Adaptive scoring in nnml groups

Adaptive scoring in nnml groups

[Hans de Graaff]

For a while now I've noticed that no adaptive scoring occurs in my
nnml groups. At first I thought I missed some new feature, but a
fairly thorough search through the documentation didn't turn up
anything.

In my .gnus I have turned on adaptive scoring like this:

(setq gnus-use-adaptive-scoring '(word line)
      gnus-score-expiry-days 14
      gnus-summary-mark-below -500)

I don't have any 'adapt' settings in any of my SCORE files.

If I read the manual correctly, this would have to give me adaptive
scoring in all groups, but it only turns on adaptive scoring in the
nntp groups.

Could somebody point me to my obvious error?

Hans

Re: Adaptive scoring in nnml groups

[Richard Pieri]

-----BEGIN PGP SIGNED MESSAGE-----

>>>>> "HdG" == Hans de Graaff <J.J.deGraaff@twi.tudelft.nl> writes:

HdG> I don't have any 'adapt' settings in any of my SCORE files.

You need to explicitly enable adaptive scoring in mail groups, I think.

-----BEGIN PGP SIGNATURE-----
Version: 2.6.3
Charset: noconv

iQCVAwUBMlFLQp6VRH7BJMxHAQER1QQAypOSR1+OliaI6x3M7uF1tBUi+8GeknpV
imA99XeIw1SzE23QCKE7kvIYiOzSmgU8X+EYpXtMSUKG/z3TOVfXIheqDurImEap
BU57YzJ7bNVZVQyY63Mjh097kOU4cJyakXSE9ccq1vgUCmfBgLxPwYeOfKShcegX
sczDEwdvUvY=
=L3XO
-----END PGP SIGNATURE-----
-- 
Richard Pieri/Information Services \ Curiosity never killed anything, except
<ratinox@unilab.dfci.harvard.edu>   \ maybe a few hours. -A cat's guide to life
http://www.dfci.harvard.edu/         \ 

Re: Adaptive scoring in nnml groups

[Lars Magne Ingebrigtsen]

Hans de Graaff <J.J.deGraaff@twi.tudelft.nl> writes:

> If I read the manual correctly, this would have to give me adaptive
> scoring in all groups, but it only turns on adaptive scoring in the
> nntp groups.

Hm...  I couldn't find anything in the adaptive score code that would
make adaptive scoring be switched on only in nntp groups.  Are you
sure you're not switching it off in some hook or other?

-- 
  "Yes.  The journey through the human heart 
     would have to wait until some other time."

Re: Adaptive scoring in nnml groups

[Hans de Graaff]

Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

> Hm...  I couldn't find anything in the adaptive score code that
> would make adaptive scoring be switched on only in nntp groups.  Are
> you sure you're not switching it off in some hook or other?

I don't have any other references to adaptive scoring whatsoever. Not
in my ~/News directory, not in my .newsrc.eld.

I have experimented a bit, mainly by switching to
gnus-use-long-file-name. (This is a Solaris 2 system). I figured that
would get me started all new. But new newsgroup.ADAPT files only get
created whenever I leave an nntp group. Not when I leave a nnml group.

Maybe this is a weird and unexpected side-effect of something else?
I'm not sure where to start looking for solutions. Any hints?

Hans

Re: Adaptive scoring in nnml groups

[Wesley.Hardaker]

Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

> > If I read the manual correctly, this would have to give me adaptive
> > scoring in all groups, but it only turns on adaptive scoring in the
> > nntp groups.
> 
> Hm...  I couldn't find anything in the adaptive score code that would
> make adaptive scoring be switched on only in nntp groups.  Are you
> sure you're not switching it off in some hook or other?

I don't know if you rememeber or not, but I have been having this
exact same problem for a long time now.  I turned on adaptive scoring
way back in .20 or so of sgnus and starting using nnml later on.  I
wrote to you and told you about this problem, and you said something
to the effect of 'I can't repeat it using your settings', so I figured
I must be doing something wierd.

I had hoped that when I moved to CH and started anew (though I did
transfer some files, like my .gnus file and possibly the scoring files
as well, I don't remember.  I did zap my .eld file though) that this
problem would disappear.  Since no one else has ever complained about
it, I just figured it was me.  I'm incredibly relieved to see someone
else having the same problem (no offence Hans!).

Anyway, as an example I just checked the variable
gnus-use-adaptive-scoring in the 'ding' nnml group (which it doesn't
work on) and it is set to 't' as it should be.  All of my summary
lines, however, display the score as '0' (I actually display the
value).  Odd...

Note:
(setq gnus-kill-files-directory "~/News/scoring/")
"ls nnml*.ADAPT" shows nothing.

Wes

Re: Adaptive scoring in nnml groups

[Steven L Baur]

>>>>> "Wesley" == Wesley Hardaker <Wesley.Hardaker@sphys.unil.ch> writes:

Hans> If I read the manual correctly, this would have to give me adaptive
Hans> scoring in all groups, but it only turns on adaptive scoring in the
Hans> nntp groups.

Lars> Hm...  I couldn't find anything in the adaptive score code that
Lars> would make adaptive scoring be switched on only in nntp groups.
Lars> Are you sure you're not switching it off in some hook or other?

Wesley> I don't know if you rememeber or not, but I have been having
Wesley> this exact same problem for a long time now.  I turned on
Wesley> adaptive scoring way back in .20 or so of sgnus and starting
Wesley> using nnml later on.  I wrote to you and told you about this
Wesley> problem, and you said something to the effect of 'I can't
Wesley> repeat it using your settings', so I figured I must be doing
Wesley> something wierd.

I doubt you're doing something weird, I remember how much of a hassle
it was completely turning off scoring in mail groups at one point.

Try using this scorefile (named as nnml:all.SCORE, or nnml.all.SCORE
in your News directory if you're using gnus-score-find-bnews):

((local (gnus-use-scoring t)
	(gnus-nov-is-evil nil))
 (exclude-files "all.SCORE")
 (read-only))

This works for me with:
(setq gnus-use-adaptive-scoring t)
in my .gnus.
-- 
steve@miranova.com baur
Unsolicited commercial e-mail will be billed at $250/message.
What are the last two letters of "doesn't" and "can't"?
Coincidence?  I think not.

Re: Adaptive scoring in nnml groups

[Lars Magne Ingebrigtsen]

Wesley.Hardaker@sphys.unil.ch writes:

> Anyway, as an example I just checked the variable
> gnus-use-adaptive-scoring in the 'ding' nnml group (which it doesn't
> work on) and it is set to 't' as it should be.  All of my summary
> lines, however, display the score as '0' (I actually display the
> value).  Odd...

Whether a group is adaptive can be checked by looking at the
`gnus-newsgroup-adaptive' variable.  (It is initialized from
`gnus-use-adaptive-scoring'.)

I've again tried to check this, but when I set
`gnus-use-adaptive-scoring' to t, all my groups (including the nnml
group) become adaptive.

-- 
  "Yes.  The journey through the human heart 
     would have to wait until some other time."

Re: Adaptive scoring in nnml groups

[Wesley.Hardaker]

Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

> Whether a group is adaptive can be checked by looking at the
> `gnus-newsgroup-adaptive' variable.  (It is initialized from
> `gnus-use-adaptive-scoring'.)
> 
> I've again tried to check this, but when I set
> `gnus-use-adaptive-scoring' to t, all my groups (including the nnml
> group) become adaptive.

Welp, I just checked both variables in my ding group and they show up
as 't'.  I have put the following in my nnml:all.SCORE file (at the
suggestion of Steven), but that hasn't seemed to work either:

((local (gnus-use-scoring t)))

Note that this variable is also 't' according to 'C-h v'.  sigh...
Something is definately odd here.

Maybe I'll have to put in everything he suggested, but I didn't see
how the rest mattered.

Wes

PS.  This afternoon campus is changing from inn 1.4 to dnews...
Someone tell me I'm not screwed and there are no bugs with this news server...

Re: Adaptive scoring in nnml groups

[Hans de Graaff]

Wesley.Hardaker@sphys.unil.ch writes:

> Anyway, as an example I just checked the variable
> gnus-use-adaptive-scoring in the 'ding' nnml group (which it doesn't
> work on) and it is set to 't' as it should be.  All of my summary
> lines, however, display the score as '0' (I actually display the
> value).  Odd...

I've been paying a bit more attention to this problem. I don't get
ADAPT files for any of the nnml groups (and hence no scoring).
However, there are some nntp groups for which I do get an ADAPT file,
but for which scoring still doesn't work. All scores list as 0. I've
tried looking into the .newsrc.eld file to see if these groups are
different from other nntp groups, but there was nothing apparent.

Hans

Re: Adaptive scoring in nnml groups

[Lars Magne Ingebrigtsen]

Hans de Graaff <J.J.deGraaff@twi.tudelft.nl> writes:

> I've been paying a bit more attention to this problem. I don't get
> ADAPT files for any of the nnml groups (and hence no scoring).

Hm.  Could you try to `M-x edebug-defun' on `gnus-score-adaptive' in
"gnus-score.el" and then SPC through the function when you exit from
the group?  Try to see what goes wrong in some groups.

-- 
  "Yes.  The journey through the human heart 
     would have to wait until some other time."

Re: Adaptive scoring in nnml groups

[Hans de Graaff]

It seems I'm starting to understand what has been happening here.

Wesley.Hardaker@sphys.unil.ch writes:

> Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:
> 
> > Hm...  I couldn't find anything in the adaptive score code that would
> > make adaptive scoring be switched on only in nntp groups.  Are you
> > sure you're not switching it off in some hook or other?

Adaptive scoring was on for those groups.

> I don't know if you rememeber or not, but I have been having this
> exact same problem for a long time now.  I turned on adaptive
> scoring way back in .20 or so of sgnus and starting using nnml later
> on.  I wrote to you and told you about this problem, and you said
> something to the effect of 'I can't repeat it using your settings',
> so I figured I must be doing something wierd.

Probably the weird thing is that you are using something like 

    (setq gnus-auto-expirable-newsgroups "mail\\..*")

I just turned this off, and lo and behold: ADAPT files for my nnml
groups. What seems to happen is that all articles get another mark
because of this, and this mark isn't in the adaptive alist. If there
are no scores, the cache data isn't 'touched, and no ADAPT file gets
written.

In addition to this I was ticking other articles, and there is no
default adaptive alist rule for gnus-ticked-mark articles. So, again,
no scores.

Lars, I hope this pins this down a bit more. It seems to be the marks
should be run through adaptive scoring first, before auto-expriable
gets its hands on them and changes them to gnus-expirable-mark.
Otherwise I suppose I should switch to total-expire, but in that case
a warning about this on the Expiring Mail page of the info file would
be nice.

Hans

Re: Adaptive scoring in nnml groups

[Wesley.Hardaker]

Hans de Graaff <J.J.deGraaff@twi.tudelft.nl> writes:

> Probably the weird thing is that you are using something like 
> 
>     (setq gnus-auto-expirable-newsgroups "mail\\..*")

That would make sense...  I'm glad someone finally figured it out.  It
was driving me nuts!

Thanks!
Wes

Re: Adaptive scoring in nnml groups

[Lars Magne Ingebrigtsen]

Hans de Graaff <J.J.deGraaff@twi.tudelft.nl> writes:

> Lars, I hope this pins this down a bit more. It seems to be the marks
> should be run through adaptive scoring first, before auto-expriable
> gets its hands on them and changes them to gnus-expirable-mark.
> Otherwise I suppose I should switch to total-expire, but in that case
> a warning about this on the Expiring Mail page of the info file would
> be nice.

I've now added warnings about this to both that section and the
adaptive scoring section.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@ifi.uio.no * Lars Ingebrigtsen

Re: Adaptive scoring in nnml groups

[Wesley.Hardaker]

Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

> I've now added warnings about this to both that section and the
> adaptive scoring section.

Um, wouldn't it be better to find a way to fix it instead?

Wes

Re: Adaptive scoring in nnml groups

[Lars Magne Ingebrigtsen]

Wesley.Hardaker@sphys.unil.ch writes:

> Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:
> 
> > I've now added warnings about this to both that section and the
> > adaptive scoring section.
> 
> Um, wouldn't it be better to find a way to fix it instead?

I don't know how to fix it.  The adaptive thingies do their work based
on article marks, and using auto-expiry will mean that all articles
get the same mark.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@ifi.uio.no * Lars Ingebrigtsen

Re: Adaptive scoring in nnml groups

[Wesley.Hardaker]

Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

> I don't know how to fix it.  The adaptive thingies do their work based
> on article marks, and using auto-expiry will mean that all articles
> get the same mark.

Grr....  Hmm...  The person that found it (sorry...  I've forgotten
your name already) was right...  We need to run the adaptive stuff
before the expiratory stuff...  Either that or make it possible for an
article to be both read and expired.  That way you could deduct on
expired articles and increase on expired and read...

Of course, I don't have any clue how this is currently handled, so I'm
sure its probably impossible to have two marks...

Hmmm...

Hmm....

There must be a way.

There has got to be a way...

I need to go home...

Its too late to be here on a friday...

Must breath...

(sorry)

Wes

Re: Adaptive scoring in nnml groups

[Hans de Graaff]

Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

> I don't know how to fix it.  The adaptive thingies do their work
> based on article marks, and using auto-expiry will mean that all
> articles get the same mark.

But you could rewrite auto-expiry as a pass after doing adaptive
scoring. (I think right now it just changes the marks as you go
through the group.) When you would do this after running the summary
through adaptive scoring, I don't think anything would break?

Hans

Re: Adaptive scoring in nnml groups

[Lars Magne Ingebrigtsen]

Hans de Graaff <J.J.deGraaff@twi.tudelft.nl> writes:

> But you could rewrite auto-expiry as a pass after doing adaptive
> scoring.

No -- auto-expiry works by marking the articles with `E' when the user
reads an article.  If the user presses `d' instead, the `E' mark will
be gone and the article will no longer be expirable.

What you're suggesting would make it difficult for users to see
whether articles are expirable or not.

-- 
  "Yes.  The journey through the human heart 
     would have to wait until some other time."

Re: Adaptive scoring in nnml groups

[Yair Friedman]

Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

> I don't know how to fix it.  The adaptive thingies do their work based
> on article marks, and using auto-expiry will mean that all articles
> get the same mark.
> 


When annotations are implemented maybe adaptive scoring could use its
own set of marks?

-- 
Yair Friedman.                  John Bryce (Systems) 1984 Ltd.           
                                7 Hamarpeh St.                           
                                Har-Hotzvom Science Based Industry Campus
                                Jerusalem 97 774, Israel                 
<URL:mailto:YFriedma@JohnBryce.Co.Il>

Re: Adaptive scoring in nnml groups

[Lars Magne Ingebrigtsen]

Yair Friedman <yfriedma@johnbryce.co.il> writes:

> When annotations are implemented maybe adaptive scoring could use its
> own set of marks?

Uhm...  No, I don't think so.  Those would be separate issues -- you
wouldn't want to rely on being able to make annotations (which would
in all likelihood be slow) to do adaptive scoring (which has to be
fast). 

-- 
  "Yes.  The journey through the human heart 
     would have to wait until some other time."