankur31]

@all-contributors please add @HengyeZhu for code, design, review

[allcontributors]

@sanjayankur31

I've put up a pull request to add @HengyeZhu! 🎉