Support for lexing/parsing XPath type tests.

Unlike what I thought before syntax such as "node()" is not a function call.
Instead this is a special node test that tests the *types* of nodes, not their
names.

lib/oga/xpath/lexer.rl

@@ -249,9 +249,7 @@ module Oga
           # added on lexer level to make it easier to handle these cases on
           # parser/evaluator level.
           if AXIS_EMIT_NODE.include?(value)
-            add_token(:T_IDENT, 'node')
-            add_token(:T_LPAREN)
-            add_token(:T_RPAREN)
+            add_token(:T_NODE_TYPE, 'node')
 
             if AXIS_EMIT_EXTRA_SLASH.include?(value) and te != eof
               add_token(:T_SLASH)
@@ -305,10 +303,25 @@ module Oga
           | op_sub
           ;
 
+        # Node types
+        #
+        # While these look like functions they are actually node tests. For
+        # example, comment() matches all comment nodes.
+        #
+        # See http://www.w3.org/TR/xpath/#NT-NodeType for more information.
+
+        node_type = 'comment' | 'text' | 'processing-instruction' | 'node';
+
+        action emit_node_type {
+          emit(:T_NODE_TYPE, ts, te - 2)
+        }
+
         main := |*
           operator;
           whitespace | slash | lparen | rparen | comma | colon;
 
+          node_type '()' => emit_node_type;
+
           '[' => { add_token(:T_LBRACK) };
           ']' => { add_token(:T_RBRACK) };
 

lib/oga/xpath/parser.y

@@ -3,7 +3,7 @@
 #
 class Oga::XPath::Parser
 
-token T_AXIS T_COLON T_COMMA T_FLOAT T_INT T_IDENT
+token T_AXIS T_COLON T_COMMA T_FLOAT T_INT T_IDENT T_NODE_TYPE
 token T_LBRACK T_RBRACK T_LPAREN T_RPAREN T_SLASH T_STRING
 token T_PIPE T_AND T_OR T_ADD T_DIV T_MOD T_EQ T_NEQ T_LT T_GT T_LTE T_GTE
 token T_SUB T_MUL
@@ -76,6 +76,11 @@ rule
   node_test
     : node_name           { s(:test, *val[0]) }
     | node_name predicate { s(:test, *val[0], val[1]) }
+    | node_type           { val[0] }
+    ;
+
+  node_type
+    : T_NODE_TYPE { s(:node_type, val[0]) }
     ;
 
   node_name