System stuff
- Did someone neglect to tell me that this stuff only runs on Linux? Kilosort requires CUDA so Mac is out, and here’s a
readlink call in makeSymLink.m and that’s not going to work on Windows, so, Linux it is?
Naming and prose style
- Do you prefer “data set” or “dataset” when talking about IMEC and Kilosort data sets? (“dataset” might be confused with the Matlab
dataset type/class.)
- Do you prefer “IDs” or “Ids” or “ids”?
Code
- You’ve got some calls to
debug() scattered around the example code. It’s broken because that function doesn’t exist. Would you like me to add some real logging support?
- I could do either a simple native-Matlab thing with just “info” and “debug” and a central on/off switch for the debug output.
- Or I could hook you up with a full logging framework that’s built on one of the established Java logging frameworks, and gives you lots of extra control and output options.
- Do you want to try to be Mlint-inspection-clean?
- Can we pin down a code style?
- 4 space indents, I guess?
- How long do you want your lines to be? 80, 90, 100 chars? I see a lot of long non-wrapped lines hanging around.
Example code and data
- The workflows in the examples and tutorial are non-idempotent! They write back to the original input files. This means they’ll produce different results if run more than once, and would produce changes that could get checked back in to an example data set repo.
- About paths: A lot of the paths in the example code use
/data. That’s a root-controlled mount point; many users are unlikely to have it or have access to create it. I think the example code should all use paths under the user’s home directory as a default, and preferably factor all these out to a central configuration point.
Miscellaneous
- Requires a Linux box (and a bare-metal one at that). Let’s add Windows support!
- CUDA = NVIDIA, so no Macs for Kilosort. And npxutils uses Unix symlink commands, so no Windows.
- Kilosort2.5 is out, and Kilosort3 is under development. Want to add support for them?