Nicolas <n.goaz...@gmail.com> wrote:
> Hello,
>
> Nick Dokos <nicholas.do...@hp.com> writes:
>
> >> The only solution I could think of is somewhere in the patchwork server
> >> ("Auto-fill shouldn't insert new items").
> >>
> >
> > Are you saying that the patchwork server introduced this? Or that the
> > patchwork server should solve it? In the case I stumbled upon (a Worg
> > file), the patchwork server is nowhere in sight, I think: the change was
> > (I'm guessing here) directly committed, so maybe the original writer had
> > auto-fill turned on, but maybe not and just typed it that way. So
> > neither the patchwork server nor making auto-fill smarter would provide
> > a guaranteed solution. But maybe there is no such thing.
>
> I'm not sure to understand. I mean that there is a patch waiting to be
> processed in the patchwork server, that should get us as close to the
> solution as I can think of (so, probably not very close).
>
Ah, OK - I was just dense: I misread it completely.
> It is true that the patch only solves the problem if it would happen
> with auto-fill (or fill-paragraph). On the other hand, if the original
> writer typed it that way, it's an user problem, and I don't think we
> should add cruft to solve it.
>
> > Perhaps a heuristic in html export to just take care of the error: if
> > it looks like a numbered list item but there is no item body, bail and
> > assume it's part of the text.
>
> I introduced support for empty items on purpose. That would be
> a regression.
>
> > Or a combination of that together with the item number > 1. Would that
> > work?
>
> It would partly work. But:
> - You would get false positives (rarely, but still),
> - It wouldn't handle case when line, beside the bullet, isn't empty,
> - It isn't only about HTML export, not even about export, as the same
> situation happens in Org buffers.
>
> Again, preventing the problem to happen automatically, i.e. without the
> user knowing for sure about it, is taken care of by the patch (provided
> it doesn't break anything else, thus its journey on the patchwork
> server). In every other case, we can assume that the user knows (should
> know) both Org syntax and what he is doing.
>
> Now, for already introduced problems, it is indeed possible to build
> a function using heuristics to scan files, and when matching, offer to
> join lines and fill-paragraph. I don't think there are enough cases to
> justify this, though.
>
OK: agreed - we just fix existing org files on a case-by-case basis.
So manually inserting "14." at the beginning of the line is user error,
auto-fill is taken care of by your patch. However, there is still the
problem that an empty item in a numbered list will trigger the error
that I posted, so org-html-export-list-line will need to be taught how
to deal with an empty list item - correct?
Nick
