Re-implemented Stream#serialize and Chunk#serialize in C.

diff --git a/ChangeLog b/ChangeLog
index f124ff3d0f1d933b52e62d70cfaf17f6b45e55f8..9d70b36d220d642e4cb46169b5f3d362b2d96cb7 100644 (file)
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,9 +1,11 @@
 --
 --

-$Id: ChangeLog 33 2005-04-30 11:10:16Z tilman $
+$Id: ChangeLog 34 2005-04-30 13:15:19Z tilman $

 ++
 
 2005-04-30 Tilman Sauerbeck (tilman at code-monkey de)
         * lib/eet.rb: Tweaked Stream#serialize for performance
 ++
 
 2005-04-30 Tilman Sauerbeck (tilman at code-monkey de)
         * lib/eet.rb: Tweaked Stream#serialize for performance

+        * lib/eet.rb, ext/ext.c: Re-implemented Stream#serialize and
+          Chunk#serialize in C

 
 2005-04-19 Tilman Sauerbeck (tilman at code-monkey de)
         * Rakefile: Code cleanup
 
 2005-04-19 Tilman Sauerbeck (tilman at code-monkey de)
         * Rakefile: Code cleanup

diff --git a/ext/ext.c b/ext/ext.c
index d85cd56088783be0f2be26c30c44fdfd2a7a7017..31beceada0ffa392d0459e47069873000b2ebf0a 100644 (file)
--- a/ext/ext.c
+++ b/ext/ext.c
@@ -1,5 +1,5 @@
 /*
 /*

- * $Id: ext.c 31 2005-04-12 19:03:50Z tilman $
+ * $Id: ext.c 34 2005-04-30 13:15:19Z tilman $

  *
  * Copyright (c) 2005 Tilman Sauerbeck (tilman at code-monkey de)
  *
  *
  * Copyright (c) 2005 Tilman Sauerbeck (tilman at code-monkey de)
  *


@@ -35,6 +35,15 @@
        if (!*(ef)) \
                rb_raise (rb_eIOError, "closed stream");
 
        if (!*(ef)) \
                rb_raise (rb_eIOError, "closed stream");
 

+#ifdef WORDS_BIGENDIAN
+# define BSWAP32(x) \
+       ((((x) & 0xff000000) >> 24) | (((x) & 0x00ff0000) >> 8) | \
+       (((x) & 0x0000ff00) << 8) | (((x) & 0x000000ff) << 24))
+#else /* !WORDS_BIGENDIAN */
+# define BSWAP16(x) (x)
+# define BSWAP32(x) (x)
+#endif /* WORDS_BIGENDIAN */
+

 static VALUE c_close (VALUE self);
 
 static void
 static VALUE c_close (VALUE self);
 
 static void


@@ -419,6 +428,79 @@ c_write_image (int argc, VALUE *argv, VALUE self)
                return INT2FIX (n);
 }
 
                return INT2FIX (n);
 }
 

+static VALUE
+stream_serialize (VALUE self)
+{
+       VALUE ret;
+       struct RArray *stream;
+       static ID id_serialize;
+       long i;
+
+       ret = rb_str_new2 ("");
+
+       stream = RARRAY (self);
+       if (!stream->len)
+               return ret;
+
+       if (!id_serialize)
+               id_serialize = rb_intern ("serialize");
+
+       for (i = 0; i < stream->len; i++) {
+               VALUE str = rb_funcall (stream->ptr[i], id_serialize, 0, NULL);
+
+               rb_str_append (ret, str);
+       }
+
+       return ret;
+}
+
+static VALUE
+chunk_serialize (VALUE self)
+{
+       VALUE tmp, ret;
+       unsigned int size, buf_len;
+       unsigned char *buf;
+       struct RString *tag, *data;
+       static ID id_tag, id_data;
+
+       if (!id_tag)
+               id_tag = rb_intern ("@tag");
+
+       if (!id_data)
+               id_data = rb_intern ("@data");
+
+       tmp = rb_ivar_get (self, id_tag);
+       StringValue (tmp);
+       tag = RSTRING (tmp);
+
+       tmp = rb_ivar_get (self, id_data);
+       StringValue (tmp);
+       data = RSTRING (tmp);
+
+       buf_len = 9 + tag->len + data->len;
+       ret = rb_str_buf_new (buf_len);
+
+       buf = RSTRING (ret)->ptr;
+       RSTRING (ret)->len = buf_len;
+
+       memcpy (buf, "CHnK", 4);
+       buf += 4;
+
+       size = tag->len + data->len + 1;
+       size = BSWAP32 (size);
+       memcpy (buf, &size, 4);
+       buf += 4;
+
+       memcpy (buf, tag->ptr, tag->len);
+       buf += tag->len;
+
+       *buf++ = 0;
+
+       memcpy (buf, data->ptr, data->len);
+
+       return ret;
+}
+

 void
 Init_eet_ext ()
 {
 void
 Init_eet_ext ()
 {


@@ -437,4 +519,10 @@ Init_eet_ext ()
        rb_define_method (c, "write", c_write, -1);
        rb_define_method (c, "read_image", c_read_image, 1);
        rb_define_method (c, "write_image", c_write_image, -1);
        rb_define_method (c, "write", c_write, -1);
        rb_define_method (c, "read_image", c_read_image, 1);
        rb_define_method (c, "write_image", c_write_image, -1);

+
+       c = rb_define_class_under (m, "Stream", rb_cArray);
+       rb_define_method (c, "serialize", stream_serialize, 0);
+
+       c = rb_define_class_under (m, "Chunk", rb_cObject);
+       rb_define_method (c, "serialize", chunk_serialize, 0);

 }
 }

diff --git a/lib/eet.rb b/lib/eet.rb
index ddac645d5404945b4a11379c30a2a35568aa7075..8a949845063a5baf1ef8e0c75fa3ec7b9741dcfa 100644 (file)
--- a/lib/eet.rb
+++ b/lib/eet.rb
@@ -1,5 +1,5 @@
 #--
 #--

-# $Id: eet.rb 33 2005-04-30 11:10:16Z tilman $
+# $Id: eet.rb 34 2005-04-30 13:15:19Z tilman $

 #
 # Copyright (c) 2005 Tilman Sauerbeck (tilman at code-monkey de)
 #
 #
 # Copyright (c) 2005 Tilman Sauerbeck (tilman at code-monkey de)
 #


@@ -166,10 +166,6 @@ module Eet
                        super(chunk.nil? ? 0 : 1, chunk)
                end
 
                        super(chunk.nil? ? 0 : 1, chunk)
                end
 

-               def serialize
-                       map { |c| c.serialize }.join
-               end
-

                def Stream.deserialize(data)
                        data = data.to_str.dup
                        s = Stream.new
                def Stream.deserialize(data)
                        data = data.to_str.dup
                        s = Stream.new


@@ -194,21 +190,13 @@ module Eet
                        @tag = tag.to_str.dup.freeze
                        @data = data.to_str.dup.freeze
 
                        @tag = tag.to_str.dup.freeze
                        @data = data.to_str.dup.freeze
 

-                       @size = @tag.length + 1 + @data.length
-

                        # libeet uses a signed 32bit integer to store the
                        # chunk size, so make sure we don't overflow it
                        # libeet uses a signed 32bit integer to store the
                        # chunk size, so make sure we don't overflow it

-                       if @size >= (1 << 31)
+                       if (@tag.length + 1 + @data.length) >= (1 << 31)

                                raise(ArgumentError, "tag or data too long")
                        end
                end
 
                                raise(ArgumentError, "tag or data too long")
                        end
                end
 

-               def serialize
-                       buf = "CHnK"
-                       buf << [@size].pack("V")
-                       buf << @tag << "\0" << @data
-               end
-

                def Chunk.deserialize(data)
                        if data.length < 8 || data[0, 4] != "CHnK"
                                raise(ChunkError, "invalid data")
                def Chunk.deserialize(data)
                        if data.length < 8 || data[0, 4] != "CHnK"
                                raise(ChunkError, "invalid data")