diff options
author | Jari Vetoniemi <jari.vetoniemi@indooratlas.com> | 2020-03-16 18:49:26 +0900 |
---|---|---|
committer | Jari Vetoniemi <jari.vetoniemi@indooratlas.com> | 2020-03-30 00:39:06 +0900 |
commit | fcbf63e62c627deae76c1b8cb8c0876c536ed811 (patch) | |
tree | 64cb17de3f41a2b6fef2368028fbd00349946994 /jni/ruby/lib/rexml/xpath.rb |
Fresh start
Diffstat (limited to 'jni/ruby/lib/rexml/xpath.rb')
-rw-r--r-- | jni/ruby/lib/rexml/xpath.rb | 80 |
1 files changed, 80 insertions, 0 deletions
diff --git a/jni/ruby/lib/rexml/xpath.rb b/jni/ruby/lib/rexml/xpath.rb new file mode 100644 index 0000000..0f99808 --- /dev/null +++ b/jni/ruby/lib/rexml/xpath.rb @@ -0,0 +1,80 @@ +require 'rexml/functions' +require 'rexml/xpath_parser' + +module REXML + # Wrapper class. Use this class to access the XPath functions. + class XPath + include Functions + # A base Hash object, supposing to be used when initializing a + # default empty namespaces set, but is currently unused. + # TODO: either set the namespaces=EMPTY_HASH, or deprecate this. + EMPTY_HASH = {} + + # Finds and returns the first node that matches the supplied xpath. + # element:: + # The context element + # path:: + # The xpath to search for. If not supplied or nil, returns the first + # node matching '*'. + # namespaces:: + # If supplied, a Hash which defines a namespace mapping. + # variables:: + # If supplied, a Hash which maps $variables in the query + # to values. This can be used to avoid XPath injection attacks + # or to automatically handle escaping string values. + # + # XPath.first( node ) + # XPath.first( doc, "//b"} ) + # XPath.first( node, "a/x:b", { "x"=>"http://doofus" } ) + # XPath.first( node, '/book/publisher/text()=$publisher', {}, {"publisher"=>"O'Reilly"}) + def XPath::first element, path=nil, namespaces=nil, variables={} + raise "The namespaces argument, if supplied, must be a hash object." unless namespaces.nil? or namespaces.kind_of?(Hash) + raise "The variables argument, if supplied, must be a hash object." unless variables.kind_of?(Hash) + parser = XPathParser.new + parser.namespaces = namespaces + parser.variables = variables + path = "*" unless path + element = [element] unless element.kind_of? Array + parser.parse(path, element).flatten[0] + end + + # Iterates over nodes that match the given path, calling the supplied + # block with the match. + # element:: + # The context element + # path:: + # The xpath to search for. If not supplied or nil, defaults to '*' + # namespaces:: + # If supplied, a Hash which defines a namespace mapping + # variables:: + # If supplied, a Hash which maps $variables in the query + # to values. This can be used to avoid XPath injection attacks + # or to automatically handle escaping string values. + # + # XPath.each( node ) { |el| ... } + # XPath.each( node, '/*[@attr='v']' ) { |el| ... } + # XPath.each( node, 'ancestor::x' ) { |el| ... } + # XPath.each( node, '/book/publisher/text()=$publisher', {}, {"publisher"=>"O'Reilly"}) \ + # {|el| ... } + def XPath::each element, path=nil, namespaces=nil, variables={}, &block + raise "The namespaces argument, if supplied, must be a hash object." unless namespaces.nil? or namespaces.kind_of?(Hash) + raise "The variables argument, if supplied, must be a hash object." unless variables.kind_of?(Hash) + parser = XPathParser.new + parser.namespaces = namespaces + parser.variables = variables + path = "*" unless path + element = [element] unless element.kind_of? Array + parser.parse(path, element).each( &block ) + end + + # Returns an array of nodes matching a given XPath. + def XPath::match element, path=nil, namespaces=nil, variables={} + parser = XPathParser.new + parser.namespaces = namespaces + parser.variables = variables + path = "*" unless path + element = [element] unless element.kind_of? Array + parser.parse(path,element) + end + end +end |