Andrew Williams

Hi @ddbkoll ! I'll take a look now, it could well be something wrong with my numba interface (still getting used to it). +1 for the tests!

@ddbkoll very weird! I thiiiink this might be to do with how numba deals with `global` instances, see [here](https://numba.pydata.org/numba-doc/dev/developer/caching.html#caching-limitations). > Global variables are treated as constants. The cache will remember...

Sweet! Yeah I recognized a few of the functions from PyTran actually. I'll start thinking about a clean-up this week! (Need to make some lists of what's being used) Another...

Interesting! Thanks for the overview! Maybe I'm missing something here, but in `DATA/HITRAN_DATA/HITRAN2016/ThermalOnly_0-5000cm.MainIsotopesOnly` the files are all quite small? for example `01_hit16.par` is only 5.3Mb? I'm a bit confused by...

Another option could be to call `loadSpectralLines()` in the `OpticalThickness` calculation, before it loops over the temperatures/pressures. Currently it loads in the same spectral lines every iteration of the loop....

Nice idea! I'll look into that now In terms of tests I was thinking primarily of checking that the assertions fail correctly (if the user gives an incorrect kernel name)...

@duncanwp I've made a small change to `_get_gpflow_kernel` so that it can also return the list of individual kernels if asked to. Now calling this in the `kernel_plot()` function rather...