[jpbrodrick89]

Relevant issue or PR

• Add UI validation support for ranged numerical inputs #3

Description of changes

This PR parses the pydantic keys minimum, maximum and multipleOf and applies them as constraints in number_input through min_value, max_value and step.

Testing done

Added some relevant keys to mock schemas. I also reformatted them to make them more maintainable and human-readable.

I have not yet done manual testing.

@jacanchaplais I'd appreciate your input on how else we can test this, should we have an example (mock) Tesseract that uses min and max values?

I'd also like to default to using a slider (w/ synced number input if possible) when min and max is provided, but thought I'd leave that for a future PR. Do you have any thoughts on whether that's a good idea?

[jacanchaplais]

Wow @jpbrodrick89, this looks amazing, thank you! I've added some comments, a nit, and a suggestion for some cleaner data structuring, hope that's alright. :)

As for the slider, would this respond well to logarithmically scaled data? I also get a bit wigged out with sliders if the resolution needs to be precise, and I can't hit the number I want haha. Let me know what you think! :)

[jacanchaplais]

Also @jpbrodrick89 I like your idea of adding a Tesseract with min / max values. You can use Streamlit's + Typer's testing frameworks as used in test_cli.py to spin up the Tesseract, generate and run the app, then try adding values outside of the numerical bounds.

[jpbrodrick89]

Thanks for the quick review @jacanchaplais , it seems that UI functionality can't really be tested with .set_value() as this just overrides any constraints. I checked manually and it seems to be working for min and max. It does not work out of the box for step but I think we should leave any custom validation of that for a future PR as quite complex.

[jacanchaplais]

@jpbrodrick89

it seems that UI functionality can't really be tested with .set_value() as this just overrides any constraints.

What about iteratively calling the .increment() / .decrement() methods until the bounds are reached, as documented here?

It does not work out of the box for step but I think we should leave any custom validation of that for a future PR as quite complex.

As in the step doesn't work when you manually test in the web UI? Sounds like the feature is just broken, then? If so, we should either investigate why or remove that feature, since we don't want to introduce non-operational code.

Testing-wise, you could probably also use the .increment() / .decrement() methods along with the .value attribute to check the step, right?

[jpbrodrick89]

I mean step correctly steps up in the correct increments but it doesn't perform validation if you manually input incorrect values. I'll try the increment/decrement methods you recommend though to toughen up the testing.

[jpbrodrick89]

Unfortunately increment and decrement override min_value and max_value too 🙁 ... but not step so added that. I think this is as good as we can get.

[jacanchaplais]

Sorry for the slow reply @jpbrodrick89! LGTM, thanks :)