{
$SADI::Simple::Base::VERSION = '0.15';
}
use strict;
use URI;
use Log::Log4perl qw(:no_extra_logdie_message);
our $AUTOLOAD;
# names of attribute's types
use constant STRING => 'string';
use constant INTEGER => 'integer';
use constant FLOAT => 'float';
use constant BOOLEAN => 'boolean';
use constant DATETIME => 'datetime';
# names of attribute's properties
use constant TYPE => 'type';
use constant POST => 'post';
use constant ISARRAY => 'is_array';
use constant READONLY => 'readonly';
#-----------------------------------------------------------------
# These methods are called by set/get methods of the sub-classes. If
# it comes here, it indicates that an attribute being get/set does not
# exist.
#-----------------------------------------------------------------
{
my %_allowed = ();
sub _accessible {
my ( $self, $attr ) = @_;
exists $_allowed{$attr};
}
sub _attr_prop {
my ( $self, $attr_name, $prop_name ) = @_;
my $attr = $_allowed{$attr_name};
return ref($attr) ? $attr->{$prop_name} : $attr if $attr;
return undef;
}
}
#-----------------------------------------------------------------
# new
#-----------------------------------------------------------------
sub new {
my ( $class, @args ) = @_;
# $LOG->debug ("NEW: $class - " . join (", ", @args)) if $LOG->is_debug;
# create an object
my $self = bless {}, ref($class) || $class;
# initialize the object
$self->init();
# set all @args into this object with 'set' values
my (%args) = ( @args == 1 ? ( value => $args[0] ) : @args );
foreach my $key ( keys %args ) {
no strict 'refs';
$self->$key( $args{$key} );
}
# done
return $self;
}
#-----------------------------------------------------------------
# init
#-----------------------------------------------------------------
sub init {
my ($self) = shift;
}
#-----------------------------------------------------------------
# toString
#-----------------------------------------------------------------
sub toString {
my ( $self, $something_else ) = @_;
require Data::Dumper;
if ($something_else) {
return Data::Dumper->Dump( [$something_else], ['M'] );
} else {
return Data::Dumper->Dump( [$self], ['M'] );
}
}
#-----------------------------------------------------------------
#
# Error handling
#
#-----------------------------------------------------------------
my $DEFAULT_THROW_WITH_LOG = 0;
my $DEFAULT_THROW_WITH_STACK = 1;
#-----------------------------------------------------------------
# throw
#-----------------------------------------------------------------
sub throw {
my ( $self, $msg ) = @_;
$msg .= "\n" unless $msg =~ /\n$/;
my $LOG = Log::Log4perl->get_logger(__PACKAGE__);
# make an instance, if called as a class method
unless ( ref $self ) {
no strict 'refs';
$self = $self->new;
}
# add (optionally) stack trace
$msg ||= 'An error.';
my $with_stack = (
defined $self->enable_throw_with_stack
? $self->enable_throw_with_stack
: $DEFAULT_THROW_WITH_STACK
);
my $result = ( $with_stack ? $self->format_stack($msg) : $msg );
# die or log and die?
my $with_log = (
defined $self->enable_throw_with_log
? $self->enable_throw_with_log
: $DEFAULT_THROW_WITH_LOG
);
if ($with_log) {
$LOG->logdie($result);
} else {
die($result);
}
}
#-----------------------------------------------------------------
# Some throwing options
#
# These options are not set by using AUTOLOAD (as other regular
# attributes) because AUTOLOAD could raise exception and we would be
# in a deep..., well deep recursion.
#
# Default values are: NO enable_throw_with_log
# YES enable_throw_with_stack
# (but they are globally changeable by calling
# default_throw_with_log and default_throw_with_stack)
#
#-----------------------------------------------------------------
sub enable_throw_with_log {
my ( $self, $value ) = @_;
$self->{enable_throw_with_log} = ( $value ? 1 : 0 )
if ( defined $value );
return $self->{enable_throw_with_log};
}
sub default_throw_with_log {
my ( $self, $value ) = @_;
$DEFAULT_THROW_WITH_LOG = ( $value ? 1 : 0 )
if defined $value;
return $DEFAULT_THROW_WITH_LOG;
}
sub enable_throw_with_stack {
my ( $self, $value ) = @_;
$self->{enable_throw_with_stack} = ( $value ? 1 : 0 )
if defined $value;
return $self->{enable_throw_with_stack};
}
sub default_throw_with_stack {
my ( $self, $value ) = @_;
$DEFAULT_THROW_WITH_STACK = ( $value ? 1 : 0 )
if defined $value;
return $DEFAULT_THROW_WITH_STACK;
}
#-----------------------------------------------------------------
# format_stack
#-----------------------------------------------------------------
sub format_stack {
my ( $self, $msg ) = @_;
my $stack = $self->_reformat_stacktrace($msg);
my $class = ref($self) || $self;
my $title = "------------- EXCEPTION: $class -------------";
my $footer = "\n" . '-' x CORE::length($title);
return "\n$title\nMSG: $msg\n" . $stack . $footer . "\n";
}
#-----------------------------------------------------------------
# _reformat_stacktrace
# Taken from bioperl.
#
# Takes one argument - an error message. It uses it to remove its
# repeated occurences from each line (not to print it).
#
# Reformatting of the stack:
# 1. Shift the file:line data in line i to line i+1.
# 2. change xxx::__ANON__() to "try{} block"
# 3. skip the "require" and "Error::subs::try" stack entries (boring)
# This means that the first line in the stack won't have
# any file:line data.
#-----------------------------------------------------------------
sub _reformat_stacktrace {
my ( $self, $msg ) = @_;
my $stack = Carp->longmess;
$stack =~ s/\Q$msg//;
my @stack = split( /\n/, $stack );
my @new_stack = ();
my ( $method, $file, $linenum, $prev_file, $prev_linenum );
my $stack_count = 0;
foreach my $i ( 0 .. $#stack ) {
if ( ( $stack[$i] =~ /^\s*([^(]+)\s*\(.*\) called at (\S+) line (\d+)/ )
|| ( $stack[$i] =~ /^\s*(require 0) called at (\S+) line (\d+)/ ) )
{
( $method, $file, $linenum ) = ( $1, $2, $3 );
$stack_count++;
} else {
next;
}
if ( $stack_count == 1 ) {
push @new_stack, "STACK: $method";
( $prev_file, $prev_linenum ) = ( $file, $linenum );
next;
}
if ( $method =~ /__ANON__/ ) {
$method = "try{} block";
}
if ( ( $method =~ /^require/ and $file =~ /Error\.pm/ )
|| ( $method =~ /^Error::subs::try/ ) )
{
last;
}
push @new_stack, "STACK: $method $prev_file:$prev_linenum";
( $prev_file, $prev_linenum ) = ( $file, $linenum );
}
push @new_stack, "STACK: $prev_file:$prev_linenum";
return join "\n", @new_stack;
}
#-----------------------------------------------------------------
# Set methods test whether incoming value is of a correct type.
# Here we return message explaining that it isn't.
#-----------------------------------------------------------------
sub _wrong_type_msg {
my ( $self, $given_type_or_value, $expected_type, $method ) = @_;
my $msg = 'In method ';
if ( defined $method ) {
$msg .= $method;
} else {
$msg .= ( caller(1) )[3];
}
return (
"$msg: Trying to set '$given_type_or_value' but '$expected_type' is expected."
);
}
#-----------------------------------------------------------------
# Deal with 'set', 'get' and 'add_' methods.
#-----------------------------------------------------------------
sub AUTOLOAD {
my ( $self, @new_values ) = @_;
my $ref_sub;
if ( $AUTOLOAD =~ /.*::(\w+)/ && $self->_accessible("$1") ) {
# get/set method
my $attr_name = "$1";
my $attr_type = $self->_attr_prop( $attr_name, TYPE ) || STRING;
my $attr_post = $self->_attr_prop( $attr_name, POST );
my $attr_is_array = $self->_attr_prop( $attr_name, ISARRAY );
my $attr_readonly = $self->_attr_prop( $attr_name, READONLY );
$ref_sub = sub {
local *__ANON__ = "__ANON__$attr_name" . "_" . ref($self);
my ( $this, @values ) = @_;
return $this->_getter($attr_name) unless @values;
$self->throw("Sorry, the attribute '$attr_name' is read-only.")
if $attr_readonly;
# here we continue with 'set' method:
if ($attr_is_array) {
my @result =
( ref( $values[0] ) eq 'ARRAY' ? @{ $values[0] } : @values );
foreach my $value (@result) {
$value = $this->check_type( $AUTOLOAD, $attr_type, $value );
}
$this->_setter( $attr_name, $attr_type, \@result );
} else {
$this->_setter(
$attr_name,
$attr_type,
$this->check_type(
$AUTOLOAD, $attr_type, @values
)
);
}
# call post-procesing (if defined)
$this->$attr_post( $this->{$attr_name} ) if $attr_post;
return $this->{$attr_name};
};
} elsif ( $AUTOLOAD =~ /.*::add_(\w+)/ && $self->_accessible("$1") ) {
# add_XXXX method
my $attr_name = "$1";
my $attr_post = $self->_attr_prop( $attr_name, POST );
if ( $self->_attr_prop( $attr_name, ISARRAY ) ) {
my $attr_type = $self->_attr_prop( $attr_name, TYPE ) || STRING;
$ref_sub = sub {
local *__ANON__ = "__ANON__$attr_name" . "_" . ref($self);
my ( $this, @values ) = @_;
if (@values) {
my @result = (
ref( $values[0] ) eq 'ARRAY'
? @{ $values[0] }
: @values );
foreach my $value (@result) {
$value =
$this->check_type( $AUTOLOAD, $attr_type, $value );
}
$this->_adder( $attr_name, $attr_type, @result );
}
# call post-procesing (if defined)
$this->$attr_post( $this->{$attr_name} ) if $attr_post;
return $this;
}
} else {
$self->throw(
"Method '$AUTOLOAD' is allowed only for array-type attributes."
);
}
} else {
$self->throw("No such method: $AUTOLOAD");
}
no strict 'refs';
*{$AUTOLOAD} = $ref_sub;
use strict 'refs';
return $ref_sub->( $self, @new_values );
}
#-----------------------------------------------------------------
# The low level get/set methods. They are called from AUTOLOAD, and
# they are separated here so they can be overriten - as they are in
# the service skeletons, for example. Also, there may be situation
# that one can call them if other features (such as type checking) are
# not required.
#-----------------------------------------------------------------
sub _getter {
my ( $self, $attr_name ) = @_;
return $self->{$attr_name};
}
sub _setter {
my ( $self, $attr_name, $attr_type, $value ) = @_;
$self->{$attr_name} = $value;
}
sub _adder {
my ( $self, $attr_name, $attr_type, @values ) = @_;
push( @{ $self->{$attr_name} }, @values );
}
#-----------------------------------------------------------------
# Keep it here! The reason is the existence of AUTOLOAD...
#-----------------------------------------------------------------
sub DESTROY {
}
#-----------------------------------------------------------------
#
# Check type of @value against $expected_type. Return checked $value
# (perhaps trimmed, or otherwise corrected - e.g. wrapped in an
# appropriate object), or undef if the $value is of a wrong type.
#
#-----------------------------------------------------------------
sub check_type {
my ( $self, $name, $expected_type, @values ) = @_;
my $value = $values[0];
# first process cases when an expected type is a simple string,
# integer etc. (not SADI::Data::String etc.) - e.g. when an ID
# attribute is being set
if ( $expected_type eq STRING ) {
return $value;
} elsif ( $expected_type eq INTEGER ) {
$self->throw( $self->_wrong_type_msg( $value, $expected_type, $name ) )
unless $value =~ m/^\s*[+-]?\s*\d+\s*$/;
$value =~ s/\s//g;
return $value;
} elsif ( $expected_type eq FLOAT ) {
$self->throw( $self->_wrong_type_msg( $value, $expected_type, $name ) )
unless $value =~
m/^\s*[+-]?\s*(\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?\s*$/;
$value =~ s/\s//g;
return $value;
} elsif ( $expected_type eq BOOLEAN ) {
return ( $value =~ /true|\+|1|yes|ano/ ? '1' : '0' );
} elsif ( $expected_type eq DATETIME ) {
my $iso;
eval {
$iso = (
HTTP::Date::time2isoz(
HTTP::Date::str2time( HTTP::Date::parse_date($value) )
)
);
};
$self->throw( $self->_wrong_type_msg( $value, 'ISO-8601', $name ) )
if $@;
return $iso; ### $iso =~ s/ /T/; ??? TBD
} else {
# Then process cases when the expected type is a name of a
# real object (e.g. SADI::Data::String); for these cases the
# $value[0] can be already such object - in which case nothing
# to be done; or $value[0] can be HASH, or @values can be a
# list of name/value pairs, in which case a new object (of
# type $expected_type) has to be created and initialized by
# @values; and, still in the latter case, if the @values has
# just one element (XX), this element is considered a 'value':
# it is treated as a a hash {value => XX}.
return $value if UNIVERSAL::isa( $value, $expected_type );
$value = { value => $value }
unless ref($value) || @values > 1;
my ($value_ref_type) = ref($value);
if ( $value_ref_type eq 'HASH' ) {
# e.g. $sequence->Length ( { value => 12, id => 'IR64'} )
return $self->create_member( $name, $expected_type, %$value );
} elsif ( $value_ref_type eq 'ARRAY' ) {
# e.g. $sequence->Length ( [ value => 12, id => 'IR64'] )
return $self->create_member( $name, $expected_type, @$value );
} elsif ($value_ref_type) {
# e.g. $sequence->Length ( new SADI::Data::Integer ( value => 12) )
$self->throw(
$self->_wrong_type_msg(
$value_ref_type, $expected_type, $name
)
) unless UNIVERSAL::isa( $value, $expected_type );
return $value;
} else {
# e.g. $sequence->Length (value => 12, id => 'IR64')
return $self->create_member( $name, $expected_type, @values );
}
}
}
#-----------------------------------------------------------------
#
#-----------------------------------------------------------------
sub create_member {
my ( $self, $name, $expected_type, @values ) = @_;
eval "require $expected_type";
$self->throw( $self->_wrong_type_msg( $values[0], $expected_type, $name ) )
if $@;
return "$expected_type"->new(@values);
}
1;
__END__
=head1 NAME
SADI::Simple::Base - Hash-based abstract super-class for all SADI objects
=head1 SYNOPSIS
use base qw( SADI::Base );
$self->throw ("This is an error");
=head1 DESCRIPTION
This is a hash-based implementation of a general sadi
super-class. Most SADI objects should inherit from this.
=head1 ACKNOWLEDGEMENTS
Thanks to Martin Senger, the original author of this code, and
also to Edward Kawas who refactored it into it's current form.
=head1 ACCESSIBLE ATTRIBUTES
Most of the SADI objects are just containers of other objects (attributes,
members). Therefore, in order to crete a new SADI object it
is often enough to inherit from this C<SADI::Base> and to list allowed
attributes. The object lists only new, additional, attributes (those
defined in its parent classes are already available).
This is done by creating a I<closure> with a list of allowed attribute
names. These names correspond with the allowed I<get> and I<set>
methods. For example:
{
my %_allowed =
(
id => undef,
namespace => undef,
);
}
The closure above allows to call:
$obj->id; # a get method
$obj->id ('my id'); # a set method
$obj->namespace; # a get method
$obj->namespace ('my ns'); # a set method
Well, not yet. The closure also needs two methods that access these
(and only these - that is why it is a closure, after all)
attributes. Here they are:
{
my %_allowed =
(
id => undef,
namespace => undef,
);
sub _accessible {
my ($self, $attr) = @_;
exists $_allowed{$attr} or $self->SUPER::_accessible ($attr);
}
sub _attr_prop {
my ($self, $attr_name, $prop_name) = @_;
my $attr = $_allowed {$attr_name};
return ref ($attr) ? $attr->{$prop_name} : $attr if $attr;
return $self->SUPER::_attr_prop ($attr_name, $prop_name);
}
}
More about these methods in a moment.
Each attribute has also associated some properties (that is why we
need the second method in the closure, the C<_attr_prop>). For example:
{
my %_allowed =
(
id => undef,
namespace => undef,
date => {type => SADI::Base->DATETIME},
numbers => {type => SADI::Base->INTEGER, is_array => 1},
primitive => {type => SADI::Base->BOOLEAN},
);
...
}
The recognized property names are:
=over
=item B<type>
It defines a type of its attribute. It can be a primitive type - one
of those defined as constants in C<SADI::Base>
(e.g. "SADI::Base->INTEGER") - or a name of a real object
(e.g. C<SADI::RDF::Core>).
When an attribute new value is being set it is checked against this
type, and an exception is thrown if the value does not comply with the
type.
Default type (used also when the whole properties are undef) is
"SADI::Base->STRING".
=item B<is_array>
A boolean property. If set to true it allows to set more values to
this attribute. It also allows to call a method prefixed with C<add_>
to add a new value (or values) to this attribute.
Default value is C<false>.
Recognized values for C<true> are: C<1>, C<yes>, C<true>, C<+> and
C<ano>. Anything else is considered C<false>.
=item B<readonly>
A boolean property. If set to true the atribute can only be read.
=item B<post>
A property containing a reference to a subroutine. This subroutine is
called after a new value was set. It allows to do some
post-processing. For example:
{
my %_allowed =
(
value => {post => sub { shift->{isValueCDATA} = 0; } },
);
...
}
=back
Now we know what attribute properties are - so we can define what
these methods in closure do (even though you do not need to know -
unless C<The Law of Leaky Abstractions> starts showing).
=over
=item C<_accessible ($attr_name)>
Return 1 if the parameter C<$attr_name> is an allowed name to be
set/get in this class; otherwise, pass it to the parent class.
=item C<_attr_prop ($attr_name, $prop_name)>
Return a value of a property given by name $prop_name for given
attribute $attr_name; if such attribute does not exist here, pass it
to the parent class.
=back
=head1 THROWING EXCEPTIONS
One of the functionalities that C<SADI::Base> provides is the ability
to B<throw()> exceptions with pretty stack traces.
=head2 throw
Throw an exception. An argument is an error message.
=head2 format_stack
Return a nicely formatted stack trace. The resul includes also an
error message given as a scalar argument. Usually, this method is not
called directly but via C<throw> (unless C<enable_throw_with_stack>
was set to true).
print $self->format_stack ("Something terrible happen.");
=head1 OTHER SUBROUTINES
=head2 new
Create an empty hash-based object. Then call B<init()> in order to do
any initializing steps. This class provides only an empty C<init()>
but sub-classes may have it richer. Finally, fill the new object with
the given arguments (name/value pairs). The filling is done via C<set>
methods - which means that only attributes allowed for this particular
object can be used.
Arguments are name/value pairs. A special case is allowed: when a
single element argument occurs, it is treated as a "value". For
example, it is allowed to write:
$sadiint = new SADI::Data::Integer (42);
instead of a long way (doing the same):
$sadiint = new SADI::Data::Integer (value => 42);
=head2 init
Called after an object has been created (in B<new()>) and before the
values given in the constructor have been set. No arguments.
If your sub-class implements this method, make sure that it calls also
the same method of its super class:
sub init {
my ($self) = shift;
$self->SUPER::init();
# ... here do what you wish to do
# ...
}
=head2 toString
Return an (almost) human-readable description of any object.
Without any parameter, it stringifies the caller object
(self). Otherwise it stringifies the object given as parameter.
print $self->toString;
my $good_stuff = { yes => [1,2,3],
no => { net => 'R', nikoliv => 'C' },
};
print $self->toString ($good_stuff);
=cut