search

Jack-H-Buckner / UniversalDiffEq.jl

Universal differential equations for ecologists
MIT License
3 stars 0 forks source link

Build light version with core functionality and fewer dependencies #52

Closed Jack-H-Buckner closed 1 week ago

Jack-H-Buckner commented 1 week ago

I have been having trouble with precompilation issues I want to see if this helps.

jarroyoe commented 1 week ago

I think @19fitzna has done a lot of work figuring out dependencies in the toml file. The main compiling issue I see is the documentation, which I'm not sure what the pathway is to solve.

19fitzna commented 1 week ago

I think @19fitzna has done a lot of work figuring out dependencies in the toml file. The main compiling issue I see is the documentation, which I'm not sure what the pathway is to solve.

The issue with the documentation is due to #48. We just need to add the docstrings in the code somewhere in the documentation. I can do this, but it might be better if the people who wrote the functions provide the descriptions just to ensure accuracy.

19fitzna commented 1 week ago

I have been having trouble with precompilation issues I want to see if this helps.

I wonder if it would potentially be better to do the opposite? We strip this package down to its core and move the extra stuff to additional packages which act as plug-ins?

jarroyoe commented 1 week ago

I have been having trouble with precompilation issues I want to see if this helps.

I wonder if it would potentially be better to do the opposite? We strip this package down to its core and move the extra stuff to additional packages which act as plug-ins?

This could be an option. Bayesian UDEs could be a separate package, but I wonder if at this point that would be an issue of splitting the audience too much.

Jack-H-Buckner commented 1 week ago

I just made this branch because I was having some trouble getting the package to precompile on my machine and wanted to see if a lighter version would help (it didn't). I think we should move forward with all the current features. I think they are useful and the pacakge only takes a minute or two to precompile on my machine.

On Wed, Jun 26, 2024 at 10:57 AM Jorge Arroyo-Esquivel < @.***> wrote:

I have been having trouble with precompilation issues I want to see if this helps.

I wonder if it would potentially be better to do the opposite? We strip this package down to its core and move the extra stuff to additional packages which act as plug-ins?

This could be an option. Bayesian UDEs could be a separate package, but I wonder if at this point that would be an issue of splitting the audience too much.

— Reply to this email directly, view it on GitHub https://github.com/Jack-H-Buckner/UniversalDiffEq.jl/issues/52#issuecomment-2191927599, or unsubscribe https://github.com/notifications/unsubscribe-auth/AOM6CCODRR7DIFXFFAJ7BMDZJLJGHAVCNFSM6AAAAABJ2HNMPKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCOJRHEZDONJZHE . You are receiving this because you authored the thread.Message ID: @.***>

zechmeunier commented 1 week ago

The precompilation for UniversalDiffEq and all the other packages in my reef analysis workflow only takes a couple minutes on my computer as well.

jarroyoe commented 1 week ago

As a clarification note, it would be hard to reduce the precompilation of the package by a significant amount because most of its dependencies (Lux and Plots come to mind) already take a long time precompiling.