apjanke / neuropixel-jankify

Project management for Andrew's 2021 Neuropixel Utilities enhancement project

Neuropixel Utils Jankification

This is a project management repo for Andrew Janke's 2021 project to enhance the Neuropixel Utils Matlab library.

This repo will contain work products, notes, project management stuff, and other things related to the project, but which are not going in to the repo fork for potential upstreaming.

This repo is mostly a GitHub Pages site. View it online at https://apjanke.github.io/neuropixel-jankify.

The repo for this project's actual code is at https://github.com/apjanke/npxutils-apj-wip-01.

This project is supported by MathWorks.

Acknowledgments

Thank you to MathWorks for sponsoring this project.

Thank you to Phish and Rosé for powering my coding.

Thank you to the Matlab Discord for helping me out when I get stuck.

neuropixel-jankify's Issues

Make sure workflow still works after code org and API changes

Ensure reproduced workflow (item #1) and live script (item #2) continue to function following changes via items #3, #4, and #5.

Use and enhance npy-matlab

On a separate fork, implement (that is, switch to using and being a client of) kwikteam/npy-matlab: via subtree to develop a proof-of-concept for discussion with the npy-matlab owners. npy-matlab is experimental code to read/write NumPy .NPY files in Matlab.

Introduce function argument validation

Apply the new MATLAB capabilities for function argument validation using argument blocks and declarative property constraints across at least two functions where inputParser is currently used (e.g. in runKilosort2).

Streamline readXXX functions

Streamline some of the readXXX functions, e.g. to use string array processing.

Tighten up public class API

Implement class organization updates in at least two classes towards the goal of a smaller and more inspectable public API for class consumers/library users. This is expected to include changes such as:

1. Demarcation of the public API from the hidden/private API in the class file, e.g. the public API before the class internals
2. Initial deprecation steps (e.g. hidden implementation for trivially coded (‘one liner’ or similar) public properties, e.g. nChannels which can be determined by class consumers simply via length(channels_id).
3. Demarcation of groups of related properties (e.g. derived/computed values) through a property naming strategy (e.g. prefixes) or the creation of a new class linked to existing class via composite pattern

Live script version of reproduced workflow

Create a Matlab Live Script version of the workflow from #1.

Code tree reorganization

Implement code organization updates towards the goal of fewer files per folder (package folders or otherwise). For instance:

1. In Utils package: consider separating out the pure coding utilities from the domain-oriented npix utilities
2. In root package: consider separation of some/all functions from the classes, perhaps via an additional sub-package
3. In Neuropixels package: consider separation of related function such as for file reading/writing/exporting (e.g. readNPY & writeNPY) & consider whether some functions (e.g. runKilosort2) can become public methods of classes defined in the package

Reproduce workflow from example walkthrough

Reproduce the full workflow described in the wiki example (https://djoshea.github.io/neuropixel-utils/#example-walkthrough) which spans 1) IMEC file load, 2) IMEC file investigation, 3) Kilosort2 run, and 4) Kilosort derived data load and visualization.

This will need to be done in a way that can be run in a modest home lab, and should include all the lab setup code.

Work product submission

Submission of all Contributions to the Project hosted at the Public Hosting Site (GitHub, with upstreaming to https://github.com/djoshea/neuropixel-utils) under the License Terms (MIT License).