package Class::XPath; use 5.006; use strict; use warnings; our $VERSION = '1.4'; use Carp qw(croak); use constant DEBUG => 0; # regex fragment for names in XPath expressions our $NAME = qr/[\w:]+/; # declare prototypes sub foreach_node (&@); # handle request to build methods from 'use Class::XPath'. sub import { my $pkg = shift; return unless @_; my $target = (caller())[0]; # hand off to add_methods $pkg->add_methods(@_, target => $target, from_import => 1); } { # setup lists of required params my %required = map { ($_,1) } qw(get_name get_parent get_children get_attr_names get_attr_value get_content get_root call_match call_xpath); # add the xpath and match methods to sub add_methods { my $pkg = shift; my %args = (call_match => 'match', call_xpath => 'xpath', @_); my $from_import = delete $args{from_import}; my $target = delete $args{target}; croak("Missing 'target' parameter to ${pkg}->add_methods()") unless defined $target; # check args local $_; for (keys %args) { croak("Unrecognized parameter '$_' " . ($from_import ? " on 'use $pkg' line. " : "passed to ${pkg}->add_methods()")) unless $required{$_}; } for (keys %required) { croak("Missing required parameter '$_' " . ($from_import ? " on 'use $pkg' line. " : "in call to ${pkg}->add_methods()")) unless exists $args{$_}; } # translate get_* method names to sub-refs for (grep { /^get_/ } keys %args) { next if ref $args{$_} and ref $args{$_} eq 'CODE'; $args{$_} = eval "sub { shift->$args{$_}(\@_) };"; croak("Unable to compile sub for '$_' : $@") if $@; } # install code into requested names to call real match/xpath with # supplied %args { no strict 'refs'; *{"${target}::$args{call_match}"} = sub { $pkg->match($_[0], \%args, $_[1]) }; *{"${target}::$args{call_xpath}"} = sub { $pkg->xpath($_[0], \%args) } } }} sub match { my ($pkg, $self, $args, $xpath) = @_; my ($get_root, $get_parent, $get_children, $get_name) = @{$args}{qw(get_root get_parent get_children get_name)}; croak("Bad call to $args->{call_match}: missing xpath argument.") unless defined $xpath; print STDERR "match('$xpath') called.\n" if DEBUG; # / is the root. This should probably work as part of the # algorithm, but it doesn't. return $get_root->($self) if $xpath eq '/'; # . is self. This should also work as part of the algorithm, # but it doesn't. return $self if $xpath eq '.'; # break up an incoming xpath into a set of @patterns to match # against a list of @target elements my (@patterns, @targets); # target aquisition if ($xpath =~ m!^//(.*)$!) { $xpath = $1; # this is a match-anywhere pattern, which should be tried on # all nodes foreach_node { push(@targets, $_) } $get_root->($self), $get_children; } elsif ($xpath =~ m!^/(.*)$!) { $xpath = $1; # this match starts at the root @targets = ($get_root->($self)); } elsif ($xpath =~ m!^\.\./(.*)$!) { $xpath = $1; # this match starts at the parent @targets = ($get_parent->($self)); } elsif ($xpath =~ m!^\./(.*)$!) { $xpath = $1; @targets = ($self); } else { # this match starts here @targets = ($self); } # pattern breakdown my @parts = split('/', $xpath); my $count = 0; for (@parts) { $count++; if (/^$NAME$/) { # it's a straight name match push(@patterns, { name => $_ }); } elsif (/^($NAME)\[(-?\d+)\]$/o) { # it's an indexed name push(@patterns, { name => $1, index => $2 }); } elsif (/^($NAME)\[\@($NAME)\s*=\s*"([^"]+)"\]$/o or /^($NAME)\[\@($NAME)\s*=\s*'([^']+)'\]$/o) { # it's a string attribute match push(@patterns, { name => $1, attr => $2, value => $3 }); } elsif (/^($NAME)\[\@($NAME)\s*(=|>|<|<=|>=|!=)\s*(\d+)\]$/o) { # it's a numeric attribute match push(@patterns, { name => $1, attr => $2, op => $3, value => $4 }); } elsif (/^($NAME)\[($NAME|\.)\s*=\s*"([^"]+)"\]$/o or /^($NAME)\[($NAME|\.)\s*=\s*'([^']+)'\]$/o) { # it's a string child match push(@patterns, { name => $1, child => $2, value => $3 }); } elsif (/^($NAME)\[($NAME|\.)\s*(=|>|<|<=|>=|!=)\s*(\d+)\]$/) { # it's a numeric child match push(@patterns, { name => $1, child => $2, op => $3, value => $4 }); } elsif (/^\@($NAME)$/) { # it's an attribute name push(@patterns, { attr => $1 }); # it better be last croak("Bad call to $args->{call_match}: '$xpath' contains an attribute selector in the middle of the expression.") if $count != @parts; } else { # unrecognized token croak("Bad call to $args->{call_match}: '$xpath' contains unknown token '$_'"); } } croak("Bad call to $args->{call_match}: '$xpath' contains no search tokens.") unless @patterns; # apply the patterns to all available targets and collect results my @results = map { $pkg->_do_match($_, $args, @patterns) } @targets; return @results; } # the underlying match engine. this takes a list of patterns and # applies them to child elements sub _do_match { my ($pkg, $self, $args, @patterns) = @_; my ($get_parent, $get_children, $get_name, $get_attr_value, $get_attr_names, $get_content) = @{$args}{qw(get_parent get_children get_name get_attr_value get_attr_names get_content)}; local $_; print STDERR "_do_match(" . $get_name->($self) . " => " . join(', ', map { '{' . join(',', %$_) . '}' } @patterns) . ") called.\n" if DEBUG; # get pattern to apply to direct descendants my $pat = shift @patterns; # find matches and put in @results my @results; my @kids; { no warnings 'uninitialized'; @kids = grep { $get_name->($_) eq $pat->{name} } $get_children->($self); } if (defined $pat->{index}) { # get a child by index push @results, $kids[$pat->{index}] if (abs($pat->{index}) <= $#kids); } elsif (defined $pat->{attr}) { if (defined $pat->{name}) { # default op is 'eq' for string matching my $op = $pat->{op} || 'eq'; # do attribute matching foreach my $kid (@kids) { my $value = $get_attr_value->($kid, $pat->{attr}); push(@results, $kid) if ($op eq 'eq' and $value eq $pat->{value}) or ($op eq '=' and $value == $pat->{value}) or ($op eq '!=' and $value != $pat->{value}) or ($op eq '>' and $value > $pat->{value}) or ($op eq '<' and $value < $pat->{value}) or ($op eq '>=' and $value >= $pat->{value}) or ($op eq '<=' and $value <= $pat->{value}); } } else { my $attr = $pat->{attr}; push(@results, $get_attr_value->($self, $attr)) if grep { $_ eq $attr } $get_attr_names->($self); } } elsif (defined $pat->{child}) { croak("Can't process child pattern without name") unless defined $pat->{name}; # default op is 'eq' for string matching my $op = $pat->{op} || 'eq'; # do attribute matching foreach my $kid (@kids) { foreach ( $pat->{child} eq "." ? $kid : grep {$get_name->($_) eq $pat->{child}} $get_children->($kid) ) { my $value; foreach_node { my $txt = $get_content->($_); $value .= $txt if defined $txt; } $_, $get_children; next unless defined $value; push(@results, $kid) if ($op eq 'eq' and $value eq $pat->{value}) or ($op eq '=' and $value == $pat->{value}) or ($op eq '!=' and $value != $pat->{value}) or ($op eq '>' and $value > $pat->{value}) or ($op eq '<' and $value < $pat->{value}) or ($op eq '>=' and $value >= $pat->{value}) or ($op eq '<=' and $value <= $pat->{value}); } } } else { push @results, @kids; } # all done? return @results unless @patterns; # apply remaining patterns on matching kids return map { $pkg->_do_match($_, $args, @patterns) } @results; } sub xpath { my ($pkg, $self, $args) = @_; my ($get_parent, $get_children, $get_name) = @{$args}{qw(get_parent get_children get_name)}; my $parent = $get_parent->($self); return '/' unless defined $parent; # root's xpath is / # get order within same-named nodes in the parent my $name = $get_name->($self); my $count = 0; for my $kid ($get_children->($parent)) { last if $kid == $self; $count++ if $get_name->($kid) eq $name; } # construct xpath using parent's xpath and our name and count return $pkg->xpath($parent, $args) . ($get_parent->($parent) ? '/' : '') . $name . '[' . $count . ']'; } # does a depth first traversal in a stack sub foreach_node (&@) { my ($code, $node, $get_children) = @_; my @stack = ($node); while (@stack) { local $_ = shift(@stack); $code->(); push(@stack, $get_children->($_)); } } 1; __END__ =head1 NAME Class::XPath - adds xpath matching to object trees =head1 SYNOPSIS In your node class, use Class::XPath: # generate xpath() and match() using Class::XPath use Class::XPath get_name => 'name', # get the node name with the 'name' method get_parent => 'parent', # get parent with the 'parent' method get_root => \&get_root, # call get_root($node) to get the root get_children => 'kids', # get children with the 'kids' method get_attr_names => 'param', # get names and values of attributes get_attr_value => 'param', # from param get_content => 'data', # get content from the 'data' method ; Now your objects support XPath-esque matching: # find all pages, anywhere in the tree @nodes = $node->match('//page'); # returns an XPath like "/page[1]/paragraph[2]" $xpath = $node->xpath(); =head1 DESCRIPTION This module adds XPath-style matching to your object trees. This means that you can find nodes using an XPath-esque query with C from anywhere in the tree. Also, the C method returns a unique path to a given node which can be used as an identifier. To use this module you must already have an OO implementation of a tree. The tree must be a true tree - all nodes have a single parent and the tree must have a single root node. Also, the order of children within a node must be stable. B This module is not yet a complete XPath implementation. Over time I expect the subset of XPath supported to grow. See the SYNTAX documentation for details on the current level of support. =head1 USAGE This module is used by providing it with information about how your class works. Class::XPath uses this information to build the C and C methods for your class. The parameters passed to 'use Class::XPath' may be set with strings, indicating method names, or subroutine references. They are: =over =item get_name (required) Returns the name of this node. This will be used as the element name when evaluating an XPath match. The value returned must matches /^[\w:]+$/. =item get_parent (required) Returns the parent of this node. The root node must return undef from the get_parent method. =item get_children (required) Returns a list of child nodes, in order. =item get_attr_names (required) Returns a list of available attribute names. The values returned must match /^[\w:]+$/). =item get_attr_value (required) Called with a single parameter, the name of the attribute. Returns the value associated with that attribute. The value returned must be C if no value exists for the attribute. =item get_content (required) Returns the contents of the node. In XML this is text between start and end tags. =item get_root (required) Returns the root node of this tree. =item call_match (optional) Set this to the name of the C method to generate. Defaults to 'match'. =item call_xpath (optional) Set this to the name of the C method to generate. Defaults to 'xpath'. =back =head2 ALTERNATE USAGE If you're using someone else's OO tree module, and you don't want to subclass it, you can still use Class::XPath to add XPath matching to it. This is done by calling Cadd_methods()> with all the options usually passed to C and one extra one, C. For example, to add xpath() and match() to HTML::Element (the node class for HTML::TreeBuilder): # add Class::XPath routines to HTML::Element Class::XPath->add_methods(target => 'HTML::Element', get_parent => 'parent', get_name => 'tag', get_attr_names => sub { my %attr = shift->all_external_attr; return keys %attr; }, get_attr_value => sub { my %attr = shift->all_external_attr; return $attr{$_[0]}; }, get_children => sub { grep { ref $_ } shift->content_list }, get_content => sub { grep { not ref $_ } shift->content_list }, get_root => sub { local $_=shift; while($_->parent) { $_ = $_->parent } return $_; }); Now you can load up an HTML file and do XPath matching on it: my $root = HTML::TreeBuilder->new; $root->parse_file("foo.html");1 # get a list of all paragraphs my @paragraphs = $root->match('//p'); # get the title element my ($title) = $root->match('/head/title'); =head1 GENERATED METHODS This module generates two public methods for your class: =over =item C<< @results = $node->match('/xpath/expression') >> This method performs an XPath match against the tree to which this node belongs. See the SYNTAX documentation for the range of supported expressions. The return value is either a list of node objects, a list of values (when retrieving specific attributes) or an empty list if no matches could be found. If your XPath expression cannot be parsed then the method will die. You can change the name of this method with the 'call_match' option described above. =item C<< $xpath = $node->xpath() >> Get an xpath to uniquely identify this node. Can be used with match() to find the element later. The xpath returned is guaranteed to be unqiue within the element tree. For example, the third node named "paragraph" inside node named "page" has the xpath "/page[1]/paragraph[2]". You can change the name of this method with the 'call_xpath' option described above. =back =head1 SYNTAX This module supports a small subset of XPath at the moment. Here is a list of the type of expressions it knows about: =over =item . Selects and returns the current node. =item name =item ./name Selects a list of nodes called 'name' in the tree below the current node. =item /name Selects a list of nodes called 'name' directly below the root of the tree. =item //name Selects all nodes with a matching name, anywhere in the tree. =item parent/child/grandchild Selects a list of grandchildren for all children of all parents. =item parent[1]/child[2] Selects a single child by indexing into the children lists. =item parent[-1]/child[0] Selects the first child of the last parent. In the real XPath they spell this 'parent[last()]/child[0]' but supporting the Perl syntax is practically free here. Eventually I'll support the XPath style too. =item ../child[2] Selects the second child from the parent of the current node. Currently .. only works at the start of an XPath, mostly because I can't imagine using it anywhere else. =item child[@id=10] Selects the child node with an 'id' attribute of 10. =item child[@id>10] Selects all the child nodes with an 'id' attribute greater than 10. Other supported operators are '<', '<=', '>=' and '!='. =item child[@category="sports"] Selects the child with an 'category' attribute of "sports". The value must be a quoted string (single or double) and no escaping is allowed. =item child[title="Hello World"] Selects the child with a 'title' child element whose content is "Hello World". The value must be a quoted string (single or double) and no escaping is allowed. e.g. Hello World =item //title[.="Hello World"] Selects all 'title' elements whose content is "Hello World". =item child/@attr Returns the list of values for all attributes "attr" within each child. =item //@attr Returns the list of values for all attributes "attr" within each node. =back B this module has no support for Unicode. If this is a problem for you please consider sending me a patch. I'm certain that I don't know enough about Unicode to do it right myself. =head1 BUGS I know of no bugs in this module. If you find one, please file a bug report at: http://rt.cpan.org Alternately you can email me directly at sam@tregar.com. Please include the version of the module and a complete test case that demonstrates the bug. =head1 TODO Planned future work: =over =item * Support more of XPath! =item * Do more to detect broken get_* functions. Maybe use Carp::Assert and a special mode for use during development? =back =head1 ACKNOWLEDGMENTS I would like to thank the creators of XPath for their fine work and the W3C for supporting them in their efforts. The following people have sent me patches and/or suggestions: Tim Peoples Mark Addison Timothy Appnel =head1 COPYRIGHT AND LICENSE Copyright (C) 2002 Sam Tregar This program is free software; you can redistribute it and/or modify it under the same terms as Perl 5 itself. =head1 AUTHOR Sam Tregar =head1 SEE ALSO The XPath W3C Recommendation: http://www.w3.org/TR/xpath =cut