# <@LICENSE> # Licensed to the Apache Software Foundation (ASF) under one or more # contributor license agreements. See the NOTICE file distributed with # this work for additional information regarding copyright ownership. # The ASF licenses this file to you under the Apache License, Version 2.0 # (the "License"); you may not use this file except in compliance with # the License. You may obtain a copy of the License at: # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # =head1 NAME Mail::SpamAssassin::Plugin - SpamAssassin plugin base class =head1 SYNOPSIS =head2 SpamAssassin configuration: loadplugin MyPlugin /path/to/myplugin.pm =head2 Perl code: package MyPlugin; use Mail::SpamAssassin::Plugin; our @ISA = qw(Mail::SpamAssassin::Plugin); sub new { my ($class, $mailsa) = @_; # the usual perlobj boilerplate to create a subclass object $class = ref($class) || $class; my $self = $class->SUPER::new($mailsa); bless ($self, $class); # then register an eval rule, if desired... $self->register_eval_rule ("check_for_foo"); # and return the new plugin object return $self; } ...methods... 1; =head1 DESCRIPTION This is the base class for SpamAssassin plugins; all plugins must be objects that implement this class. This class provides no-op stub methods for all the callbacks that a plugin can receive. It is expected that your plugin will override one or more of these stubs to perform its actions. SpamAssassin implements a plugin chain; each callback event is passed to each of the registered plugin objects in turn. Any plugin can call C<$self-Einhibit_further_callbacks()> to block delivery of that event to later plugins in the chain. This is useful if the plugin has handled the event, and there will be no need for later plugins to handle it as well. If you're looking to write a simple eval rule, skip straight to C, below. =head1 INTERFACE In all the plugin APIs below, C refers to a reference to a hash containing name-value pairs. This is used to ensure future-compatibility, in that we can add new options in future without affecting objects built to an earlier version of the API. For example, here would be how to print out the C item in a C method: sub parse_config { my ($self, $opts) = @_; print "MyPlugin: parse_config got ".$opts->{line}."\n"; } =head1 METHODS The following methods can be overridden by subclasses to handle events that SpamAssassin will call back to: =over 4 =cut package Mail::SpamAssassin::Plugin; use Mail::SpamAssassin; use Mail::SpamAssassin::Logger; use strict; use warnings; use bytes; use vars qw{ @ISA $VERSION }; @ISA = qw(); $VERSION = 'bogus'; ########################################################################### =item $plugin = MyPluginClass->new ($mailsaobject) Constructor. Plugins that need to register themselves will need to define their own; the default super-class constructor will work fine for plugins that just override a method. Note that subclasses must provide the C<$mailsaobject> to the superclass constructor, like so: my $self = $class->SUPER::new($mailsaobject); Lifecycle note: plugins that will need to store per-scan state should not store that on the Plugin object; see C below. It is also likewise recommended that configuration settings be stored on the Conf object; see C. =cut sub new { my $class = shift; my $mailsaobject = shift; $class = ref($class) || $class; if (!defined $mailsaobject) { die "plugin: usage: Mail::SpamAssassin::Plugin::new(class,mailsaobject)"; } my $self = { main => $mailsaobject, _inhibit_further_callbacks => 0 }; bless ($self, $class); $self; } # --------------------------------------------------------------------------- # now list the supported methods we will call into. NOTE: we don't have # to implement them here, since the plugin can use "can()" to introspect # the object and determine if it's capable of calling the method anyway. # Nifty! =item $plugin->parse_config ( { options ... } ) Parse a configuration line that hasn't already been handled. C is a reference to a hash containing these options: =over 4 =item line The line of configuration text to parse. This has leading and trailing whitespace, and comments, removed. =item key The configuration key; ie. the first "word" on the line. =item value The configuration value; everything after the first "word" and any whitespace after that. =item conf The C object on which the configuration data should be stored. =item user_config A boolean: C<1> if reading a user's configuration, C<0> if reading the system-wide configuration files. =back If the configuration line was a setting that is handled by this plugin, the method implementation should call C<$self-Einhibit_further_callbacks()>. If the setting is not handled by this plugin, the method should return C<0> so that a later plugin may handle it, or so that SpamAssassin can output a warning message to the user if no plugin understands it. Lifecycle note: it is suggested that configuration be stored on the C object in use, instead of the plugin object itself. That can be found as C<$plugin-E{main}-E{conf}>. This allows per-user and system-wide configuration to be dealt with correctly, with per-user overriding system-wide. =item $plugin->finish_parsing_end ( { options ... } ) Signals that the configuration parsing has just finished, and SpamAssassin is nearly ready to check messages. C is a reference to a hash containing these options: =over 4 =item conf The C object on which the configuration data should be stored. =back Note: there are no guarantees that the internal data structures of SpamAssassin will not change from release to release. In particular to this plugin hook, if you modify the rules data structures in a third-party plugin, all bets are off until such time that an API is present for modifying that configuration data. =item $plugin->signal_user_changed ( { options ... } ) Signals that the current user has changed for a new one. =over 4 =item username The new user's username. =item user_dir The new user's home directory. (equivalent to C<~>.) =item userstate_dir The new user's storage directory. (equivalent to C<~/.spamassassin>.) =back =item $plugin->services_authorized_for_username ( { options ... } ) Validates that a given username is authorized to use certain services. In order to authorize a user, the plugin should first check that it can handle any of the services passed into the method and then set the value for each allowed service to true (or any non-negative value). The current supported services are: bayessql =over 4 =item username A username =item services Reference to a hash containing the services you want to check. { 'bayessql' => 0 } =item conf The C object on which the configuration data should be stored. =back =item $plugin->compile_now_start ( { options ... } ) This is called at the beginning of Mail::SpamAssassin::compile_now() so plugins can do any necessary initialization for multi-process SpamAssassin (such as spamd or mass-check -j). =over 4 =item use_user_prefs The value of $use_user_prefs option in compile_now(). =item keep_userstate The value of $keep_userstate option in compile_now(). =back =item $plugin->compile_now_finish ( { options ... } ) This is called at the end of Mail::SpamAssassin::compile_now() so plugins can do any necessary initialization for multi-process SpamAssassin (such as spamd or mass-check -j). =over 4 =item use_user_prefs The value of $use_user_prefs option in compile_now(). =item keep_userstate The value of $keep_userstate option in compile_now(). =back =item $plugin->check_start ( { options ... } ) Signals that a message check operation is starting. =over 4 =item permsgstatus The C context object for this scan. Lifecycle note: it is recommended that rules that need to track test state on a per-scan basis should store that state on this object, not on the plugin object itself, since the plugin object will be shared between all active scanners. The message being scanned is accessible through the C<$permsgstatus-Eget_message()> API; there are a number of other public APIs on that object, too. See C perldoc. =back =item $plugin->check_main ( { options ... } ) Signals that a message should be checked. Note that implementations of this hook should return C<1>. =over 4 =item permsgstatus The C context object for this scan. =back =item $plugin->check_tick ( { options ... } ) Called periodically during a message check operation. A callback set for this method is a good place to run through an event loop dealing with network events triggered in a C method, for example. =over 4 =item permsgstatus The C context object for this scan. =back =item $plugin->check_post_dnsbl ( { options ... } ) Called after the DNSBL results have been harvested. This is a good place to harvest your own asynchronously-started network lookups. =over 4 =item permsgstatus The C context object for this scan. =back =item $plugin->check_post_learn ( { options ... } ) Called after auto-learning may (or may not) have taken place. If you wish to perform additional learning, whether or not auto-learning happens, this is the place to do it. =over 4 =item permsgstatus The C context object for this scan. =back =item $plugin->check_end ( { options ... } ) Signals that a message check operation has just finished, and the results are about to be returned to the caller. =over 4 =item permsgstatus The C context object for this scan. The current score, names of rules that hit, etc. can be retrieved using the public APIs on this object. =back =item $plugin->finish_tests ( { options ... } ) Called via SpamAssassin::finish and should clear up any tests that a plugin has added to the namespace. In certain circumstances, plugins may find it useful to compile perl functions from the ruleset, on the fly. It is important to remove these once the C object is deleted, however, and this API allows this. Each plugin is responsible for its own generated perl functions. =over 4 =item conf The C object on which the configuration data should be stored. =item $plugin->extract_metadata ( { options ... } ) Signals that a message is being mined for metadata. Some plugins may wish to add their own metadata as well. =over 4 =item msg The C object for this message. =back =item $plugin->parsed_metadata ( { options ... } ) Signals that a message's metadata has been parsed, and can now be accessed by the plugin. =over 4 =item permsgstatus The C context object for this scan. =back =item $plugin->start_rules ( { options ... } ) Called before testing a set of rules of a given type and priority. =over 4 =item permsgstatus The C context object for this scan. =item ruletype The type of the rules about to be performed. =item priority The priority level of the rules about to be performed. =back =item $plugin->hit_rule ( { options ... } ) Called when a rule fires. =over 4 =item permsgstatus The C context object for this scan. =item ruletype The type of the rule that fired. =item rulename The name of the rule that fired. =item score The rule's score in the active scoreset. =back =item $plugin->ran_rule ( { options ... } ) Called after a rule has been tested, whether or not it fired. When the rule fires, the hit_rule callback is always called before this. =over 4 =item permsgstatus The C context object for this scan. =item ruletype The type of the rule that was tested. =item rulename The name of the rule that was tested. =back =item $plugin->autolearn_discriminator ( { options ... } ) Control whether a just-scanned message should be learned as either spam or ham. This method should return one of C<1> to learn the message as spam, C<0> to learn as ham, or C to not learn from the message at all. =over 4 =item permsgstatus The C context object for this scan. =back =item $plugin->autolearn ( { options ... } ) Signals that a message is about to be auto-learned as either ham or spam. =over 4 =item permsgstatus The C context object for this scan. =item isspam C<1> if the message is spam, C<0> if ham. =back =item $plugin->per_msg_finish ( { options ... } ) Signals that a C object is being destroyed, and any per-scan context held on that object by this plugin should be destroyed as well. Normally, any member variables on the C object will be cleaned up automatically -- but if your plugin has made a circular reference on that object, this is the place to break them so that garbage collection can operate correctly. =over 4 =item permsgstatus The C context object for this scan. =back =item $plugin->have_shortcircuited ( { options ... } ) Has the current scan operation 'short-circuited'? In other words, can further scanning be skipped, since the message is already definitively classified as either spam or ham? Plugins should return C<0> to indicate that scanning should continue, or C<1> to indicate that short-circuiting has taken effect. =over 4 =item permsgstatus The C context object for this scan. =back =item $plugin->bayes_learn ( { options ... } ) Called at the end of a bayes learn operation. This phase is the best place to map the raw (original) token value to the SHA1 hashed value. =over 4 =item toksref Reference to hash returned by call to tokenize. The hash takes the format of: { 'SHA1 Hash Value' => 'raw (original) value' } NOTE: This data structure has changed since it was originally introduced in version 3.0.0. The values are no longer perl anonymous hashes, they are a single string containing the raw token value. You can test for backwards compatability by checking to see if the value for a key is a reference to a perl HASH, for instance: if (ref($toksref->{$sometokenkey}) eq 'HASH') {... If it is, then you are using the old interface, otherwise you are using the current interface. =item isspam Boolean value stating what flavor of message the tokens represent, if true then message was specified as spam, false is nonspam. Note, when function is scan then isspam value is not valid. =item msgid Generated message id of the message just learned. =item msgatime Received date of the current message or current time if received date could not be determined. In addition, if the receive date is more than 24 hrs into the future it will be reset to current datetime. =back =item $plugin->bayes_forget ( { options ... } ) Called at the end of a bayes forget operation. =over 4 =item toksref Reference to hash returned by call to tokenize. See bayes_learn documentation for additional information on the format. =item isspam Boolean value stating what flavor of message the tokens represent, if true then message was specified as spam, false is nonspam. Note, when function is scan then isspam value is not valid. =item msgid Generated message id of the message just forgotten. =back =item $plugin->bayes_scan ( { options ... } ) Called at the end of a bayes scan operation. NOTE: Will not be called in case of error or if the message is otherwise skipped. =over 4 =item toksref Reference to hash returned by call to tokenize. See bayes_learn documentation for additional information on the format. =item probsref Reference to hash of calculated probabilities for tokens found in the database. { 'SHA1 Hash Value' => { 'prob' => 'calculated probability', 'spam_count' => 'Total number of spam msgs w/ token', 'ham_count' => 'Total number of ham msgs w/ token', 'atime' => 'Atime value for token in database' } } =item score Score calculated for this particular message. =item msgatime Calculated atime of the message just learned, note it may have been adjusted if it was determined to be too far into the future. =item significant_tokens Array ref of the tokens found to be significant in determining the score for this message. =back =item $plugin->plugin_report ( { options ... } ) Called if the message is to be reported as spam. If the reporting system is available, the variable C<$options-{report}->report_available}> should be set to C<1>; if the reporting system successfully reported the message, the variable C<$options-{report}->report_return}> should be set to C<1>. =over 4 =item report Reference to the Reporter object (C<$options-{report}> in the paragraph above.) =item text Reference to a markup removed copy of the message in scalar string format. =item msg Reference to the original message object. =back =item $plugin->plugin_revoke ( { options ... } ) Called if the message is to be reported as ham (revokes a spam report). If the reporting system is available, the variable C<$options-{revoke}->revoke_available}> should be set to C<1>; if the reporting system successfully revoked the message, the variable C<$options-{revoke}->revoke_return}> should be set to C<1>. =over 4 =item revoke Reference to the Reporter object (C<$options-{revoke}> in the paragraph above.) =item text Reference to a markup removed copy of the message in scalar string format. =item msg Reference to the original message object. =back =item $plugin->whitelist_address( { options ... } ) Called when a request is made to add an address to a persistent address list. =over 4 =item address Address you wish to add. =back =item $plugin->blacklist_address( { options ... } ) Called when a request is made to add an address to a persistent address list. =over 4 =item address Address you wish to add. =back =item $plugin->remove_address( { options ... } ) Called when a request is made to remove an address to a persistent address list. =over 4 =item address Address you wish to remove. =back =item $plugin->spamd_child_init () Called when a new child starts up under spamd. =item $plugin->log_scan_result ( { options ... } ) Called when spamd has completed scanning a message. Currently, only spamd calls this API. =over 4 =item result The C<'result: ...'> line for this scan. Format is as described at B. =back =item $plugin->spamd_child_post_connection_close () Called when child returns from handling a connection. If there was an accept failure, the child will die and this code will not be called. =item $plugin->finish () Called when the C object is destroyed. =back =cut sub finish { my ($self) = @_; %{$self} = (); } =head1 HELPER APIS These methods provide an API for plugins to register themselves to receive specific events, or control the callback chain behaviour. =over 4 =item $plugin->register_eval_rule ($nameofevalsub) Plugins that implement an eval test will need to call this, so that SpamAssassin calls into the object when that eval test is encountered. See the B section for full details. =cut sub register_eval_rule { my ($self, $nameofsub) = @_; $self->{main}->{conf}->register_eval_rule ($self, $nameofsub); } =item $plugin->register_generated_rule_method ($nameofsub) In certain circumstances, plugins may find it useful to compile perl functions from the ruleset, on the fly. It is important to remove these once the C object is deleted, however, and this API allows this. Once the method C<$nameofsub> has been generated, call this API with the name of the method (including full package scope). This indicates that it's a temporary piece of generated code, built from the SpamAssassin ruleset, and when C is called, the method will be destroyed. This API was added in SpamAssassin 3.2.0. =cut sub register_generated_rule_method { my ($self, $nameofsub) = @_; push @Mail::SpamAssassin::PerMsgStatus::TEMPORARY_METHODS, $nameofsub; } =item $plugin->register_method_priority($methodname, $priority) Indicate that the method named C<$methodname> on the current object has a callback priority of C<$priority>. This is used by the plugin handler to determine the relative order of callbacks; plugins with lower-numbered priorities are called before plugins with higher-numbered priorities. Each method can have a different priority value. The default value is C<0>. The ordering of callbacks to methods with equal priority is undefined. Typically, you only need to worry about this if you need to ensure your plugin's method is called before another plugin's implementation of that method. It should be called from your plugin's constructor. This API was added in SpamAssassin 3.2.0. =cut sub register_method_priority { my ($self, $methname, $pri) = @_; $self->{method_priority}->{$methname} = $pri; } =item $plugin->inhibit_further_callbacks() Tells the plugin handler to inhibit calling into other plugins in the plugin chain for the current callback. Frequently used when parsing configuration settings using C. =cut sub inhibit_further_callbacks { my ($self) = @_; $self->{_inhibit_further_callbacks} = 1; } =item dbg($message) Output a debugging message C<$message>, if the SpamAssassin object is running with debugging turned on. I This function is not available in the package namespace of general plugins and can't be called via $self->dbg(). If a plugin wishes to output debug information, it should call C. =item info($message) Output an informational message C<$message>, if the SpamAssassin object is running with informational messages turned on. I This function is not available in the package namespace of general plugins and can't be called via $self->info(). If a plugin wishes to output debug information, it should call C. =cut 1; =back =head1 REGISTERING EVAL RULES Plugins that implement an eval test must register the methods that can be called from rules in the configuration files, in the plugin class' constructor. For example, $plugin->register_eval_rule ('check_for_foo') will cause C<$plugin-Echeck_for_foo()> to be called for this SpamAssassin rule: header FOO_RULE eval:check_for_foo() Note that eval rules are passed the following arguments: =over 4 =item - The plugin object itself =item - The C object calling the rule =item - standard arguments for the rule type in use =item - any and all arguments as specified in the configuration file =back In other words, the eval test method should look something like this: sub check_for_foo { my ($self, $permsgstatus, ...arguments...) = @_; ...code returning 0 or 1 } Note that the headers can be accessed using the C method on the C object, and the body by C and other similar methods. Similarly, the C object holding the current configuration may be accessed through C<$permsgstatus-E{main}-E{conf}>. The eval rule should return C<1> for a hit, or C<0> if the rule is not hit. State for a single message being scanned should be stored on the C<$checker> object, not on the C<$self> object, since C<$self> persists between scan operations. See the 'lifecycle note' on the C method above. =head1 STANDARD ARGUMENTS FOR RULE TYPES Plugins will be called with the same arguments as a standard EvalTest. Different rule types receive different information by default: =over 4 =item - header tests: no extra arguments =item - body tests: fully rendered message as array reference =item - rawbody tests: fully decoded message as array reference =item - full tests: pristine message as scalar reference =back The configuration file arguments will be passed in after the standard arguments. =head1 BACKWARDS COMPATIBILITY Note that if you write a plugin and need to determine if a particular helper method is supported on C, you can do this: if ($self->can("name_of_method")) { eval { $self->name_of_method(); # etc. } } else { # take fallback action } The same applies for the public APIs on objects of other types, such as C. =head1 SEE ALSO C C http://wiki.apache.org/spamassassin/PluginWritingTips http://issues.apache.org/SpamAssassin/show_bug.cgi?id=2163 =cut