new(\%config)

    # hash reference of named params
    my $object = Your::Badger::Module->new({
        arg1 => 'value1',
        arg2 => 'value2',
        ...etc...
    });

    # list of named params
    my $object = Your::Badger::Module->new(
        arg1 => 'value1',
        arg2 => 'value2',
        ...etc...
    );

The constructor creates a new object by blessing a hash reference and then calling the "init()" method. A reference to the hash array of named parameters is passed as a single argument. In most cases you should be able to re-use the existing new() method and define your own init() method to initialise the object.

The "new()" method returns whatever the init() method returns. This will normally be the $self object reference, but your own init() methods are free to return whatever they like. However, it must be a true value of some kind or the new() method will throw an error indicating that the init() method failed.

init(\%config)

The "init()" method is passed a reference to a hash array of named configuration parameters. The method may perform any configuration or initialisation processes and should generally return the $self reference to indicate success.

    sub init {
        my ($self, $config) = @_;

        # set the 'answer' parameter or default to 42
        $self->{ answer } = $config->{ answer } || 42;

        return $self;
    }

The "init()" method can return any true value which will then be sent back as the return value from new(). In most cases you'll want to return the $self object reference, but the possibility exists of returning other values instead (e.g. to implement singletons, prototypes, or some other clever object trickery).

If something goes wrong in the "init()" method then you should call the error() method (or error_msg()) to throw an error.

    sub init {
        my ($self, $config) = @_;

        # set the 'answer' parameter or report error
        $self->{ answer } = $config->{ answer }
            || return $self->error('no answer supplied');

        return $self;
    }

The only function of the default init() method in "Badger::Base" is to save a reference to the $config hash array in "$self->{ config }". If you use the default init() method, or add an equivalent line to your own init() method, then you can defer inspection of the configuration parameters until later. For example, you might have a method which does something like this:

    our $DATE_FORMAT = '%Y-%d-%b';

    sub date_format {
        my $self = shift;
        return @_
            ? ($self->{ date_format } = shift)      # set from argument
            :  $self->{ date_format }               # get from self...
           ||= $self->{ config }->{ date_format }   #  ...or config...
           ||  $DATE_FORMAT;                        #  ...or pkg var
    }

This allows you to use "$self->{date_format}" as a working copy of the value while keeping the original configuration value (if any) intact in "$self->{config}->{date_format}". The above method will set the value if you pass an argument and return the current value if you don't. If no current value is defined then it defaults to the value in the config hash or the $DATE_FORMAT package variable. Now any other methods that require access to a date format need only call to the "date_format()" method to have it Do The Right Thing.

The benefit here is that you don't have to waste time (and memory) in the init() method copying the "date_format" parameter from $config into $self. That doesn't mean that it's always the right thing to do, but it can be useful for configuration options that are infrequently used.

The on_error() and on_warn() methods follow this protocol. They look for an "ON_ERROR" or "ON_WARN" item in $self or an "on_error" or "on_warn" parameter in "$self->{config}". If you want to define "on_error" and/or "on_warn" handlers as configuration parameters then you'll need to either copy the $config reference into "$self->{config}" or copy the individual items into "$self->{ON_ERROR}" and/or "$self->{ON_WARN}", respectively.

    # either copy the config...
    sub init {
        my ($self, $config) = @_;
        $self->{ config } = $config;
        # ...more code...
        return $self;
    }

    # ...or the individual items
    sub init {
        my ($self, $config) = @_;

        # no need to check if either of these are defined because the
        # on_warn() and on_error() methods will Do The Right Thing.
        $self->{ ON_WARN  } = $config->{ on_warn  };
        $self->{ ON_ERROR } = $config->{ on_error };

        # ...more code...
        return $self;
    }

With either of the above in place, you can then define "on_warn" and "on_error" handlers and expect them to work when the error() and warn() methods are called.

    my $object = Your::Badger::Module->new(
        on_warn  => \&my_warn,
        on_error => \&my_error,
    );

    $object->warn("Rebel Alliance engaging");    # calls my_warn()
    $object->error("Exhaust port unshielded!");  # calls my_error()

warn($message)

    $object->warn("Careful with that axe, Eugene!");

on_warn($handler, $another_handler, ...)

For example, if you wanted to forward warning messages to your favourite logging tool, you might do something like this:

    my $log = My::Fave::Log::Tool->new(%log_config);

    $object->on_warn(
        sub {
            my $message = shift;
            $log->warning("$message);
            return $message;
        }
    );

The value returned from the callback is forwarded to the next handler (if there is one). If a callback returns a false value or an empty list then the remaining handlers will not be called.

The default behaviour of the "on_warn()" method is to replace any existing warning handlers with the new one(s) specified. You can prefix the handler(s) with 'before' or 'after' to add them to the existing list of handlers. e.g.

    $object->on_warn( before => \&one, \&two );
    $object->on_warn( after  => \&one, \&two );

The on_warn() method returns a reference to the list, so you can also monkey about with it directly if you want the handler(s) to go somewhere else.

    my $handlers = $object->on_warn;
    shift(@$handlers, \&one);       # add before existing handlers
    push(@$handlers, \&two);        # add after existing handlers

You can also specify a method name as a warning handler. For example, if you want to automatically upgrade all warnings to errors for a particular object, you can write this:

    $object->on_warn('error');      # calls $object->error() on warnings

You can also specify 'warn' as a handler which will call Perl's "warn()" function. This is the default value. To explicitly disable any handlers, you can use a value of 0.

    $object->on_warn('warn');       # raise warning - the default action
    $object->on_warn(0);            # no warnings

These values can be imported from Badger::Constants as the "WARN" and "NONE" constants.

    use Badger::Constants 'WARN NONE';
    $object->on_warn(WARN);         # raise warning - the default action
    $object->on_warn(NONE);         # no warnings

The on_warn() method works equally well as a class method. In this case it sets the $ON_WARN package variable for the class. This acts as the default handler list for any objects of that class that don't explicitly define their own warning handlers.

    Your::Badger::Module->on_warn(\&handler_sub);

If you prefer you can define this using the $ON_WARN package variable. This will then be used as the default for all objects of this class.

    package Your::Badger::Module;
    use base 'Badger::Base';
    our $ON_WARN = \&handler_sub;

Multiple values should be defined using a list reference. Method names and the special "warn" flag can also be included.

    our $ON_WARN = [ \&this_code_first, 'this_method_next', 'warn' ]

error($message)

    sub engage {
        my $self = shift;
        return $self->error('warp drive offline');
    }

Multiple arguments can be passed to the "error()" method. They are concatenated into a single string.

    sub engage {
        my $self = shift;
        return $self->error(
            'warp drive ',
             $self->{ engine_no },
            ' is offline'
        );
    }

The error method can also be called without arguments to return the error message previously thrown by a call to "error()". In this case it performs exactly the same function as the reason() method.

    eval { $enterprise->engage }
        || warn "Could not engage: ", $enterprise->error;

The fact that the "error()" method can be called without arguments allows you to write things like this:

    # doesn't throw anything if list is empty
    $self->error(@list_of_errors);

An existing exception object can also be passed as a single argument to the error method. In this case, the exception object is re-thrown unmodified.

    sub save_the_world {
        eval { $self->world_peace };

        if ($@) {
            $self->call_international_rescue($@);   # call Thunderbirds
            return $self->error($@);                # re-throw error
        };
    }

ASIDE: You may have noticed in these examples that I'm using the "return" keyword when raising an error. For example:

    return $self->error('warp drive offline');

The "error()" method doesn't return when you call it with arguments so the "return" keyword has no effect whatsoever. However, I like to put it there to give a clear indication of what my intentions are at that point. It also means that the code will continue to return even if a subclass should "accidentally" define a different error() method that doesn't throw an error (don't laugh - it happens). It's also useful when used in conjunction with syntax highlighting to see at a glance where the potential exit points from a method are (assuming you make "return" bright red or something obvious like I do).

The "error()" method can also be called as a class method. In this case, it updates and retrieves the $ERROR package variable in the package of the subclass module. This can be used to raise and examine errors thrown by class methods.

    # calling package error() method
    my $object = eval { Your::Badger::Module->new() }
        || warn "Could not create badger module: ",
                Your::Badger::Module->error();

    # accessing $ERROR package variable
    my $object = eval { Your::Badger::Module->new() }
        || warn 'Could not create badger module: ",
                $Your::Badger::Module::ERROR;

on_error($handler, $another_handler, ...)

    $world->on_error(
        sub {
            my $message = shift;

            Thunderbirds->call({
                priority => IR_PRIORITY_HIGH,
                message  => $message,
            });

            return $message;    # forward message to next handler
        }
    );

The value returned from the callback is forwarded to the next handler. If a callback returns a false value or an empty list then the remaining handlers will not be called. However, the error will still be raised regardless of what any of the handlers do or return.

decline($reason, $more_reasons, ...)

    sub forage {
        my ($self, $name) = @_;

        # hard error if database isn't connected
        my $db = $self->{ database }
            || return $self->error('no database')

        if ($thing = $db->fetch_thing($name)) {
            # return true value on success
            return $thing;
        }
        else {
            # soft decline if not found
            return $self->decline("not found: $name")
        }
    }

Like error(), the decline() method can be called without arguments to return the most recent decline message, although it's probably better to use reason() which is designed specifically for that purpose. The decline() method can also be called as a class method as well as an object method, as per error().

declined()

    my $result = eval { $forager->fetch('nuts') };

    if ($result) {
        print "result: $result\n";
    }
    elsif ($forager->declined) {
        print "declined: ", $forager->reason, "\n";
    }
    else {
        print "error: ", $forager->reason, "\n";
    }

reason()

    $forager->forage('nuts and berries')
        || die $forager->reason;

message($type, @args)

    # base class
    package Badger::Example::One
    use base 'Badger::Base';

    our $MESSAGES = {
        hai => 'Hello %s',
    };

    # subclass
    package Badger::Example::Two;
    use base 'Badger::Example::One';

    our $MESSAGES = {
        bye => 'Goodbye %s',
    };

    # using the classes
    package main;

    my $two = Badger::Example::Two->new();
    $two->message( hai => 'World' );    # Hello World
    $two->message( bye => 'World' );    # Goodbye World

The $two object can use message formats defined in its own package ("Badger::Example::Two") and also those of its base class ("Badger::Example::One").

The messages are formatted using the xprintf() function in Badger::Utils. This is a thin wrapper around the built-in "sprintf()" function with some additional formatting controls to simplify the process of using positional arguments.

Messages are used internally by the error_msg() and decline_msg() methods for generating error messages, but you can use them for any kind of simple message generation.

There are a number of benefits to defining messages in a central repository like this.

First, it makes it easy to reuse the same message format in different places. Also known as the "DRY" principle - Don't Repeat Yourself.

Second, it allows you to put all your messages in one place instead of dotting them all over your code. The benefit here is a clearer separation of concerns between the underlying logic of your application and the presentational aspects.

The third benefit comes as a result of this clear separation - it becomes trivially easy to change the messages generated by your application because they're all defined in one place (possibly in several different modules if that's how you choose to break it down, but at least they're in one place in each of those modules). Possible applications of this include: localising an application to different spoken languages; generating messages in colour (as the Badger::Debug and Badger::Test modules do); or formatting messages as HTML.

warn_msg($message, @args)

    our $NAME     = 'Badger';
    our $MESSAGES = {
        using_default => "Using default value for %s: %s",
    };

    sub init {
        my ($self, $config) = @_;

        if ($config->{ name }) {
            $self->{ name } = $config->{ name };
        }
        else {
            $self->warn_msg( using_default => name => $NAME );
            $self->{ name } = $NAME;
        }

        return $self;
    }

If a "name" isn't provided as a configuration parameter then the default $NAME will be used and the following warning will be generated:

    Using default value for name: Badger

error_msg($message, @args)

    package Your::Zoo;
    use base 'Badger::Base';

    our $MESSAGES = {
        not_found => "I can't find the %s you asked for: %s",
    }

    sub animal {
        my ($self, $name) = @_;

        return $self->fetch_an_animal($name)
            || $self->error_msg( missing => animal => $name );
    }

Calling the "animal()" method on this object with an animal that can't be found, like this:

    $zoo->animal('Badgerpotamus');

Will generate an error message like this:

    your.zoo error - I can't find the animal you asked for: Badgerpotamus

decline_msg($message, @args)

    our $MESSAGES = {
        not_found => 'No %s found in the forest',
    };

    sub forage {
        my ($self, $name) = @_;

        return $self->database->fetch_item($name)
            || $self->decline_msg( not_found => $name );
    }

The reason() method can be used to return the message generated.

    my $food = $forager->forage('nuts')
        || warn $forager->reason;       # No nuts found in the forest

fatal_msg($message, @args)

throw($type, $info, %more_info)

    # error message - same thing as error()
    $object->throw('an error has occurred');

    # exception object
    $e = Badger::Exception->new(
        type => 'engine',
        info => 'warp drive offline'
    );
    $object->throw($e);

In the first case, the exception() and throws() methods will be called to determine the exception class (Badger::Exception by default) and type for the exception, respectively.

The method can also be called with two arguments. The first defines the exception "type", the second the error message.

    $object->throw( engine => 'warp drive offline' );

The second argument can also be another exception object. If the exception has the same type as the first argument then it is re-thrown unchanged.

    $e = Badger::Exception->new(
        type => 'engine',
        info => 'warp drive offline'
    );
    $object->throw( engine => $e ) };

In the example above, the $e exception already has a type of "engine" and so is thrown without change. If the exception types don't match, or if the exception isn't the right kind of exception object that we're expecting (as reported by exception()) then a new exception is thrown with the old one attached via the "info" field.

     $object->throw( propulsion => $e );

Here a new "propulsion" exception is thrown, with the previous "engine" exception linked in via the "info" field. The exception object has type() and info() methods that allow you to inspect its value, iteratively if necessary. Or you can just print an exception and rely on its overloaded stringification operator to call the text() method. For the error thrown in the previous example, that would be:

    propulsion error - engine error - warp drive offline

throw_msg($type, $message, @args)

    our $MESSAGES = {
        offline => '%s is offline',
    };

    sub engage {
        my $self = shift;
        $self->throw_msg( warp => offline => 'warp drive' )
            unless $self->warp_drive_ready;
        # make it so
    }

    # throws the following exception:
    warp error - warp drive is offline

try($method, @args)

    my $result = $object->try( fetch => 'answer' ) || 42;

This example is equivalent to:

    my $result = eval { $object->fetch('answer') } || 42;

The error thrown can be retrieved using the "reason()" method.

    my $result = $object->try( fetch => 'answer' )|| do {
        warn "Could not fetch answer: ", $object->reason;
        42;     # a sensible default
    };

If you call the "try()" method without any arguments then it will return a "Badger::Base::Trial" object as a wafer thin wrapper around the original object. Any methods called on this delegate object will be forwarded to the original object, wrapped up in an "eval" block to catch any errors thrown.

    my $result = $object->try->fetch('answer') ||= do {
        ...
    };

catch($type, $method, @args)

throws($type)

    # object method
    $object->throws('food');
    $object->throw('No nuts');              # food error - No nuts
    $object->error('No nuts');              # food error - No nuts

    # class method
    Badger::Example->throws('food');
    Badger::Example->throw('No berries');   # food error - No berries
    Badger::Example->error('No berries');   # food error - No berries

    # all objects of this class now throw food by default
    my $badger = Badger::Example->new;
    $badger->throw('No cheese');            # food error - No cheese
    $badger->error('No cheese');            # food error - No cheese

You can also set this value for an object by passing a "throws" configuration parameter to the new() constructor method.

    my $badger = Badger::Example->new(
        throws => 'food',
    );

This relies on the default behaviour of the init() method which stores a reference to the original configuration parameters in "$self->{config}". If you want to use this feature then you should ensure that any specialised init() method you define does the same thing, or copies the "throws" value from $config into "$self->{THROWS}".

    # example 1: store entire config for later
    sub init {
        my ($self, $config) = @_;
        $self->{ config } = $config;
        # do other stuff
        return $self;
    }

    # example 2: extract specific parameter up front
    sub init {
        my ($self, $config) = @_;
        $self->{ THROWS } = $config->{ throws };
        # do other stuff
        return $self;
    }

You can set the default exception type for your own modules that inherit from "Badger::Base" by adding a $THROWS package variable;

    package Badger::Example;
    use base 'Badger::Base';
    our $THROWS = 'food';

If you don't specify an exception type then one will be generated from the module's class name using the id() method in Badger::Class.

exception($class)

    use Badger::Example;
    use Some::Other::Exception;
    Badger::Example->exception('Some::Other::Exception');

    # now Badger::Example objects throw Some::Other::Exception

You can set the default exception class for your own modules that inherit from "Badger::Base" by adding a $EXCEPTION package variable;

    package Badger::Example;
    use base 'Badger::Base';
    use Some::Other::Exception;
    our $EXCEPTION = 'Some::Other::Exception';

fatal($info, $more_info, ...)

The most common reason for a fatal error being raised is calling the message() method (or either of the error_msg() or decline_msg() wrapper methods) with a message format that doesn't exist. The stack backtrace will tell you where in your code you're making the call so you can easily find and fix it.

not_implemented($what)

    sub example_method {
        shift->not_implemented;
    }

Calling the "example_method()" would result in an error message similar to this (shown here split across two lines):

    your.badger.module error - example_method() is not implemented
    for Your::Badger::Module in /path/to/your/script.pl at line 42

Note that it tells you where the "example_method()" was called from, not where the method is defined.

The "not_implemented()" method is typically used in methods defined in a base classes that subclasses are expected to re-define (a.k.a. pure virtual methods or abstract methods).

You can pass an argument to be more specific about what it is that isn't implemented.

    sub example_method {
        shift->not_implemented('in base class');
    }

The argument is added to the generated error message following the method name. A single space is also added to separate them.

    your.badger.module error - example_method() is not implemented in
    base class for Your::Badger::Module in ...etc...

todo($what)

    sub not_yet_working {
        shift->todo;
    }

The error message generated looks something like this:

    your.badger.module error - not_yet_working() is TODO in
    Your::Badger::Module at line 42

You can pass an argument to be more specific about what is still TODO.

    sub not_yet_working {
        my ($self, $x) = @_;
        if (ref $x) {
            $self->todo('support for references');
        }
        else {
            # do something
        }
    }

The error message generated would then be:

    your.badger.module error - not_yet_working() support for
    references is TODO in Your::Badger::Module at line 42

debug($msg1,$msg2,...)

    sub example {
        my $self = shift;
        $self->debug('entered example()');
        # ... some code ...
        $self->debug('leaving example()');
    }

debug_msg($message, @args)

    our $MESSAGES = {
        here => 'You are in %s',
    };

    sub example {
        my $self = shift;

        $self->debug_msg(
            here => 'a maze of twisty little passages, all alike'
        ) if DEBUG;

        # ... some code ...

        $self->debug_msg(
            here => 'boat, floating on a sea of purest green'
        ) if DEBUG;
    }

debug_up($level,$msg1,$msg2,...)

    sub parse {
        my $self = shift;

        while (my ($foo, $bar) = $self->get_foo_bar) {
            $self->trace($foo, $bar);               # report line here
            # do something
        }
    }

    sub trace {
        my ($self, $foo, $bar) = @_;
        $self->debug_up(2, "foo: $foo  bar: $bar"); # not here
    }

See Badger::Debug for further details.

$DEBUG

    $Badger::Base::DEBUG = 1;

The "Badger::Base" module does not use or define any $DEBUG variable in the subclasses derived from it. However, you may want to do something similar in your own modules to assist in debugging.

    package Your::Badger::Module;
    use base 'Badger::Base';

    # allow flag to be set before this module is loaded
    our $DEBUG = 0 unless defined $DEBUG;

    sub gnarly_method {
        my ($self, $item) = @_;
        $self->debug("gnarly_method($item)\n") if $DEBUG;
        # your gnarly code
    }

The "Badger::Class" module defines the "debug" method and import hook which will automatically define a $DEBUG variable for you.

    package Your::Badger::Module;

    use Badger::Class
        base  => 'Badger::Base',
        debug => 0;

$DECLINED

$ERROR

$EXCEPTION

Subclasses may define an $EXCEPTION package variable to change this value.

    package Your::Badger::Module;
    use base 'Badger::Base';
    use Your::Exception;
    our $EXCEPTION = 'Your::Exception';

Those that don't explicitly define an $EXCEPTION will inherit the value from any of their base classes, possibly coming all the way back up to the default value in "Badger::Base".

Calling the "exception()" class method with an argument will update the $EXCEPTION package variable in that class.

    # sets $Your::Badger::Module::EXCEPTION
    Your::Badger::Module->exception('Your::Exception');

$MESSAGES

    our $MESSAGES = {
        not_found       => '%s not found: %s',
        not_found_in    => '%s not found in %s',
        not_implemented => '%s is not implemented %s',
        no_component    => 'No %s component defined',
        bad_method      => "Invalid method '%s' called on %s at %s line %s",
        invalid         => 'Invalid %s specified: %s',
        unexpected      => 'Invalid %s specified: %s (expected a %s)',
        missing_to      => 'No %s specified to %s',
        missing         => 'No %s specified',
        todo            => '%s is TODO %s',
        at_line         => '%s at line %s',
        at_file_line    => '%s in %s at line %s',
    };

The message() method searches for $MESSAGES in the current class and those of any base classes. That means that any objects derived from "Badger::Base" can use these message formats.

    package Your::Badger::Module;
    use base 'Badger::Base';

    sub init {
        my ($self, $config) = @_;
        $self->{ name } = $config->{ name }
            || $self->error_msg( missing => $name );
        return $self;
    }

You can define additional $MESSAGES for your own classes.

    package Your::Badger::Module;
    use base 'Badger::Base';

    our $MESSAGES = {
        life_jim  => "It's %s Jim, but not as we know it",
    }

    sub bones {
        my ($self, $thing)= @_;
        $self->warn_msg( life_jim => $thing );
        return $self;
    }

Calling the "bones()" method like this:

    $object->bones('a badger');

will generate a warning like this:

    It's a badger Jim, but not as we know it.

$ON_ERROR

The "Badger::Base" module doesn't define any $ON_ERROR package variable by default. The on_error() method can be called as a class method to set the $ON_ERROR package variable.

    Your::Badger::Module->on_error(\&my_handler);

You can also define an $ON_ERROR handler or list of handlers in your module.

    package Your::Badger::Module;
    use base 'Badger::Base';

    # one of the following...
    our $ON_ERROR = 'warn';         # call Perl's warn()
    our $ON_ERROR = 'method_name';
    our $ON_ERROR = \&code_ref;
    our $ON_ERROR = [ 'warn', 'method_name', \&code_ref ];

    # code refs get message as first argument
    sub code_ref {
        my $message = shift;
        # do something...
    }

    # methods get implicit $self, then message argument
    sub method_name {
        my ($self, $message) = @_;
        # do something...
    }

$ON_WARN

$THROWS

    Your::Badger::Module->throws('food');

You can define $THROWS in your own modules that are derived from "Badger::Base".

    package Your::Badger::Module;
    use base 'Badger::Base';
    our $THROWS = 'food';

If the $THROWS value is not defined in the current class or any of an object's base classes, then the id() method is used to construct an identifier for the module to use instead.

config

If you've defined your own init() method then this item won't exist unless your init() method adds it explicitly.

DECLINED

ERROR

EXCEPTION

ON_ERROR

ON_WARN

THROWS