RE: coping with huge xml-saxon

[ Paul Tremblay]

> It seem that if you have 256 MB of memory, you ought to be able to
> handle a document of around 50 MB. (I think the processor actually
> requires something like 4 times the size of the document--or 
> was it 10?)
> 
> I see the memory as the only drawback for xslt transformations. After
> all, I have documents that are around 1 MB in size, and I can 
> easily see
> real documents being many times that size
> 

Factors of 10 to 20 seem to be realistic, I think.  A lot depends on the
design of the stylesheet.  if you have performance O(n^2), you will
probably be waiting forever.  If you have a lot of recursion, and data
from the entire document have to be pulled together, you will overflow
any reasonable stack.

Databases are designed to pull data in from disk as they process, so
that it all does not have to be in memory at once.  It is hard to get
this to work well, which is partly why there are only a few big
commercial relational databases out there, I suppose.  So far, we don;t
have a standard xslt processor which can work in this way. Even with
databases like that, you have to design the query very carefully, and
place that data on the disk(s) properly, otherwise you can bring the
system to its knees.

Cheers,

Tom P

 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list