Hi Daniel,

Daniel Carrera wrote:

The real question, however, is whether having one or ten characters in the tag makes any significant difference in the current parsing process, which quite frankly I have no idea about and would need to be measured before it's overly debated, if you want my opinion.


Measuring is always a good idea. I did some measurements (calculated the size savings from using smaller tags and compared them to the speed of an IDE drive) and concluded that the probable savings were negligible.


Well, there is a significant difference in uncompressed size.

Take a look at Tools-Options-Load/Save-General. There is an option 'optimize XML for size'. This option defaults to 'optimize' and iirc it was introduced in an early effort to make the file load process faster. I would even assume that actual profiling led to this idea.

All this does is eliminate sequences of blanks or tabs and newlines which are needed to pretty-print the XML so it is readable for humas in an ordinary text editor, leaving one huge unstructured line of XML. I'd assume that the effect of this on compressed size is limited, but certainly does affect uncompressed size.

This suggests that uncompressed XML size does have a measurable effect.

And that also suggests a more relvant experiment than stopping at byte counts and speculating whether or not they have a significant effect: Take the same (big) document saved with and without pretty-printing, compare their sizes and measuring the load times. Depending on details of the parser used, the same change in uncompressed size due to tag lengths could have a slightly larger effect than pure whitespace, because tags go through somewhat more processing (e.g. token lookups).

Ciao, Joerg

--
Joerg Barfurth              Sun Microsystems - Desktop - Hamburg
>>>>>>>>>>>>>>>>>> using std::disclaimer <<<<<<<<<<<<<<<<<<<<<<<
Software Engineer                         [EMAIL PROTECTED]
OpenOffice.org Configuration          http://util.openoffice.org


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to