matrixlabtools / potentialcalculation.jl Goto Github PK

Add comprehensive testing to all functionality.

There are some inconsistencies in samplin clusters that could be cleaned. In example, there are several sampling methods that could be done with one method using multiple dispatch.

• load_clusters_and_sample_input
• load_clusters_and_make_input
• load_clusters_and_sample_input

Insted of these 3 there should be on one sample-method that would do everyting

Use Documenter.jl

Implement cluster rotations with Rotations.jl.

With this it's easy to add rotations around arbitrary axis and and also cleaner random rotations.

There are some cases outside clusters.jl that call .xyz directly. It would be good to get rid of these

Restart file get save only at the end of calculation, which is not the intended behavior.

Logging does not flush atm and that mean there is no information of how calculations are going. This is problematic when trying to get things going on supercomputer.

Here is a related discussion.

calculate_adaptive_sample_inputs gets stuck when sampling only one line.

https://github.com/PainterQubits/Unitful.jl

As title says.

This package should not depend on JLD2, but removing it causes tests to fail.

As is stands now a lot of not needed low level stuff is exported.

It would be good idea to reorganize exports so that default export only exports high level stuff, like content of restarttools.jl and some other needed functionality, and low level stuff would located eg. at PotentialCalculation.LowLevel or not exported at all.

Sampling has problems in distance calculation. At this moment distance calculation adaptive_line_sampler is not ok and needs to be checked. Now it does not depend on actual distance between molecules, but on how much molecule2 has been moved. This is not the intended behavior.

Do calculations with PSI4.

Practical implementaition will have to use PyCall.jl and some refactoring of calculator implementation is needed.

As title says. Thise are need for fitting complex potentials ( https://github.com/tjjarvinen/PotentialFitting.jl )

TODO

• Distance between two atom in different clusters via index and unitrange
• Angle between atoms in same cluster
• Angle between two atoms in cluster and external point
• Dihedralangle

Restart files are saved, but they appear to be at random times. Escesially first restart file is saved later than expected.

Documentation of structures does not include documentation examples of how to create them. It would be good to add them.