Open junglegobs opened 3 years ago
Note that most of SpineOpt is lazy loaded using Requires.jl. That means the first run is slower than normal. Can you please confirm that the second run is also that slow?
Yes it is. I need to investigate more however. I put it here as a placeholder in case someone eventually complains about this.
In fact now I've noticed it even gets stuck at the preprocessing step. Not sure what this could be.
EDIT: it seems to take ages simply creating the the representative periods from SpineOpt.time_slice()
EDIT: This seems to have been fixed by removing the start and end date of the "master" temporal block (which I call 2017). Not sure why.
This may be simply due to the having to interact with a database, but these can easily take ~ 2 minutes compared to the 30 seconds to solve the ordering problem. Would be nice to speed this up somehow.