You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Wade Rossmann <ro...@realestatewebmasters.com> on 2014/02/20 21:07:24 UTC

SQL preferences when invoked via perl

I'm using SpamAssassin via MimeDefang and trying to load user 
preferences in from a mySQL database. I've tried adding the 
user_scores_sql_* config items in the config file, but these seem to be 
ignored by everything except spamd -q. I've more or less given up on 
making that work [but if someone can clue me in that would be great] so 
I've written my own bit of code into mimedefang-filter to pull the 
preferences out of the database and then appending them with 
post_config_text.

However, it appears that none of these preferences are being applied 
other than required_hits. For example, the preferences pulled for a test 
user are:

    required_hits 7
    report_safe 0
    rewrite_header Subject [SPAM]
    score BANG_OPRAH 3.0
    required_hits 8
    required_hits 9

Where required_hits is set to 5 in the config file, 7 is in the global 
settings in DB, 8 in domain in DB, and 9 in the user's settings in DB. A 
test message <https://gist.github.com/Sammitch/9084301> I've devised 
triggers the message to be tagged, and the report shows a required score 
of 9, but the subject is not modified, and the BANG_OPRAH rule is not 
listed as matching.

Can someone enlighten me on the proper way to specify per-user settings 
pulled from a database when SpamAssassin is invoked via perl?

-- 
Wade Rossmann
Systems Administrator
rossmann.wade@realestatewebmasters.com


Re: SQL preferences when invoked via perl

Posted by "Kevin A. McGrail" <KM...@PCCC.com>.
On 2/20/2014 3:07 PM, Wade Rossmann wrote:
> I'm using SpamAssassin via MimeDefang and trying to load user 
> preferences in from a mySQL database. I've tried adding the 
> user_scores_sql_* config items in the config file, but these seem to 
> be ignored by everything except spamd -q. I've more or less given up 
> on making that work [but if someone can clue me in that would be 
> great] so I've written my own bit of code into mimedefang-filter to 
> pull the preferences out of the database and then appending them with 
> post_config_text.
>
> However, it appears that none of these preferences are being applied 
> other than required_hits. For example, the preferences pulled for a 
> test user are:
>
>     required_hits 7
>     report_safe 0
>     rewrite_header Subject [SPAM]
>     score BANG_OPRAH 3.0
>     required_hits 8
>     required_hits 9
>
> Where required_hits is set to 5 in the config file, 7 is in the global 
> settings in DB, 8 in domain in DB, and 9 in the user's settings in DB. 
> A test message <https://gist.github.com/Sammitch/9084301> I've devised 
> triggers the message to be tagged, and the report shows a required 
> score of 9, but the subject is not modified, and the BANG_OPRAH rule 
> is not listed as matching.
>
> Can someone enlighten me on the proper way to specify per-user 
> settings pulled from a database when SpamAssassin is invoked via perl?

Likely a better question for the MD list but here's some answers:

A - The API is different than using Spamc or SpamAssassin.

For example, required_hits will be used.  And you can set a config file 
with BANG_OPRAH 3.0.

But these settings ONLY affect SA's implementation of the API:

report_safe 0
rewrite_header Subject [SPAM]

Think of the API as returning a 0/1 for yes or no this is spam.

Then in MD, you've got to implement report_safe and rewrite_header for 
the subject if that's what you want.  Many people on MD list can help 
with that as I think it's been documented over the years.



B - Two ways to implement SA with MD:

One: I spawn out and call spamc from MD.  I already run a bunch of spamd 
servers so this makes most sense to me: 
http://www.pccc.com/downloads/MIMEDefang/contrib/spamc/SpamassassinSpamcSpamd%20-%20Kwiki.pdf 


Two:

load_scoreonly_sql() and signal_user_changed() are likely what you need 
such as these pseudo cut and paste of a live filter:



my $sa_object;


if ($Features{"SpamAssassin"}) {
     #$SALocalTestsOnly = 1;  # I do NOT want network tests
     $SALocalTestsOnly = 0;  # I DO want network tests

     #Switched to $sa_object as returned by SA per DFS
     $sa_object = spam_assassin_init() if defined(spam_assassin_init());

     # If you want to use auto-whitelisting:
     if (defined($sa_object)) {
        use Mail::SpamAssassin::SQLBasedAddrList;
        my $awl = Mail::SpamAssassin::SQLBasedAddrList->new();
        $sa_object->set_persistent_address_list_factory($awl) if 
defined($awl);
     }

}




Then in filter_end you need logic to figure out your username:

  #IF WE HAVE A USER, A SPAMASSASSIN OBJECT, AND THEY AREN'T AN 
AUTHORIZED USER, SET THE USER PREFERENCES
     if ($username ne '' && $Features{"SpamAssassin"} && 
$authorized_sender < 1) {
       $sa_object->load_scoreonly_sql($username) || md_syslog('warning', 
"load_scoreonly_sql failed for $recip");
       $sa_object->signal_user_changed({ username => $username}, 
user_dir => undef);
       md_syslog('warning', "Enabling SPAM Filter from MD for $username 
- $recip_list");
     }



then call  ($hits, $req, $names, $report) = spam_assassin_check();

regards,
KAM