updated SAX API

README.md

@@ -22,7 +22,7 @@ A fast XML parser and Object marshaller as a Ruby gem.
 
 ### Release 1.3.2
 
- - SAX parser bug fix that missed element attributes, 7.5 times faster than Nokogiri and 2.6 times faster than LibXML
+ - Changed SAX parser API for element and instruction attributes
 
 ## <a name="description">Description</a>
 
@@ -56,8 +56,8 @@ files may result in slightly different times.
 As an Object serializer Ox is up to 6 times faster than the standard Ruby
 Marshal.dump() and up to 3 times faster than Marshal.load().
 
-The SAX like stream parser is over 7 times faster than Nokogiri and more than
-2.5 times faster than LibXML when using a trivial Ruby side set of
+The SAX like stream parser is 40 times faster than Nokogiri and more than 13
+times faster than LibXML when validating a file with minimal Ruby
 callbacks. Unlike Nokogiri and LibXML, Ox can be tuned to use only the SAX
 callbacks that are of interest to the caller. (See the perf_sax.rb file for an
 example.)

ext/ox/ox.c

@@ -46,6 +46,7 @@ void Init_ox();
 VALUE    Ox = Qnil;
 
 ID      at_id;
+ID      attr_id;
 ID      attributes_id;
 ID      beg_id;
 ID      cdata_id;
@@ -616,6 +617,7 @@ void Init_ox() {
 
     rb_require("time");
     at_id = rb_intern("at");
+    attr_id = rb_intern("attr");
     attributes_id = rb_intern("@attributes");
     beg_id = rb_intern("@beg");
     cdata_id = rb_intern("cdata");

ext/ox/ox.h

@@ -206,6 +206,7 @@ extern void     write_obj_to_file(VALUE obj, const char *path, Options copts);
 extern VALUE    Ox;
 
 extern ID       at_id;
+extern ID       attr_id;
 extern ID       attributes_id;
 extern ID       beg_id;
 extern ID       cdata_id;

ext/ox/sax.c

@@ -56,6 +56,7 @@ typedef struct _SaxDrive {
         VALUE   io;
     };
     int         has_instruct;
+    int         has_attr;
     int         has_doctype;
     int         has_comment;
     int         has_cdata;
@@ -80,7 +81,7 @@ static int      read_cdata(SaxDrive dr);
 static int      read_comment(SaxDrive dr);
 static int      read_element(SaxDrive dr);
 static int      read_text(SaxDrive dr);
-static int      read_attrs(SaxDrive dr, VALUE *attrs, char c, char termc, char term2, int gather);
+static int      read_attrs(SaxDrive dr, char c, char termc, char term2, int is_xml);
 static char     read_name_token(SaxDrive dr);
 static int      read_quoted_value(SaxDrive dr);
 
@@ -162,6 +163,7 @@ ox_sax_parse(VALUE handler, VALUE io) {
 #if 0
     printf("*** sax_parse with these flags\n");
     printf("    has_instruct = %s\n", dr.has_instruct ? "true" : "false");
+    printf("    has_attr = %s\n", dr.has_attr ? "true" : "false");
     printf("    has_doctype = %s\n", dr.has_doctype ? "true" : "false");
     printf("    has_comment = %s\n", dr.has_comment ? "true" : "false");
     printf("    has_cdata = %s\n", dr.has_cdata ? "true" : "false");
@@ -199,6 +201,7 @@ sax_drive_init(SaxDrive dr, VALUE handler, VALUE io) {
     dr->col = 0;
     dr->handler = handler;
     dr->has_instruct = rb_respond_to(handler, instruct_id);
+    dr->has_attr = rb_respond_to(handler, attr_id);
     dr->has_doctype = rb_respond_to(handler, doctype_id);
     dr->has_comment = rb_respond_to(handler, comment_id);
     dr->has_cdata = rb_respond_to(handler, cdata_id);
@@ -373,31 +376,25 @@ read_children(SaxDrive dr, int first) {
  */
 static int
 read_instruction(SaxDrive dr) {
-    VALUE       target = Qnil;
-    VALUE       attrs = Qnil;
     char        c;
 
     if ('\0' == (c = read_name_token(dr))) {
         return -1;
     }
     if (dr->has_instruct) {
-        target = rb_str_new2(dr->str);
+        VALUE       args[1];
+
+        args[0] = rb_str_new2(dr->str);
+        rb_funcall2(dr->handler, instruct_id, 1, args);
     }
-    if (0 != read_attrs(dr, &attrs, c, '?', '?', dr->has_instruct)) {
+    if (0 != read_attrs(dr, c, '?', '?', (0 == strcmp("xml", dr->str)))) {
         return -1;
     }
     c = next_non_white(dr);
     if ('>' != c) {
         sax_drive_error(dr, "invalid format, instruction not terminated", 1);
         return -1;
     }
-    if (dr->has_instruct) {
-        VALUE       args[2];
-
-        args[0] = target;
-        args[1] = attrs;
-        rb_funcall2(dr->handler, instruct_id, 2, args);
-    }
     dr->str = 0;
 
     return 0;
@@ -519,20 +516,25 @@ read_comment(SaxDrive dr) {
 static int
 read_element(SaxDrive dr) {
     VALUE       name = Qnil;
-    VALUE       attrs = Qnil;
     char        c;
     int         closed;
 
     if ('\0' == (c = read_name_token(dr))) {
         return -1;
     }
     name = str2sym(dr->str);
+    if (dr->has_start_element) {
+        VALUE       args[1];
+
+        args[0] = name;
+        rb_funcall2(dr->handler, start_element_id, 1, args);
+    }
     if ('/' == c) {
         closed = 1;
     } else if ('>' == c) {
         closed = 0;
     } else {
-        if (0 != read_attrs(dr, &attrs, c, '/', '>', dr->has_start_element)) {
+        if (0 != read_attrs(dr, c, '/', '>', 0)) {
             return -1;
         }
         closed = ('/' == *(dr->cur - 1));
@@ -544,17 +546,14 @@ read_element(SaxDrive dr) {
             return -1;
         }
     }
-    if (dr->has_start_element) {
-        VALUE       args[2];
+    if (closed) {
+        if (dr->has_end_element) {
+            VALUE       args[1];
 
-        args[0] = name;
-        args[1] = attrs;
-        rb_funcall2(dr->handler, start_element_id, 2, args);
-        if (closed && dr->has_end_element) {
+            args[0] = name;
             rb_funcall2(dr->handler, end_element_id, 1, args);
         }
-    }
-    if (!closed) {
+    } else {
         if (0 != read_children(dr, 0)) {
             return -1;
         }
@@ -601,7 +600,7 @@ read_text(SaxDrive dr) {
 }
 
 static int
-read_attrs(SaxDrive dr, VALUE *attrs, char c, char termc, char term2, int gather) {
+read_attrs(SaxDrive dr, char c, char termc, char term2, int is_xml) {
     VALUE       name = Qnil;
     int         is_encoding = 0;
 
@@ -618,10 +617,10 @@ read_attrs(SaxDrive dr, VALUE *attrs, char c, char termc, char term2, int gather
         if ('\0' == (c = read_name_token(dr))) {
             return -1;
         }
-        if ('?' == termc && 0 == strcmp("encoding", dr->str)) {
+        if (is_xml && 0 == strcmp("encoding", dr->str)) {
             is_encoding = 1;
         }
-        if (gather) {
+        if (dr->has_attr) {
             name = str2sym(dr->str);
         }
         if (is_white(c)) {
@@ -639,18 +638,17 @@ read_attrs(SaxDrive dr, VALUE *attrs, char c, char termc, char term2, int gather
             dr->encoding = rb_enc_find(dr->str);
         }
 #endif
-        if (gather) {
-            VALUE       rstr = rb_str_new2(dr->str);
-
-            if (Qnil == *attrs) {
-                *attrs = rb_hash_new();
-            }
+        if (dr->has_attr) {
+            VALUE       args[2];
+
+            args[0] = name;
+            args[1] = rb_str_new2(dr->str);
 #ifdef HAVE_RUBY_ENCODING_H
             if (0 != dr->encoding) {
-                rb_enc_associate(rstr, dr->encoding);
+                rb_enc_associate(args[1], dr->encoding);
             }
 #endif
-            rb_hash_aset(*attrs, name, rstr);
+            rb_funcall2(dr->handler, attr_id, 2, args);
         }
         c = next_non_white(dr);
     }

lib/ox/sax.rb

@@ -26,12 +26,13 @@ module Ox
   # be made public in the subclasses. If the methods remain private they will
   # not be called during parsing.
   #
-  #    def instruct(target, attrs); end
+  #    def instruct(target); end
+  #    def attr(name, value); end
   #    def doctype(value); end
   #    def comment(value); end
   #    def cdata(value); end
   #    def text(value); end
-  #    def start_element(name, attrs); end
+  #    def start_element(name); end
   #    def end_element(name); end
   #
   class Sax
@@ -44,7 +45,10 @@ def initialize()
     # they will not be called during parsing.
     private
 
-    def instruct(target, attrs)
+    def instruct(target)
+    end
+
+    def attr(name, value)
     end
 
     def doctype(value)
@@ -59,7 +63,7 @@ def cdata(value)
     def text(value)
     end
 
-    def start_element(name, attrs)
+    def start_element(name)
     end
 
     def end_element(name)

test/perf_sax.rb

@@ -28,7 +28,6 @@
 $verbose = 0
 $ox_only = false
 $all_cbs = false
-$type_cbs = false
 $filename = nil # nil indicates new file names perf.xml will be created and used
 $filesize = 1000 # KBytes
 $iter = 100
@@ -37,7 +36,6 @@
 opts.on("-v", "increase verbosity")                            { $verbose += 1 }
 opts.on("-x", "ox only")                                       { $ox_only = true }
 opts.on("-a", "all callbacks")                                 { $all_cbs = true }
-opts.on("-t", "typical callbacks")                             { $typ_cbs = true }
 opts.on("-f", "--file [String]", String, "filename")           { |f| $filename = f }
 opts.on("-i", "--iterations [Int]", Integer, "iterations")     { |i| $iter = i }
 opts.on("-s", "--size [Int]", Integer, "file size in KBytes")  { |s| $filesize = s }
@@ -79,31 +77,27 @@ def create_file(filename, size)
 end
 
 class OxSax < ::Ox::Sax
-  def start_element(name, attrs); end
   def error(message, line, column); puts message; end
 end
 
-class OxTypSax < OxSax
-  def end_element(name); end
-  def text(value); end
-end
-
 class OxAllSax < OxSax
+  def start_element(name); end
+  def attr(name, value); end
   def end_element(name); end
-  def instruct(target, attrs); end
+  def text(value); end
+  def instruct(target); end
   def doctype(value); end
   def comment(value); end
   def cdata(value); end
-  def text(value); end
 end
 
 unless defined?(::Nokogiri).nil?
   class NoSax < Nokogiri::XML::SAX::Document
-    def start_element(name, attrs = []); end
     def error(message); puts message; end
     def warning(message); puts message; end
   end
   class NoAllSax < NoSax
+    def start_element(name, attrs = []); end
     def characters(text); end
     def cdata_block(string); end
     def comment(string); end
@@ -117,9 +111,9 @@ def xmldecl(version, encoding, standalone); end
 unless defined?(::LibXML).nil?
   class LxSax
     include LibXML::XML::SaxParser::Callbacks
-    def on_start_element(element, attributes); end
   end
   class LxAllSax < LxSax
+    def on_start_element(element, attributes); end
     def on_cdata_block(cdata); end
     def on_characters(chars); end
     def on_comment(msg); end
@@ -141,7 +135,7 @@ def on_start_element_ns(name, attributes, prefix, uri, namespaces); end
 
 def perf_stringio()
   start = Time.now
-  handler = $all_cbs ? OxAllSax.new() : ($typ_cbs ? OxTypSax.new() : OxSax.new())
+  handler = $all_cbs ? OxAllSax.new() : OxSax.new()
   $iter.times do
     input = StringIO.new($xml_str)
     Ox.sax_parse(handler, input)
@@ -188,7 +182,7 @@ def perf_fileio()
   puts "A #{$filesize} KByte XML file was parsed #{$iter} times for this test."
   puts "\n"
   start = Time.now
-  handler = $all_cbs ? OxAllSax.new() : ($typ_cbs ? OxTypSax.new() : OxSax.new())
+  handler = $all_cbs ? OxAllSax.new() : OxSax.new()
   $iter.times do
     input = IO.open(IO.sysopen($filename))
     Ox.sax_parse(handler, input)