sa-learn-cyrus(1) Train Spamassassin with spam/ham from user's imap mailboxes


sa-learn-cyrus [ options ] user-name(s)

  user-name(s)              One ore more user/mailbox name(s).
    --help                  Prints a brief help message and exits.
    --man                   Prints the manual page and exits.
    --verbose level         Be verbose if level > 0
    -v level
    --config file           Use a configuration file other than the default
    -c file                 one.
    --sa-debug              Run sa-learn in debug mode.
    --simulate              Run in simulation mode (show commands only).
    --imap-domains domains  Search mailboxes in list of domains.
    -D domains


sa-learn-cyrus feeds spam and non-spam (ham) messages to Spamassassin's database. Its main purpose is to train SA's bayes database with spam/ham messages sorted by the mailbox owners into special subfolders.

It is intended to be used on smal mail systems (e.g. home office) with a single server-wide SA configuration.

Launching sa-learn-cyrus at regular intervalls (cron job) may improve SA's hit rate considerably, provided that the users are well instructed what to move to their ham/spam folders and what not.


sa-learn-cyrus scans local mail spools as used by Cyrus IMAPd for special subfolders. These subfolders are supposed to contain mails which have been classfied as spam or ham by the mailbox owners.

Example: The users move spam mails which have not been tagged as spam by SpamAssassin (false positives) to a subfolder INBOX.Learn.Spam. Other mails, which may be classified by SA as spam in the future because of certain characteristics are copied to a subfolder INBOX.Learn.Ham.

sa-learn-cyrus feeds the content of these spam/ham folders to SA's Bayes database using the sa-learn tool which is shipped with the Spamassassin package.

Afterwards these mails are deleted (optionally) by means of ipurge which is a helper tool coming along with the Cyrus IMAPd package.


sa-learn-cyrus optionally takes a list of mailbox/user names as agruments:

  sa-learn-cyrus fred wilma fritz hjb

If not supplied all mailboxes found will be handled.


All options supplied on the comand line will override corresponding parameters given in the configuration file.

Please note that the basic parameters of sa-learn-cyrus have to be defined in a configuration file. sa-learn-cyrus cannot be controlled solely by means of command.

--config file, -c file
Use configuration file other then the default one. Always adopt the configuartion file to your needs before using sa-learn-cyrus on a live system. Otherwise you may loose data or corrupt your SA data base!
--verbose level, -v level
Specify level of verbosity. (Default = 0)
--sa-debug, -d
Run sa-learn in debug mode. This may be useful to examine problems with sa-learn.
--simulate, -s
Run sa-learn-cyrus in simulation mode. This is useful for first tests after initial configuration or if problem are encountered. In simulation mode sa-learn-cyrus doesn't execute any system commands nor does it touch any data. It just displays what it would do.
--imap-domains list-of-domains, -D list-of-domains
If your Cyrus installation uses the ``domain support'' you may use this option to tell what domains you want to be searched.


is equivalent to

  domains = 

in the configuration file.


By default sa-learn-cyrus expects its configuration file as /etc/sapmasassin/sa-learn-cyrus.conf.

One has to change this setting in the code, if another default file is wanted. Another than the default file can always be choosen with the "--config option".

A sample configuration file is shipped with sa-learn-cyrus.


The configuration file has a format as knwon from rsync or samba is very similar to the format of Windows ini files. The file consist of sequence sections. The begin of each section is designated with a section name, a word in square brackets, e.g. "[global]". The section entries consist of parameters, which are key/value pairs each on a single line. Key an value are separated by an equal sign like

  key = value

The value is a single word or a list of words each of them representing a number or a string. Words may be surrounded ba any number of spaces for better readability. Empty lines and lines with a leading hash character "#" are ingored.

Section [global]

The [global] section contains all global controll parameters.
tmp_dir = temporary-directory
sa-learn-cyrus creates some temporary files during each run. This is the directory where thes files are created.
lock_file = full-path-to-lock-file
To avoid race conditions, sa-learn-cyrus uses a simple file locking mechanism. Each new sa-learn-cyrus process looks for this file before it realy does anything. If this file exists, the process exits with a warning, assuming that another sa-learn-cyrus process is running.
verbose = level
The level of verbosity. Values range from 0 (low) to 3 (high). A reasonable level to start with is 1.
simulate = yes|no
sa-learn-cyrus should be run in simulation mode ("simulate = yes") after the first customization of the configuration to avoid loss of data or corruption of SA's database in case of wrongly configured parameters.

Section [mailbox]

Section [mailbox] contains all parameters to select the mailboxes, to specify the special subfolders, and to define the actions to apply.
include_list = list-of-mailboxes
Only spam/ham mails of these mailboxes are fed to Spamassassin's database. If this List ist empty, all mailboxes will be used. "include_list" may be used instead of the list on the command line.


  include_list = fred wilma fritz hjb
include_regexp = regular-expression
If include_list is empty, a regular expression given here is applied to all mailbox names to select mailboxes. This parameter is ignored if include_list is not empty.

Example: Include all mailboxes beginning with 'knf-'.

  include_regexp = ^knf-
exclude_list = list-of-mailboxes
A list of mailboxes wich will be excluded. If include_list is not empty, this parameter is ignored.
exclude_regexp = regular-expression
Mailbox names which match with this regular expresson are excluded from processing.

Example: Ignore all mailboxes ending with '.beie'

  exclude_regexp = \.beie$
spam_folder = folder-name
The name of the special subfolder in each mailbox which contains spam. The name should be a complete folder path relative to the root folder INBOX. The Cyrus nomenclature is applied (same as with cyradm).


  spam_folder = Learn.Spam

This is a subfolder in a folder tree like this:

    |  +--Ham
    |  +--Spam  <-- spam subfolder
ham_folder = folder-name
The name of the special subfolder in each mailbox which contains ham. (Same naming scheme as with "spam_folder", see above.)
remove_spam = yes|no
Are the spam messages in the "spam_folder" to be removed after feeding them to the SA database or not?
remove_ham = yes|no
Are the ham messages in the "ham_folder" to be removed after feeding them to the SA database or not?

Section [sa]

Spamassassin (SA) configuration items.
site_config_path = path
Path to system-wide SA preferences.


  site_config_path = /etc/spamassassin
prefs_file = file
Path of the system-wide SA configuartin file.


  prefs_file = /etc/spamassassin/
learn_cmd = path
Path to the sa-learn utility.


  learn_cmd = /usr/bin/sa-learn
user = user-id
The user id SA runs with.


  user = mail
group = group-id
The group id SA runs with.


  group = mail
debug = yes|no
Run sa-learn in debug mode or not. "debug = yes" may be useful to examine problems.

Section [imap]

The section [imap] contains the necessary configuration parameter to locate an manage the (Cyrus) IMAPd spool files.
base_dir = dir
The root of the base directory of the IMAP spool (below that the mailboxes are located).
initial_letter = yes|no
If base_dir is divided in subdirectories named with the initial letters of mailbox names set "initial_letter = yes" (default), otherwise choose no.

Examples for joe's mailbox:

  <base_dir>/j/user/joe/ : initial_letter = yes
  <base_dir>/user/joe/ : initial_letter = no
domains = list-of-domains
If your Cyrus spool uses domain hierarchy supply a list of domains. If domain support is not used leave this entry empty. The "initial_letter" option (see above) is applied to domains, too.

Example for mailboxes [email protected] and [email protected] :

The mail files within the Cyrus spool are located at


List the domains as

  domains =
unixhierarchysep = yes|no
Choose "unixhierarchysep = yes" if Cyrus is configured to accept usernames like 'hans.mueller.somedomain.tld'. Otherwise set "unixhierarchysep = no".
purge_cmd = path-to-command
The path to the Cyrus ipurge utility for purging mail messages.


  purge_cmd = /usr/sbin/ipurge
user = user
The user Cyrus-IMAPd runs as.


  user = cyrus




sa-learn (part of the SpamAssassin package), ipurge (part of Cyrus IMAPd)


Hans-Juergen Beie <[email protected]>


Copyright 2004-2008 by Hans-Juergen Beie.

This program is free software; you can redistribute it and/or modify it under the terms of the Artistic License 2.0 <> or the GNU General Public License as published by the Free Software Foundation; either version 2 of the license <>, or (at your option) any later version.


This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.


Thanks to Robert Carnecky and Jan Hauke Rahm for testing and suggestions for the implementation of the domain support.