Question about graphics units updating

[David Bateman]

I've been addressing the bug report #30519 as it is necessary to have 
the legend code working correctly. I started implementing a fix where 
the position, outerposition, etc values are stored in a particular unit 
and the get_position, set_position, etc function handles the conversion 
to the desired units. As I started with the axes and worked down to the 
root_figure class while implementing this, I didn't notice that the root 
windows screensize property is already treated, but it instead calls the 
update_units() method in the set_units() method to natively store the 
value in the desired units. So that the get_position, set_position, etc 
methods don't need to be modified.

Whether one method is better than the other seems to me more a question 
of taste, and I have essentially all the other unit conversion functions 
written. However I find it displeasing to use two different techniques 
to perform the same manipulation of the graphics properties. Is there 
any reason to pick one method over the other? If so, which? If we can 
accept that there are two different means of performing this 
functionality depending on the graphics call then I'm almost ready to 
push a changeset.

D.