Joseph Wright
Joseph Wright
This is referring to what is in the grand scheme of things a tiny optimisation: hand-tuned versus auto-generated commands. I think we are best just dropping this in the docs:...
I've re-worked that list: see latest check-in
> Also "should come last" seems too strong, after all kinds of `\exp_args:N` are provided by `l3kernel` directly. The text is specifically about the cases where optimisation is really crucial,...
> The new description ([66d321d](https://github.com/latex3/latex3/commit/66d321db10ce36446e1c2778c16fc9262ac0043b)) seems too general. Handling unchanged `n`-type argument (before changed arguments) is slower than handling unchanged `N`-type. The latter needs only an `\exp_after:wN`, but the former...
> A rule too strict is a rule more often broken, hence has less practical guidance. Following current description a optimization fanatic may wonder why `\tl_set:Nn` instead of `\tl_set:nN`, but...
@FrankMittelbach I've added your text - hopefully this is clear enough
I think we've addressed this ...
This was missed > For example, you cannot test for 2 successive spaces with a construct like That's not a limitation of `peek`, rather TeX has tokenized as a single...
Better?
> Just a historical note: we used to have `\peek_catcode_ignore_spaces:N`, `\peek_catcode_remove_ignore_spaces:N` (now listed in `doc/l3obsolete.txt`), so the sentence probably dates from back then. Yes, it does: I think it was...