James Kermode
James Kermode
Sounds like a good plan
A constrained bond length would be the first canonical example. I will set something up and let you know if/when there are problems.
Here is an attempt at using `ConstrainedOptim.jl` to fix the length of a bond in a piece of bulk SW silicon with latest JuLIP. The optimiser converges and penalty goes...
Think I've got it working, mistake was not zeroing all elements of constraint Jacobian: ```julia function con_jacobian!(J, x) J[1,:] = 0.0 J[1,I1] = (x[I1]-x[I2])/blen(x) J[1,I2] = (x[I2]-x[I1])/blen(x) end ```
No, slightly confusingly the constraint Hessian contribution is additive, so that goes on top of the one from the objective, while the constraint Jacobian needs the complete vector.
Thanks Christoph- working fine now, so no need to look at, but agree it would be good to implement a generalisation of the function into JuLIP in a more convenient...
I'd also be in favour of including up-to-date versions of ASE.jl and JuLIP.jl in the General registry. My impression is they are no longer changing that rapidly so this should...
I'm sure you're right Bob, but as you say this is a lot of work. When I started f90wrap I could not find any Python implementations of Fortran parsers that...
@jlkittle Thanks for the feedback. If you post some examples of subroutines that don't parse I'll take a look. Pull requests also very welcome :-)
This should work as you expect, I’ll take a look. Thanks for reporting!