cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Conny Krappatsch <co...@smb-tec.com>
Subject Re: Fwd: XML Compilation
Date Tue, 17 Oct 2000 13:08:26 GMT
Hi Stefano,

is that what you and Falko where talking about? (He told me!)

Our approach is almost the same. Using some sort of serialization we create
streamlike data out of XML data (DOM or SAX) for transfer between server
and client.

Depending on XML data size this may happen in multiple steps to reduce memory
consumption . The client initializes a new retrieval, asks for a chunk of data,
creates what ever it wants to create (DOM or SAX), asks for the next chunk and
so on (client has to ask, because server can't be the active part). Storage is
similar (still the client is the active one and sends data chunks).

Do you see any possibility to integrate our 'chunk approach' and your 'stream
solution'?

BTW, i would suggest to add the possibility to transfer/store comments (and
maybe some other XML constructs (CDATA ?), though the SAX ContentHandler doesn't
provide information about it - ultimately LexicalHandler will).


Regards,
Conny


Stefano Mazzocchi wrote:
> 
> Hi,
> 
> I have implemented an idea I had in mind for a while: XML bytecode
> compilation.
> 
> In short, an XML file is parsed by a regular parser (and possibly
> validated against a schema), then compiled into a binary form that is
> easier to parse. You can think as the XML equivalent of java bytecode
> compilation.
> 
> Then the document is read by an XML interpreter which behaves exactly as
> an XML SAX parser (so you can plug it into your code with no changes
> whatsoever) just it's much faster since it doesn't have to do any
> special parsing but iterating on compiled SAX events and throwing them.
> 
> The results are very interesting: speed improvement goes from 16000% for
> very small files (100bytes) to 45% for big files (650Kb) over Xerces 1.2
> in non-validating SAX mode.
> 
> XML compilation is quite fast (300 millis for the 33Kb file, 2500 millis
> for the 650Kb file) and doesn't increase the size for much (3% bigger
> for the 650Kb file).
> 
> See the attached files for a complete result. The tests I used as
> attached as well.
> 
> The code is written as a test but it's carefully optimized for speed
> without any particular JVM trick (only algorithmical optimizations such
> as string pooling and faster unicode encoding)
> 
> I'm releasing it under the Cocoon APL. If interesting, I will write a
> description of the CXML file format and release that as well.
> 
> The package is not yet identified, suggestions are welcome.

-- 
______________________________________________________________________
Conny Krappatsch                              mailto:conny@smb-tec.com
SMB GmbH                                        http://www.smb-tec.com




Mime
View raw message