[cdt-core-dev] FYI: parser performance work for 2.0.1

[John Camelon <jcamelon@xxxxxxxxxx>]

Andrew, Doug and I have been working
away in CVS HEAD, striving to make the parser framework perform better
and use less memory.  I thought I'd send an email to the list letting
the community know what exactly this entails.  

1.  We plan to reduce (if not eliminate)
our use of java.lang.String within our implementation in order to decrease
our object count and footprint.  
2.  We plan to eliminate all use
of java.util containers in order to reduce footprint, object count, and
to improve performance.  
3.  We are 2/3 of the way through
a re-write of our IScanner implementation that uses less objects and classes,
and behaves more like a typical scanner one would write in C.  

As I mentioned on our last conference
call, we shall do our best to make sure that our interface changes are
backwards compatible where it does not affect performance and scalability.
 Before we merge the HEAD back into the 2_0 branch near the end of
this month, I shall publish an email of interface changes that we feel
we cannot avoid making.  For those who plan to ship products on top
of 2.0.1, I ask for your feedback and patience.  

Until this merge happens, we are not
looking at any noncritical defects in this area. 

Thanks,
JohnC
www.eclipse.org/cdt