DBIx::Class::AuditAny - Flexible change tracking framework for DBIx::Class
my $schema = My::Schema->connect(@connect);
use DBIx::Class::AuditAny;
my $Auditor = DBIx::Class::AuditAny->track(
schema => $schema,
track_all_sources => 1,
collector_class => 'Collector::AutoDBIC',
collector_params => {
sqlite_db => 'db/audit.db',
}
);
This module provides a generalized way to track changes to DBIC databases. The aim is to provide quick/turn-key options to be able to hit the ground running, while also being highly flexible and customizable with sane APIs.
DBIx::Class::AuditAny
wants to be a general framework on top of which other Change Tracking modules for DBIC can be written, while also providing fully fleshed, end-user solutions that can be dropped in and work out-of-the-box.
This module was originally written in 2012 for an internal client project, and the process of getting it released open-source as a stand-alone, general-purpose module was started in 2013. However, I got busy with other projects and wasn't able to complete a CPAN release at that time (mainly due to missing docs and minor loose ends). I finally came back to this project (May 2015) to actually get a release out to CPAN. So, even though the release date is in 2015, the majority of the code is actually several years old (and has been running perfectly in production for several client apps the whole time).
AuditAny uses a different API than typical DBIC components. Instead of loading at the schema/result class level with load_components
, AuditAny is used by attaching an "Auditor" to an existing schema object instance:
my $schema = My::Schema->connect(@connect);
my $Auditor = DBIx::Class::AuditAny->track(
schema => $schema,
track_all_sources => 1,
collector_class => 'Collector::AutoDBIC',
collector_params => {
sqlite_db => 'db/audit.db',
}
);
The rationale of this approach is that change tracking isn't necessarily something that needs to be, or should be, defined as a built-in attribute of the schema class. Additionally, because of the object-based approach, it is possible to attach multiple Auditors to a single schema object with multiple calls to DBIx::Class::AuditAny->track.
As changes occur in the tracked schema, information is collected in the form of datapoints at various stages - or contexts - before being passed to the configured Collector. A datapoint has a globally unique name and code used to calculate its value. Code is called at the stage defined by the context of the datapoint. The available contexts are:
- set
-
- base
- change
-
- source
- column
set (AKA changeset) datapoints are specific to an entire set of changes - insert/ update/delete statements grouped in a transaction. Example changeset datapoints include changeset_ts
and other broad items. base datapoints are logically the same as set but only need to be calculated once (instead of with every change set). These include things like schema
and schema_ver
.
change datapoints apply to a specific insert
, update
or delete
statement, and range from simple items such as action
(one of 'insert', 'update' or 'delete') to more exotic and complex items like column_changes_json
. source datapoints are logically the same as change, but like base datapoints, only need to be calculated once (per source). These include things like table_name
and source
(source name).
Finally, column datapoints cover information specific to an individual column, such as column_name
, old_value
and new_value
.
There are a number of built-in datapoints (currently stored in DBIx::Class::AuditAny::Util::BuiltinDatapoints which is likely to change), but custom datapoints can also be defined. The Auditor config defines a specific set of datapoints to be calculated (built-in and/or custom). If no datapoints are specified, the default list is used (currently change_ts, action, source, pri_key_value, column_name, old_value, new_value
).
The list of datapoints is specified as an ArrayRef in the config. For example:
datapoints => [qw(action_id column_name new_value)],
Custom datapoints are specified as HashRef configs with 3 parameters:
- name
-
The unique name of the datapoint. Should be all lowercase letters, numbers and underscore and must be different from all other datapoints (across all contexts).
- context
-
The context of the datapoint: base, source, set, change or column.
- method
-
CodeRef to calculate and return the value. The CodeRef is called according to the context, and a different context object is supplied for each context. Each context has its own context object type except base which is supplied the Auditor object itself. See Audit Context Objects below.
Custom datapoints are defined in the datapoint_configs
param. After defining a new datapoint config it can then be used like any other datapoint. For example:
datapoints => [qw(action_id column_name new_value client_ip)],
datapoint_configs => [
{
name => 'client_ip',
context => 'set',
method => sub {
my $contextObj = shift;
my $c = some_func(...);
return $c->req->address;
}
}
]
Datapoint names must be unique, which means all the built-in datapoint names are reserved. However, if you really want to use an existing datapoint name, or if you want a built-in datapoint to use a different name, you can rename any datapoints like so:
rename_datapoints => {
new_value => 'new',
old_value => 'old',
column_name => 'column',
},
Once the Auditor calculates the configured datapoints it passes them to the configured Collector. There are several built-in Collectors provided, but writing a custom Collector is a trivial matter. All you need to do is write a Moo-compatible class which consumes the DBIx::Class::AuditAny::Role::Collector role and implement a record_changes()
method. This method is called with a ChangeSet object supplied as the argument at the end of every database transaction which performs a write operation.
No matter how small or large the transaction, the ChangeSet object provides APIs to a nested structure to be able to access all information regarding what changed during the given transaction. (See AUDIT CONTEXT OBJECTS below).
The following built-in collector classes are already provided:
Inspired in part by the Catalyst Context object design, the internal machinery which captures and organizes the change datapoints associated with a modifying transaction is wrapped in a nested structure of 3 kinds of "context" objects:
This provides a clean and straightforward API for which Collector classes are able to identify and act on the data in any manner they want, be it recording to a database, logging to a simple file, or taking any kind of programmatic action. Collectors can really be thought of as a structure for powerful external triggers.
Note: Documentation of all the individual attrs and methods of this class (shown below) is still TBD. However, most meaningful scenarios involving interacting with these is already covered above, or is covered further down in the Examples.
Record all changes into a *separate*, auto-generated and initialized SQLite schema/db with default datapoints (Quickest/simplest usage - SYNOPSIS example):
Uses the Collector DBIx::Class::AuditAny::Collector::AutoDBIC
my $schema = My::Schema->connect(@connect);
use DBIx::Class::AuditAny;
my $Auditor = DBIx::Class::AuditAny->track(
schema => $schema,
track_all_sources => 1,
collector_class => 'Collector::AutoDBIC',
collector_params => {
sqlite_db => 'db/audit.db',
}
);
Record all changes - into specified target sources within the *same*/tracked schema - using specific datapoints:
Uses the Collector DBIx::Class::AuditAny::Collector::DBIC
DBIx::Class::AuditAny->track(
schema => $schema,
track_all_sources => 1,
collector_class => 'Collector::DBIC',
collector_params => {
target_source => 'MyChangeSet', # ChangeSet source name
change_data_rel => 'changes', # Change source, via rel within ChangeSet
column_data_rel => 'change_columns', # ColumnChange source, via rel within Change
},
datapoints => [ # predefined/built-in named datapoints:
(qw(changeset_ts changeset_elapsed)),
(qw(change_elapsed action source pri_key_value)),
(qw(column_name old_value new_value)),
],
);
Dump raw change data for specific sources (Artist and Album) to a file, ignore immutable flags in the schema/result classes, and allow more than one DBIx::Class::AuditAny Auditor to be attached to the same schema object:
Uses 'collect' sugar param to setup a bare-bones CodeRef Collector (DBIx::Class::AuditAny::Role::Collector)
my $Auditor = DBIx::Class::AuditAny->track(
schema => $schema,
track_sources => [qw(Artist Album)],
track_immutable => 1,
allow_multiple_auditors => 1,
collect => sub {
my $cntx = shift; # ChangeSet context object
require Data::Dumper;
print $fh Data::Dumper->Dump([$cntx],[qw(changeset)]);
# Do other custom stuff...
}
);
Record all updates (but *not* inserts/deletes) - into specified target sources within the same/tracked schema - using specific datapoints, including user-defined datapoints and built-in datapoints with custom names:
DBIx::Class::AuditAny->track(
schema => CoolCatalystApp->model('Schema')->schema,
track_all_sources => 1,
track_actions => [qw(update)],
collector_class => 'Collector::DBIC',
collector_params => {
target_source => 'MyChangeSet', # ChangeSet source name
change_data_rel => 'changes', # Change source, via rel within ChangeSet
column_data_rel => 'change_columns', # ColumnChange source, via rel within Change
},
datapoints => [
(qw(changeset_ts changeset_elapsed)),
(qw(change_elapsed action_id table_name pri_key_value)),
(qw(column_name old_value new_value)),
],
datapoint_configs => [
{
name => 'client_ip',
context => 'set',
method => sub {
my $c = some_func(...);
return $c->req->address;
}
},
{
name => 'user_id',
context => 'set',
method => sub {
my $c = some_func(...);
$c->user->id;
}
}
],
rename_datapoints => {
changeset_elapsed => 'total_elapsed',
change_elapsed => 'elapsed',
pri_key_value => 'row_key',
new_value => 'new',
old_value => 'old',
column_name => 'column',
},
);
Record all changes into a user-defined custom Collector class - using default datapoints:
my $Auditor = DBIx::Class::AuditAny->track(
schema => $schema,
track_all_sources => 1,
collector_class => '+MyApp::MyCollector',
collector_params => {
foo => 'blah',
anything => $val
}
);
Access/query the audit db of Collector::DBIC and Collector::AutoDBIC collectors:
my $audit_schema = $Auditor->collector->target_schema;
$audit_schema->resultset('AuditChangeSet')->search({...});
# Print the ddl that auto-generated and deployed with a Collector::AutoDBIC collector:
print $audit_schema->resultset('DeployInfo')->first->deployed_ddl;
See the unit tests (which are extensive) for more examples.
Enable tracking multi-primary-key sources (code currently disabled)
Write more tests
Write more documentation
Add more built-in datapoints
Expand the Collector API to be able to provide datapoint configs
Separate set/change/column datapoints into 'pre' and 'post' stages
Add mechanism to enable/disable tracking (localizable global?)
Switch to use Types::Standard
DBIx::Class::Journal was the first DBIC change tracking module released to CPAN. It works, but is inflexible and mandates a single mode of operation, which is not ideal in many ways.
DBIx::Class::AuditLog takes a more casual approach than DBIx::Class::Journal, which makes it easier to work with. However, it still forces a narrow and specific manner in which it stores the change history data which doesn't fit all workflows.
AuditAny was designed specifically for flexibility. By separating the Auditor - which captures the change data as it happens - from the Collector, which handles storing the data, all sorts of different styles and manners of formatting and storing the audit data can be achieved. In fact, DBIx::Class::AuditLog could be written using AuditAny, and store the data in exactly the same manner by implementing a custom collector class.
Shadow is a different animal. It is very sophisticated, and places accuracy above all else, with the idea of being able to do things such as reliably "revive" the previous state of rows, etc. The downside of this is that it is also not flexible, in that it handles the entire change life cycle within its logic. This is different from AuditAny, which is more like a packet capture lib for DBIC (like tcpdump/libpcap is a packet capture lib for networks). Unlike the others, Shadow could not be implemented using AuditAny, because the way it captures the change data is specific and fundamentally different.
Unfortunately, DBIx::Class::Shadow is unfinished and has never been released to CPAN (as of the time of this writing, in May 2015). Its current, unfinished status can be seen in GitHub:
IRC:
Join #rapidapp on irc.perl.org.
Henry Van Styn <vanstyn@cpan.org>
This software is copyright (c) 2012-2015 by IntelliTree Solutions llc.
This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.