At 2022-07-24T11:35:45+0100, Ralph Corderoy wrote: > > You can _do_ the table of contents in one pass; relocating it to the > > front matter requires either a second pass or a post-processing > > tool. An index is a much bigger pain because it has to be sorted. > > I have idly wondered from time to time if one diversion after another > could be created, each holding a page's content without any pages > being emitted, and then the diversions printed in a different order to > their creation.
It seems like it should be possible, by opening diversions as the first thing in a head-of-page trap (`PT` in ms) and closing them as the last thing in a foot-of-page trap (`BT` in ms). I would predict some problems because various tools like tbl(1) behave badly because they expect to create diversions themselves that aren't already nested. > Memory bloat would be an obvious downside, I predict that any such bloat would look svelte indeed compared to Firefox or Chrome running on the same machine. > but it's more of an experiment ...one I still wouldn't mind seeing done, if we have an enterprising member of the mailing list... > than a serious solution as re-arranging the pages is often > straightforward and some post-processing already required for other > reasons. I think a more practical approach would be to do the page rearrangement in the "grout" domain (the output of troff(1)), and a while back I filed a ticket to this effect. https://savannah.gnu.org/bugs/?61836 Regards, Branden
signature.asc
Description: PGP signature
