What's the status of this? Is it just that we still aren't sure about the Transformable instance? If so, can you explain a bit more what the issue is?

To be honest I just forgot about.

Concerning the Transformable instance the options are just fmap the transform:

instance Transformable t => Transformable (Measured t) where
  transform t = fmap (transform t)

or do something similar to how Text gets transformed so only local measures get scaled from transforms. Something like

instance Transformable t =>  Transformable (Measured t) where
  transform t = scaleLocal n . fmap (transform t')
    where
      t' = t <> scaling (1 / avgScale t)
      n = avgScale t

but I'm not sure how much sense this option makes.

Other than that I just need to document it and it should be ready to merge.

So to be clear, under the two different options for Transformable, if you e.g. created a measured circle of size output 30 and then scaled it by 2, what would happen?

Under the first option the circle would double in size. In the second option nothing would change (but if it was a measured circle of size local 30, that would double in size).

I see. I think I would expect the latter. The former seems very strange and not that useful. What does output 30 even mean if you can change it by scaling? Under that first instance, if you wanted a diagram to have a particular size in the output you would either have to (1) avoid scaling it at all, or (2) compute what "output size" you should specify so that after scaling it will have the real output size you actually want.

Why don't you think the second instance makes sense?

Just because it's something I came up with now, I haven't tested it. I'll try implementing it make sure it does what I think it does (I'm not sure about translations for example). I agree, if the second option does work it should be the one we use.

I guess what I want is

transform t (measuredDiagram md) == measuredDiagram (transform t md)

where md :: Measured n (Diagram b v n).