I think the way we are managing this repo (in fact any PyCyto repos) are not obvious to newcomers. We should have a more thorough documentation on this, and provide a step-by-step guide on how people can meaningfully contribute to our project. Our current page offers some good guidelines, but it doesn't reflect the structure of the project. I will work on this once I have time.

Currently, the cytomulate.utilities.linear_function returns a function that does not support a list input as t. However, the Akima1DInterpolator does.

To reproduce the error

import cytomulate
import numpy as np

func =  cytomulate.utilities.linear_function(1, 2)
func(0.1) # Okay
func(np.array([0.1, 0.2]) # Okay
func([0.1, 0.2]) # Error

Behavior of Akima1DInterpolator

import cytomulate
import numpy as np

func =  cytomulate.utilities.smooth_brownian_bridge(1, 2, sigma2=0.5)
func(0.1) # Okay
func(np.array([0.1, 0.2])) # Okay
func([0.1, 0.2]) # Okay

Related Issue

The return type of cytomulate.utilities.linear_function now depends on the input. Do we want to make it return an array regardless?

Solution

Depending on what the intended use cases are, we can do one of the following:

• Check input type and cast the list to an numpy array or throw an Exception
• Make our function consistent with the Akima1DInterpolator

Our typing and documentation need to reflect the changes according.

We need to update the links on the documentation to reflect all the changes as well as the newest publication details. We should be able to get this done quite easily.

What feature would you like to propose?

It may be time to update our CI pipelines to include newer versions of Python: at least 3.11. This will ensure better compatibility going forward.

How do you envision this feature to be implemented?
We can just try it and see if there are problems.

Will all our random number generation occur with numpy? At some point, it will be helpful to implement a global seed (or at an appropriate entry point of our package), especially for predefined, reproducible workflows. We will also think more carefully how this should be tested because we don't want to break user space once this is implemented (i.e. Backwards compatibility and reproducibility).