NAME

Object::Result - Allow subs to build and return objects on-the-fly

VERSION

This document describes Object::Result version 0.000003

SYNOPSIS

use Object::Result;

sub get_time {
    # Return an object indicating failure...
    if (!load Time::HiRes => 'time') {
        result { <FAIL> }
    }

    # Set up some lexical state information...
    my $time = time();

    # Return an object with methods that use that information...
    result {
        # Named methods for returned object...
        timestamp  { return scalar localtime($time) }

        delay (Num $offset)  { $time += $offset }

        # Coercions for returned object...
        <STR>  { $self->timestamp; }  # String --> timestamp method
        <NUM>  { time() - $time;   }  # Number --> age of object
        <BOOL> { $self < 60;       }  # True for the first minute
    }
};


# And later...

# Get back an object...
my $now = get_time();

# Use it (as a string, as an object, as a number)...
say "It's $now";
say $now->timestamp;
say 'Recent' if $now < 0.1;

# Change the object's internal state...
$now->delay(50);

# Use the object as a (dynamically valued) boolean...
while ($now) {
    sleep 5;
    say "Are we there yet?";
}

DESCRIPTION

This module adds a new keyword to Perl: result

That keyword acts like a return, but instead of a list of values to return, it takes a single block which specifies the behaviour (i.e. the methods and operator overloading) of an object to be returned.

The intention is to make it much less onerous to return clean, properly encapsulated objects...instead of returning lists of values or references to arrays or hashes.

For example, instead of:

my ($ID, $name, $uptime, $load, $users, $location, $contact)
    = get_server_status($server_ID);

if ($uptime) {
    say "$ID ($name) load: $load";
}

or:

my $server = get_server_status($server_ID);

if ($server->{uptime}) {
    say "$server->{ID} ($server->{name) load: $server->{load}";
}

you can arrange the API to be object-based:

my $server = get_server_status($server_ID);

if ($server->is_up) {
    say $server->describe, ' load: ', $server->load;
}

The real advantage is that, inside the module providing get_server_status() you don't have to define a separate class implementing the objects returned by that subroutine. More importantly, if you have a dozen subroutines returning specialized objects, you don't have to define a dozen separate classes to support them.

RATIONALE

(Skip straight ahead to INTERFACE if you already get it...or just don't care why this approach is better.)

Subroutines that return lists of values make client code brittle: it's far too easy to mess up the unpacking:

my ($ID, $name, $uptime, $users, $load, $contact, $location)
    = get_server_status($server_ID);

Subroutines that return hash references aren't much better: they do keep all the information together, and eliminate the order dependency of unpacking it, but it's also very easy to misspell a key and thereby create a silent bug:

my $server = get_server_status($server_ID);

if ($server->{up_time}) {
    say "$server->{id} ($server->{name) load: $server->{load}";
}

Like hashrefs, properly encapsulated objects keep all the returned information together and allow it to be retrieved by name, but they can also provide extra methods to simplify common tasks, and the OO syntax makes it a fatal error to misspell any access attempt:

my $server = get_server_status($server_ID);

if ($server->up_time) {
    say $server->describe, ' load: ', $server->lode;
}
# dies with: "No such method 'lode' at demo.pl line 27"

The only downside is that object-based return values are tedious to set up. If you have multiple subroutines in your API, each of them may need to return a unique type of object, which means you have to define as many distinct support classes as you have subroutines. And even with helper modules (such as Moose or Object::InsideOut) that's a substantial amount of extra work:

sub get_server_status ($server_ID)  {
    my $status_ref = _acquire_status_somehow_for($server_ID);

    return GSS::Result->new(status => $status_ref);
}

# Class implementing result objects for get_server_status()...
package GSS::Result {
    use Moose;

    has status => (is => 'ro', required => 1);

    sub ID       { my $self = shift; $self->status->{ID}       }
    sub name     { my $self = shift; $self->status->{name}     }
    sub uptime   { my $self = shift; $self->status->{uptime}   }
    sub load     { my $self = shift; $self->status->{load}     }
    sub users    { my $self = shift; $self->status->{users}    }
    sub location { my $self = shift; $self->status->{location} }
    sub contact  { my $self = shift; $self->status->{contact}  }

    sub is_up    { my $self = shift; $self->uptime > 0         }

    sub describe {
        my $self = shift;
        $self->ID . ' (' . $self->name . ')';
    }
}

The Object::Result module allows you to have your cake (per-subroutine return objects) without requiring quite so much baking (of per-subroutine support clases).

INTERFACE

The module lexically inserts a new keyword (result) into any scope when it is loaded.

That keyword takes a single block of code containing zero or more method specifications, and builds an object which supplies those methods. The keyword then causes the surrounding subroutine to immediately return that object.

Defining named methods

To define a normal named method for the result object, specify its name followed by a block implementing its body. For example, to specify that the result object has two methods: succeeded() and fitness():

result {
    succeeded { return $outcome > 0 }
    fitness   { return $outcome * $sample->{metric} }
}

Methods may be specifed with parameter lists (which are implemented by the Method::Signatures module):

result {
    succeeded (Num $threshold = 0) {
        return $outcome > $threshold
    }

    fitness {
        return $outcome * $sample->{metric}
    }
}

If not specified with a parameter list, result methods are assumed to take no arguments (which is Method::Signatures' default behaviour).

Data storage for result objects

Result methods get their information (e.g. $outcome and $sample) from the lexical variables declared in the surrounding subroutine. For example:

sub estimate_fitness ($sample, $environment) {
    my $outcome = $sample->{metric} < $environment->{max_impact}
        ? $environment->{fitness_func}->($sample->{max})
        : $environment->{max_survivability};

    result {
        succeeded (Num $threshold = 0) {
            return $outcome > $threshold
        }

        fitness {
            return $outcome * $sample->{metric}
        }
    }
}

In other words, the various methods defined in the result block become closures over the variables inside the surrounding subroutine, and the result object can use those variables as its private storage (i.e. its attributes/fields).

For example, the full get_server_status() subroutine shown earlier in RATIONALE could be implemented as:

sub get_server_status ($server_ID)  {
    my $status_ref = _acquire_status_somehow_for($server_ID);

    result {
        ID       { $status_ref->{ID}         }
        name     { $status_ref->{name}       }
        uptime   { $status_ref->{uptime}     }
        load     { $status_ref->{load}       }
        users    { $status_ref->{users}      }
        location { $status_ref->{location}   }
        contact  { $status_ref->{contact}    }
        is_up    { $status_ref->{uptime} > 0 }
        describe { "$status_ref->{ID} ($status_ref->{name})" }
}

using the lexical $status_ref as the object's attribute storage.

Note that methods can also modify this private data, so the result object from get_server_status() could also support subsequent annotations on the result:

sub get_server_status ($server_ID)  {
    my $status_ref = _acquire_status_somehow_for($server_ID);
    my @annotations;

    result {
        # All the methods defined in the previous version, plus...

        add_note  ($msg) { push @annotations, $msg; }
        get_notes        { return @annotations;     }

}

Coercions

Perl allows classes to specify coercive overloadings, so that their objects can then be treated as if they were booleans, strings, numbers, integers, or references to scalars, arrays, hashes, regexes, subroutines, or typeglobs.

The result keyword supports this kind of type-coercion too. Coercion methods can be specified by naming the specific type in angle brackets. For example, get_server_status() could return a result object that is true only if the server is up, and which stringifies to a printable summary of the status, like so:

sub get_server_status ($server_ID)  {
    my $status_ref = _acquire_status_somehow_for($server_ID);

    result {
        # All the methods defined in the previous version, plus...

        <BOOL>  { return $status_ref->{uptime} > 0;          }
        <STR>   { return $self->describe . ': ' $self->load; }
}

Note that, because all methods (named or coercive) are implemented via the Method::Signatures module, they automagically get a $self variable.

The following coercions are supported:

Treat object as boolean:                  <BOOL>
Treat object as string:                   <STR>
Treat object as integer:                  <INT>
Treat object as number:                   <NUM>
Treat object as scalar ref:               <SCALAR>
Treat object as array ref:                <ARRAY>
Treat object as hash ref:                 <HASH>
Treat object as typeblob ref/filehandle:  <GLOB>
Treat object as regex:                    <REGEXP>  or <REGEX>
Treat object as subroutine ref:           <CODE>    or <SUB>

In addition, the coercions to reference types can have the suffix REF appended to their names (e.g. <HASHREF>, <ARRAYREF>, <SUBREF>, etc.)

Coercion methods take no arguments (except the implicit $self) and are expected to return either a value of the appropriate type, or some other object with a suitable coercion overloading (i.e. the same requirements as for coercions specified via use overload).

By default, if a result object is asked for a particular coercion (apart from boolean; see "Boolean coercions"), but did not have that coercion explicitly defined, then the result object immediately throws an exception. See "Default coercions" for a way to change this behaviour.

head3 Boolean coercions

By default, any result object that has at least one method or coercion defined will evaluate true in a boolean context...as if they all had the following coercion implicitly defined:

result {
    <BOOL> { return 1 }
    ...
}

However, as a special case, result objects with no methods at all:

result { }

always evaluate false (besides having several other useful features; see "Result objects for failure signalling").

You can override these default boolean coercion behaviours simply by defining an explicit <BOOL> coercion yourself:

result {
    <BOOL> { return defined $outcome }
    ...
}

Or, if a result object has other methods, but should nevertheless always evaluate false, you can define that explicitly too:

result {
    <BOOL> { return 0 }
    ...
}

Note, however, that in such cases it may be more effective to use a failure object.

Default coercions

Although "missing" coercions default to throwing an exception, it's also possible to specify that something else should happen when an unimplemented coercion is requested...by using the <DEFAULT> specifier.

For example, to convert the normal exception-throwing response into merely warning about an unimplemented coercion:

result {
    # Actual methods and coercions here

    <DEFAULT> ($requested_coercion, $obj_origin) {
        carp "Can't convert result of $obj_origin to $requested_coercion";
    }
}

Or to revert the object to Perl's built-in behaviours (i.e. address-as-integer in a numeric context, "REFTYPE=CLASSNAME(0xADDRESS)" in a string context, etc.) you could specify:

result {
    # Actual methods and coercions here

    <DEFAULT> { return $self }
}

As the first example implies, the <DEFAULT> coercion method is passed two extra arguments apart from the usual $self. The first argument is a string containing the name of the missing coercion that was requested: '<STR>', or '<NUM>', or '<HASH>', etc. The second argument is a string indicating the origin of the result object being coerced. That second argument is of the form:

'call to __SUBNAME__() at __FILE__ line __LINE__'

and may be useful for generating more informative warnings or errors within a <DEFAULT> coercion.

Cleaning up result objects

If a result object manages some external resource, you can also set up a destructor for that object, using the <DESTROY> pseudo-coercion:

sub open_output_file ($filename) {
    open my $fh, '>', $filename or croak $!;

    result {
        write   (@whatever)  { print @whatever; }
        writeln (@whatever)  {   say @whatever; }

        <BOOL> { return not $fh->eof  }
        <GLOB> { return $fh           }

        # Make sure file is flushed before closing...
        <DESTROY> {
            $fh->flush();
            $fh->close();
        }
    }
}

Result objects for failure signalling

The default behaviour of missing coercions provides an easy way to produce so-called "contingent exceptions" (a.k.a. "failure objects").

In particular, a result statement of the form:

result { <FAIL> }

or its exact equivalent, an "empty" result:

result { }

returns an object that evaluates false in boolean contexts and throws an exception when used in any other way (i.e. whenever it has a method called on it, or it is used as a string, number, regex, or reference). The result object also throws an exception if it is destroyed without having been tested in a boolean context.

That is, the two forms shown above are equivalent to something like:

{
    my $tested_as_boolean = 0;
    my $error_msg = $@ // $! // $?;
    result {
        <BOOL>    { $tested_as_boolean++; return 0;          }
        <DEFAULT> { croak $error_msg;                        }
        <DESTROY> { croak $error_msg if !$tested_as_boolean; }
    }
}

If you want the exceptions to throw something else, you can give the <FAIL> specifier a block, which will then be called instead to generate the argument(s) to <croak():

# Throw a different string as the exception...
result {
    <FAIL> { "Could not load file: $!" }
}

# Throw an exception object...
result {
    <FAIL> { X::File::NoLoad->new($!) }
}

Failure objects such as these are a useful way of signalling errors, because the client code can test the result object (in which case it evaluates false like a typical undef or 0 or "" failure value):

my $status = get_server_status();

if ($status) {     # 'if' test fails if sub returned failure object
    say $status;
}

Or the client code can decide not to bother testing it, in which case it dies when used in any non-boolean way (or not used at all):

my $status = get_server_status();

say $status;       # Dies here if sub returned failure object

Thus the exception-on-failure is contingent: it will be thrown if the object is used in (almost) any way, unless the failure object has been "defused" by testing it for in a boolean context.

DIAGNOSTICS

Compile-time diagnostics

"Invalid syntax for 'result' statement. Expected block but found %s"

The result keyword take a single block after it. You put something else there instead.

Maybe you just needed a regular return, instead of result?

"Missing definition for %s method"

You declared the name of a method or coercion within the result's block, but didn't give it an implementation (i.e. a block of code for it to execute).

"Invalid definition for %s method. Expected parameter list or method block, but found %s"

You declared the name of a method or coercion within the result's block, so the module next expected to see either a parameter list specification or else a block implementing the method or coercion. It reported the syntax error, because it found something else instead immediately after the method name.

Run-time diagnostics

"Object returned by %s can't be used as %s"

You tried to coerce a result object returned by result to some other type (a number, a string, a reference, etc.) but the result block didn't explicitly specify that coercion.

Add the appropriate coercion specification to the result block.

Or, if you just wanted the vanilla Perl behaviours when coercing such result objects, add:

<DEFAULT> { $self }

to the result block.

"Call to %s failed: %s Failure detected at %s"

You called a subroutine that returned a failure object (i.e. a result {<FAIL>}).

Such failure objects can only be tested for their boolean value. Doing anything else with them (or nothing at all with them) will produce this exception...because that's what <FAIL> is supposed to do.

CONFIGURATION AND ENVIRONMENT

Object::Result requires no configuration files or environment variables.

DEPENDENCIES

Keyword::Simple: To install the result keyword.
PPI: To parse the new result-block syntax cleanly.
Method::Signatures: To support signatures on methods declared within a result block. Also used internally within the module's own implementation.

INCOMPATIBILITIES

None reported.

BUGS AND LIMITATIONS

No bugs have been reported.

Please report any bugs or feature requests to bug-object-result@rt.cpan.org, or through the web interface at http://rt.cpan.org.

AUTHOR

Damian Conway <DCONWAY@CPAN.org>

LICENCE AND COPYRIGHT

This module is free software; you can redistribute it and/or modify it under the same terms as Perl itself. See perlartistic.

DISCLAIMER OF WARRANTY

BECAUSE THIS SOFTWARE IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY FOR THE SOFTWARE, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE SOFTWARE "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE SOFTWARE IS WITH YOU. SHOULD THE SOFTWARE PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR, OR CORRECTION.

IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR REDISTRIBUTE THE SOFTWARE AS PERMITTED BY THE ABOVE LICENCE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE SOFTWARE (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A FAILURE OF THE SOFTWARE TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.

To install Object::Result, copy and paste the appropriate command in to your terminal.

cpanm

cpanm Object::Result

CPAN shell

perl -MCPAN -e shell
install Object::Result

For more information on module installation, please visit the detailed CPAN module installation guide.

	Global
`s`	Focus search bar
`?`	Bring up this help dialog

	GitHub
`g` `p`	Go to pull requests
`g` `i`	go to github issues (only if github is preferred repository)

	POD
`g` `a`	Go to author
`g` `c`	Go to changes
`g` `i`	Go to issues
`g` `d`	Go to dist
`g` `r`	Go to repository/SCM
`g` `s`	Go to source
`g` `b`	Go to file browse

	Search terms
module: (e.g. module:Plugin)
distribution: (e.g. distribution:Dancer auth)
author: (e.g. author:SONGMU Redis)
version: (e.g. version:1.00)