Will Robertson writes:
 > On 04/01/2011, at 12:23 AM, Joseph Wright wrote:
 > 
 > > The resulting documentation might read
 > > 
 > > %  Evaluates the \meta{integer expression}, expanding any
 > > %  integer and token list variables within the \meta{expression}
 > > %  to their content (without requiring \cs{int_use:N}/\cs{tl_use:N})
 > > %  and applying the standard mathematical rules. This process requires
 > > %  two expansions. The result of the calculation is an
 > > %  \meta{internal integer} which should be treated in the same way
 > > %  as a \texttt{int} variable, \emph{i.e.}~it must be prefixed
 > > %  by \cs{int_use:N} unless used in a context which requires an
 > > %  \meta{integer expression}.
 > > 
 > > (with similar statements for \dim_eval:n and \skip_eval:n). Is this
 > > sufficiently accurate and clear? Does the entire proposal make sense? 
 > 
 > I *think* this is sensible, but I'd be surprised if this wouldn't require
 > some extra changes in some of the internal expl3 functions. The fact that
 > \int_eval:n is expandable is probably exploited in a few places (at least
 > in \int_compare:n, anyway). But I'm sure this isn't an insurmountable
 > problem. 

you can bet that making this non-expandable will break the kernel in many
places. I guess this is essential to a
lot of code that Morten has written. Perhaps you could take \number out and
put it in front in all such places, but whether that is better I wonder.

Morten are you listening right now?

 > There's not really a sensible alternative for \glueexpr is there? It would
 > be fine to have \number before the \dimexpr version as well, but without
 > consistency between all of them we should definitely (er I think) choose
 > the change that you're suggesting. 

are we sure the \the isn't the right thing to apply before \glueexpr, ie
Joesph's original thought? I rather think that would be better than taking out
exansion from \int_eval:n

frank