Re: Holes in article sequence

["François Pinard" <pinard@iro.umontreal.ca>]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> Hrvoje Niksic <hniksic@srce.hr> writes:

> > Now, if you have a big hole in your article range -- e.g. your article
> > range is 128 - 12300 because you ticked article no. 128 -- total
> > expiry will stat tens of thousands of non-existant files every time 
> > you leave the group.  That tends to get a bit slow.

> One could speed things up by having the backend look at what articles
> exist in the group, and then iterate over that instead of the other
> way around.  That would be much, much quicker for groups with big
> "holes" in them.  It might be slower in other situations, though.

If this is the only problem with big holes, it would be worth moving
iterating the other way around, and forget hole problems forever :-).

If you fear that it could a bit slow, but are able to keep the old code as
well as the new, maybe you could let the number of articles decide which
method to use, or else, have some collaboration from the backend to tell
you when it is better to change methods.  Or maybe both: if the backend has
nothing to say, use some heuristic on the number of articles.  Of course, if
the other-way-around is not supported by the backend, just use the old way.

-- 
François Pinard   http://www.iro.umontreal.ca/~pinard