Enhanced virtual groups

Enhanced virtual groups

[Norman Walsh]

[-- Attachment #1: Type: text/plain, Size: 731 bytes --]

This is mostly a curiousity question. One of my colleagues was
describing a feature of Evolution that he really finds useful: the
ability to have virtual folders based on criteria more selective than
gnus virtual groups (as I understand them).

In particular, he has a virtual folder for "all mail less than 24
hours old" and another for "all mail addressed to him".

Anyone thought of implementing something like that in gnus? Or maybe
it's already been done. Wouldn't be the first time :-)

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | A polar bear is just another way of
http://nwalsh.com/            | expressing a rectangular bear.

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: Enhanced virtual groups

[David Edmondson]

* ndw@nwalsh.com [2004-07-16 14:52:23]
> Anyone thought of implementing something like that in gnus? Or maybe
> it's already been done. Wouldn't be the first time :-)

If you use nnimap then a small twist to nnir allows you to use
arbitrary IMAP SEARCH expressions, which would get you there.  Doesn't
help for nnml, etc. though :-(

dme.

Re: Enhanced virtual groups

[David Moore]

Norman Walsh <ndw@nwalsh.com> writes:

> This is mostly a curiousity question. One of my colleagues was
> describing a feature of Evolution that he really finds useful: the
> ability to have virtual folders based on criteria more selective than
> gnus virtual groups (as I understand them).
>
> In particular, he has a virtual folder for "all mail less than 24
> hours old" and another for "all mail addressed to him".
>
> Anyone thought of implementing something like that in gnus? Or maybe
> it's already been done. Wouldn't be the first time :-)

I've been thinking about just this problem recently, and am likely to
modify nnvirtual to allow these sorts of things more easily/efficiently.
The poor man's way of doing some of it now is to have a big nnvirtual
group and using limiting to narrow things down.

I've also been thinking of adding "labels" which are sortof like more
generalized marks, ala gmail, and mixing that with nnvirtual also.

If people have nnvirtual ideas, now is a good time to mention them.

On my nnvirtual thought list:

1. combined nnvirtual with initial limit setting.
2. nnvirtual for `nnmail-split-history'
3. combined nnvirtual with various user pre-select functions
   (probably how 1&2 actually get done)
4. labels
5. redecide which marks copy and don't copy into and outof virtual
   groups back to the base groups, since new marks have been added
   since nnvirtual was last really touched.


-- david

Re: Enhanced virtual groups

[James Leifer]

David Moore <dmoore@caida.org> writes:

<snip>

> I've been thinking about just this problem recently, and am likely to
> modify nnvirtual to allow these sorts of things more easily/efficiently.
> The poor man's way of doing some of it now is to have a big nnvirtual
> group and using limiting to narrow things down.

Hi David,

That's great that you're thinking of improving nnvirtual!

I'd be really happy to have google gmail (or evolution) like
functionality for gnus to handle my 100,000+ messages.  Here are some
ideas on what we would need in order to do this.

The specification of the contents of a virtual group is rather simple:

   (
     ("mygroup1" (100 101 (102 . 107)))
     ("mygroup2" (20506))
   )
   ...

i.e. a list of pairs, where each pair consists of 

* an underlying group name;

* a list of article numbers from that group.

There are a couple of important properties of the constructed virtual
group

* mnimal retrieval

  The virtual group should retrieve only the articles from the
  underlying groups that are requested; e.g. if we have a virtual
  group consisting of 5 articles from my inbox of 50K articles, the
  retrieval should be almost instantaneous.

* mark pulling

  The virtual group's articles should inherit the marks of the
  underlying groups; e.g. if the virtual group includes article 31415
  from my inbox and this article was marked read in my inbox, it
  should be marked read in the virtual groups.

* mark pushing
 
  The virtual group should push mark changes back onto the underlying
  groups; e.g. replying to an article in the virtual group should push
  the reply mark onto the article in the underlying group.

This would give us some wonderful capabilities.  For example, we could
use indexing tools such as swish++ to rapidly assemble a list of
articles (spanning several underlying groups) which could be
aggregated into a virtual group.  Also, we could take any article in
the virtual group and "expand" it into a new virtual group to include
its complete surrounding thread (provided that external indexing tools
or the gnus registry could provide this data).

Comments?

-J

Re: Enhanced virtual groups

[James Leifer]

James Leifer <James.Leifer@inria.fr> writes:

> The specification of the contents of a virtual group is rather simple:

I meant of course "would be rather simple": I'm talking about a
proposed new version, not the existing one.

-James

Re: Enhanced virtual groups

[Ted Zlatanov]

On Fri, 16 Jul 2004, ndw@nwalsh.com wrote:

> This is mostly a curiousity question. One of my colleagues was
> describing a feature of Evolution that he really finds useful: the
> ability to have virtual folders based on criteria more selective than
> gnus virtual groups (as I understand them).
> 
> In particular, he has a virtual folder for "all mail less than 24
> hours old" and another for "all mail addressed to him".
> 
> Anyone thought of implementing something like that in gnus? Or maybe
> it's already been done. Wouldn't be the first time :-)

It would be pretty easy to use the Gnus registry to build a list of
messages that have the time or sender attribute as you describe (those
are already stored in the registry).  More attributes, e.g. size of
message, can be added very easily.

Ted