Currently install/compile of the package takes some 3 min. on my notebook. Most of the time is spent in the initialisation of the Element_M structs in the line
elements_arr = Vector{Element_M}(Element_M.(els_data)) # no idea why not working without this type casting
of the file Mendeleev.jl/src/Mendeleev.jl
I suppose the long compilation time is due to the type instability because of the missings - almost every 70-elements-tuple has a different type. One could use NaNs, empty strings etc. instead. Or probably the loop parallelization could reduce the compile time to something more acceptable like 1 min.
Currently install/compile of the package takes some 3 min. on my notebook. Most of the time is spent in the initialisation of the
Element_M
structs in the lineelements_arr = Vector{Element_M}(Element_M.(els_data)) # no idea why not working without this type casting
of the file Mendeleev.jl/src/Mendeleev.jlI suppose the long compilation time is due to the type instability because of the
missing
s - almost every 70-elements-tuple has a different type. One could use NaNs, empty strings etc. instead. Or probably the loop parallelization could reduce the compile time to something more acceptable like 1 min.