On 4/1/19 12:52, Benjamin Berg wrote:
> On Mon, 2019-04-01 at 09:41 +0200, David Kastrup wrote:
>> Benjamin Berg <[log in to unmask]> writes:
>>> Hi,
>>> On Sun, 2019-03-31 at 18:57 -0300, Phelype Oleinik wrote:
>>>> If you can't switch the catcode of & before looking for the
>>>> optional argument, as David suggests
>>> David has a good point, I could handle the optional argument parsing
>>> myself and ensure that the case works correctly. It may not be trivial,
>>> but it should be reasonably easy.
>> It should be trivial.  You just use the provided mechanism exactly as
>> you did before (though naming it differently) and call it last thing in
>> your argumentless wrapper environment after setting up catcodes.  That
>> way the argument scanning only commences once catcodes are settled.
> Well, I do have an optional parameter that disables the catcode changes
> (which I know realize might have been a bad idea). That said, it would
> still be rather safe to make the change, parse the optional parameter
> and undo the change again afterwards when desired.
> Or maybe I should just bite the bullet and create two almost identical
> environments for this.
> Benjamin

Other option: use an xparse environment with "o" signature as you are
doing, then end the code with

    \peek_charcode:NT & { ... } { ... }

to detect "&" and replace it by an active "&" (or rather by the
definition you have for it.