Skillmon

That change (adding `\relax`) breaks `collcell` (just pointing it out). Well, breaking is not really correct, `collcell` now "just" needs to strip the extra `\relax`, still requires a change by...

@u-fischer just take a look at the contents, it starts with the `\relax` that was injected. So it doesn't outright break, but depending on what you intend to do with...

@josephwright yes, I think that would be great (sorry I didn't make it, didn't find the time yesterday evening -- not because I was busy or something just got distracted)

This will have `\prg_generate_conditional_variant:Nnn` have to guess from which base to derive the alternative branching version. So your code will have to check whether the "correct" base-variant exists, and if...

@PhelypeOleinik Similar to what @Witiko suggests, the different branching versions can just be generated with `\use_i:nn` and friends appended to the found branch-version. Since `\prg_new_conditional:Nnn` might add optimisations you'd have...

> However, there does not seem to be a straightforward way to convert e.g. the variants `T` and `F` to `TF` if you treat the functions as black boxes. Assuming...

I'd say if you do something like this it shouldn't be editor specific, so EditorConfig might actually be a good idea. But an Emacs specific file would then also call...

@ysalmon the wording might be misleading a bit, but it is documented for `\tl_if_novalue:nTF`: > Tests if the ⟨*`token list`*⟩ is exactly equal to the special `\c_novalue_tl` marker. doesn't mean...

> Not as a replacement for the current l3expan functions (which I agree would be much slower), but as some special-purpose additions so that more things become expandable. For example,...

This is roughly what `beamer` does (but overzealous by removing all key=value options from the global options list), which led to unexpected behaviour in the past. I'd say that a...