Gaurav Arya

I just did, and unfortunately that fails too. Looking at the code, https://github.com/JuliaDiff/ChainRulesTestUtils.jl/blob/4050989ccdea6766f6546e1e2bcb2d77f36c6733/src/finite_difference_calls.jl#L23 it seems like the only influence of the tangent on the finite difference call is whether to...

Any pointers on doing the necessary type piracy? For `test_frule`, I think I would need to overload `to_vec`, but what should the overload be so that the check will go...

For posterity, here's what I needed to do to get `test_frule` and `test_rrule` to work on a structure with a non-differentiable, troublesome nested field (in this case of type `InnerPlan`):...

It doesn't *necessarily* have to be type unstable. For example, you could have a boolean `valid` bit in a Dimensions object, with `valid = 0` in this case. I did...

Cross-referencing https://github.com/SymbolicML/DynamicQuantities.jl/pull/42#issuecomment-1726343281: changing `one` has the potential to make a function like `Base.prod` type unstable. This wasn't a problem with Unitful because something like a multiplicative reduction on an array...

To clarify the key problem: the question is, when the user writes something like `u"fm"`, what should the resulting quantity look like? Since the quantity will be represented in SI,...

Hey, what's the status of this? It could be good to merge soon-ish to minimize the amount of code in the old style we'll have to adjust

Regarding Woodbury, see bottom of page 3 here: https://arxiv.org/pdf/2309.03060.pdf

If `L` is a lazy operator, we might want to do `A = convert(AbstractMatrix, L)` because cerrtain operations might be faster with the matrix. But written like I just did,...

yes, happy to since I've already looked at a lot of the operator code there. Would need a merge of #143 though