[Spambayes] X-Spambayes-Exception in classifier.py

[LUK ShunTim]

Meyer, Tony wrote:

[snipped]

> 
> 
>>BTW, I found that (accidentally) training the same message 
>>twice will increase the number of trained mail message count 
>>by 2 so this number in  fact may not reflect accurately the 
>>amount of real training.
> 
> 
> This is still with sb_mboxtrain, I presume?  I don't use mboxtrain

No, this time I did it using the web interface.

> myself, but I thought that it added a header to messages to ensure that
> it doesn't train messages twice (unless this is overridden).  Did that
> not happen?
> 
> By the "number of trained mail message count", do you mean the total
> counts in the hammie.db database?  The one that, when it is wrong,
> causes the exception?  It's the "total emails trained" count on the main

Yes, that's what I meant.

> page of the web interface.  If so, then this should happen - if you
> train the message twice, then all the tokens for the message will be
> incremented twice and the total count should be incremented twice.
> Perhaps I'm missing what you meant?

If so then how can one tell how many *distinct* massage are actually
trained? It may be a little confusing if people try to use this
information to follow the recommendation of "number of ham and spam of
equal order".

> 
> 
>>Finally, I just found these error messages appearing
>><quote>
>>error: uncaptured python exception, closing channel 
>><spambayes.Dibbler._HTTPHandler connected at 0xb586ab8c> 
>>(socket.error:(32, 'Broken pipe') 
>>[/usr/lib/python2.3/asynchat.py|initiate_send|218]
>>[/usr/lib/python2.3/asyncore.py|send|337])
>></quote>
>>Quite a lot of them. Not harmful, I suppose?
> 
> 
> I don't recall seeing this before, offhand.  At first glance, it looks
> like problems with the connection to/from the local web interface
> breaking.  If everything still works, then it should be ok.  If you're

Yes, at least up to now; nothing breaks.

> concerned, open a bug report <http://sf.net/projects/spambayes> and
> someone will look at it when they can (could be a while at the moment).
> 
> =Tony.Meyer
> 

Thanks very much,
ST
--