search

SciML / DiffEqOperators.jl

Linear operators for discretizations of differential equations and scientific machine learning (SciML)
https://docs.sciml.ai/DiffEqOperators/stable/
Other
285 stars 74 forks source link

Uninitialized field `opnorm` in JacVecOperator #519

Open danielwe opened 2 years ago

danielwe commented 2 years ago

opnorm is left uninitialized here:

https://github.com/SciML/DiffEqOperators.jl/blob/4b1bd632f717aee8be6f26601b37322f11f6b94a/src/jacvec_operators.jl#L24-L33

Another instance of JuliaLang/julia#24943

danielwe commented 2 years ago

Same issue in vecjac_operators.jl:

https://github.com/SciML/DiffEqOperators.jl/blob/4b1bd632f717aee8be6f26601b37322f11f6b94a/src/vecjac_operators.jl#L23-L32

ChrisRackauckas commented 2 years ago

These should probably just be deprecated and deleted in favor of the ones that are more updated in SparseDiffTools.

xtalax commented 2 years ago

Are we sure that they are completely obselete? I'd prefer to leave the operators up for now until we have stencil interfaces in MOL for the high point count problems and warn of future deprecation, with links to successor packages. Do MethodOfLines and SparseDiffTools cover all use cases?

ChrisRackauckas commented 2 years ago

Yes, these ones are not robust to alternative tags like SparseDiffTools is, so pretty much any use case should use that. There's a few things in here that aren't in there to match the AbstractSciMLOperator interface, but what we probably need to do is just make that interface package a lean package and then let SparseDiffTools extend that, add the one function, and then there's no reason for the operators here.