bogofilter on opensolaris with sqlite

[timejumper at hush.ai]

On Sat, 23 Jan 2010 14:04:25 -0800 David Relson 
<relson at osagesoftware.com> wrote:
>On Sat, 23 Jan 2010 11:51:19 -0800
>timejumper at hush.ai wrote:
>
>> 
>> I am using bogofilter 1.2.1 with sqlite 3.6.21 on opensolaris 
>5.11 
>> snv_111b
>> 
>> When I run bogoutil -r sqllite.db it returns -NaN
>> 
>> Any ideas?  Is it a sqlite problem or bogofilter problem?
>> Thanks for any tips.
>> 
>> I tried compiling sqlite with -DSQLITE_HAVE_ISNAN
>
>No obvious answer comes to mind.  
>
>Are you a new user or have you been using bogofilter for a while?  

I've been using bogofilter for a while. I have never used it with 
sqlite.
I haven't really had to do much at all with it for a long time. And 
so I must have done something wrong when migrating from my previous 
.db

One odd thing that threw me off was that when I first started using 
sqlite I ran bogoutil -r .db and it printed a number. Then 
subsequent attemps always gave me a -NaN.  

>I'd
>guess the latter since the "-r" option isn't useful until there's 
>a
>database with a decent number of spam and ham messages in it. 
>
>How big is your database?  More specifically how many tokens and
>messages does it have?  The following commands will provide the 
>numbers:
>
>  bogoutil -p {path_to_your_wordlist} .MSG_COUNT

I ran this and got spam=249 good=0.
I've re-trained with ham/spam and now -r returns a number again.

now spam=81 good=1073

>  bogoutil -d {path_to_your_wordlist} | wc -l

~54,000

Thanks for the tips, I think I'm underway again.

>
>Bogoutil has a number of debugging flags that will print tons of
>information.  We'll drill down into that area if we can't find an
>answer using a less verbose method.
>
>Regards,
>
>David