Re: Continuing the 'Not Training' trend --> bug in mailfilter.crm(?)

:-(( and I wouldn't be me if I didn't add a mistake in a patch file:

>  @@ -728,7 +725,8 @@
>                     (:classify_status:)
>             match [:classify_status:] <nomultiline> \
>                     /^#0.* pR: ([-. 0-9]+)/ (:: :pr:)
>  -           eval /:@: :*:pr: < :*:thick_threshold: : /
>  +            # eval /:@: :*:pr: < (0 - :*:thick_threshold:) : /
>  +            eval /:@: :*:pr: < :*:spam_threshold: : /
>             #
>             #     write out the pre-mutilation text, with newlines
>             #

should read (note flip of < to > comparison)

>  @@ -728,7 +725,8 @@
>                     (:classify_status:)
>             match [:classify_status:] <nomultiline> \
>                     /^#0.* pR: ([-. 0-9]+)/ (:: :pr:)
>  -           eval /:@: :*:pr: < :*:thick_threshold: : /
>  +            # eval /:@: :*:pr: > (0 - :*:thick_threshold:) : /
>  +            eval /:@: :*:pr: > :*:spam_threshold: : /
>             #
>             #     write out the pre-mutilation text, with newlines
>             #

which also means my analysis is slightly off: false positives are not
learned in mailfilter.crm because of this:

pR < thick_threshold --> pR < 5

evaluates to FALSE, thus skipping the subsequent learn() operation.
Any other spam should be trained, as the comparison will evaluate to TRUE.

Nevertheless, it's still buggy. But so am I, tonight. :-(


Note that mailreaver.crm does not suffer from _that_ issue _at_ _all_;
the _only_ thing in there (= mail_reaver_), that _may_ be subject to
dispute is whether the initial classify precheck results should be
checked against thick_threshold or spam/good_threshold values. I'd say
the latter, because that increases the 'certainty' for a user that
anything unsure will be pushed out of the unsure zone through
training, no matter the threshold settings. But that's seat of the
pants logic...

Sorry for the screwup. Luckily the brain got it before I shut down
completely. Phew!


-- 
Met vriendelijke groeten / Best regards,

Ger Hobbelt

--------------------------------------------------
web: http://www.hobbelt.com/
 http://www.hebbut.net/
mail: [email protected]
mobile: +31-6-11 120 978
--------------------------------------------------

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone