Support Pyodide

[fumitoh]

Pyodide allows Python to run in a browser. Suppporting Pyodide should make it possible for modelx to integrate into xlwing lite.
modelx is a pure Python package but there are parts that may be CPython specific. Test modelx with Pyodide an see if it works with modifications.

[declann]

Pyodide is a CPython port. I encountered no problems running either nomx models or modelx itself in Pyodide, so I'd expect xlwing lite to just work (not tried!).

It also permits an online interactive playground for modelx (independent of servers) - like the Actuarial Playground contains for calculang (though too subtle!).

Pyodide is an amazing project, and performs well in interactive loads that I tried and I'm testing.

I ported the playground calculang model to modelx and then to Python (via export/nomx). And using Pyodide I hooked it into the Actuarial Playground interactive UI. Performance is not super good but I guess this isn't the main concern for the usecase here, and I think performance will be really interesting to analyse/improve/compare (outside Pyodide there are so many more projects to get more efficient execution of Python code too - in native and browser/wasm land)

[declann]

A simple mostly vibe coded example of modelx+pyodide is in a single html file here:

https://preeminent-brigadeiros-ce64c7.netlify.app/messin-pyodide.html (clearly based on your call option model code)

• wait a few seconds for pyodide to download and code to run, then result appears
• then drag sliders

The hot/update path is the runMonteCarlo function: setting mx params, making a call for a new result, then JS gets that result (pyodideInstance.globals.get('result')) to put on UI.

[fumitoh]

Thanks for the note. It's good news that modelx works with Pyodide. I’ve also installed and tested xlwings lite with modelx, and I’m now wondering what actuarial use cases xlwings lite might be useful for.

[declann]

Models formalized in modelx and run/inspected in Excel.
Actuaries will also like this for setting up exercises; for those less familiar with Python environments its a much more accessible way to get started.

I've made a calculang addin for Excel too. It exports every formula in the model as a UDF. So very very flexible about what users can create and how layout/output can be arranged.

For modelx a UDF may take params and function signatures as inputs - with all that exposed, I guess you get interesting flexibility in layout/running exercises too.

[declann]

Unfortunately export doesn't work for me in Pyodide - code and error message below.

I tried older Pyodide to verify it's not a python 3.13 issue.

import modelx as mx
import numpy as np
model = mx.new_model()
space = model.new_space("a")

@mx.defcells(space)
def a():
    return 2

model.export('a_nomx')

Error: PythonError: Traceback (most recent call last):
  File "/lib/python313.zip/_pyodide/_base.py", line 597, in eval_code_async
    await CodeRunner(
    ...<9 lines>...
    .run_async(globals, locals)
  File "/lib/python313.zip/_pyodide/_base.py", line 411, in run_async
    coroutine = eval(self.code, globals, locals)
  File "<exec>", line 10, in <module>
  File "/lib/python3.13/site-packages/modelx/core/model.py", line 502, in export
    Exporter(self, path).export()
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
  File "/lib/python3.13/site-packages/modelx/export/exporter.py", line 90, in export
    SpaceTranslator(parent, io_manager).code,
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/lib/python313.zip/functools.py", line 1042, in __get__
    val = self.func(instance)
  File "/lib/python3.13/site-packages/modelx/export/exporter.py", line 185, in code
    class_defs=self.class_defs,
               ^^^^^^^^^^^^^^^
  File "/lib/python313.zip/functools.py", line 1042, in __get__
    val = self.func(instance)
  File "/lib/python3.13/site-packages/modelx/export/exporter.py", line 464, in class_defs
    defs.append(self._get_class_def(space))
                ~~~~~~~~~~~~~~~~~~~^^^^^^^
  File "/lib/python3.13/site-packages/modelx/export/exporter.py", line 481, in _get_class_def
    if is_lambda_expr(src):
       ~~~~~~~~~~~~~~^^^^^
  File "/lib/python3.13/site-packages/modelx/export/transformer.py", line 309, in is_lambda_expr
    return source.strip()[:6] == "lambda"
           ^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'strip'

When I call a.formula instead of export, I get:

Error: PythonError: Traceback (most recent call last):
  File "/lib/python313.zip/_pyodide/_base.py", line 597, in eval_code_async
    await CodeRunner(
    ...<9 lines>...
    .run_async(globals, locals)
  File "/lib/python313.zip/_pyodide/_base.py", line 411, in run_async
    coroutine = eval(self.code, globals, locals)
  File "<exec>", line 10, in <module>
TypeError: __str__ returned non-string (type NoneType)

and when I call a.source it's:

Error: PythonError: Traceback (most recent call last):
  File "/lib/python313.zip/_pyodide/_base.py", line 597, in eval_code_async
    await CodeRunner(
    ...<9 lines>...
    .run_async(globals, locals)
  File "/lib/python313.zip/_pyodide/_base.py", line 411, in run_async
    coroutine = eval(self.code, globals, locals)
  File "<exec>", line 10, in <module>
AttributeError: 'Cells' object has no attribute 'source'

So it seems like support in Pyodide might be a bit more complicated than I thought, unless this is a separate/environment related issue we can still control.

Do those errors ring any bells with you?

[fumitoh]

modelx uses getsource, getsourcefile, and findsource from the standard inspect library. Whether these functions work is interpreter dependent. So, modelx doesn't work with the default interpreter that comes with CPython 3.12 or older. It always works with IPython. CPython 1.13 comes with an improved interpreter, so modelx works with CPython 3.13 interpreter.
I guess that with Pyodide you used, the interpreter is still old even Pytho is 3.13, so it didn't work.

[declann]

Thanks for the info!