On 13/02/2010, at 11:37 AM, Philipp Stephani wrote:

> expl3 itself uses \exp_arg:x in disguise, e.g. in l3tl.dtx, l. 1504:
> 
> \cs_set_nopar:Npn \tl_tmp:w #1 {
>  \tl_set:Nx \l_tmpa_tl {
>    \exp_not:N \prg_new_conditional:Npnn \exp_not:c {tl_if_eq:#1}
>      ####1 ####2 {p,TF,T,F} {
>        \exp_not:N \tex_ifnum:D
>        \exp_not:c {tl_compare:#1} {####1}{####2}
>        \exp_not:n{ =\c_zero \prg_return_true: \else: \prg_return_false: \fi: }
>    }
>  }
>  \l_tmpa_tl
> }
> 
> Codes like that are very common. I also think that this is not too elegant, but in many cases there is no other way. \edef lets you specify completely arbitrary combinations of expansions, something that is impossible otherwise. It doesn't matter that it is not expandable because the result must be executed anyway.


I've proposed adding this version of \exp_arg:x to expl3 relatively recently.
(It's the first on the list of l3candidates.dtx.)

Your feedback helps us to decide whether it should be included.
Any thoughts from others?

-- Will