[antlr-interest] TreeNodeStream implementation

Kirk Woods woods.kirk at gmail.com
Fri Dec 22 22:19:08 PST 2006


I vote for faster and simpler. Memory is cheap and continues to get
cheaper.  If someone is on an older machine lacking memory, they can use the
UnbufferedTreeNodeStream.

Kirk

On 12/22/06, Terence Parr <parrt at cs.usfca.edu> wrote:
>
> Heh, I've been building a node stream that is much faster and simpler
> than current CommonTreeNodeStream, but requires that I make an array
> of node pointers that point into the tree.  It will be much more
> suitable to interpreters etc... because it can jump around.  So it's
> faster and simpler, but costs memory for that array.
>
> Q: Shall I make this the CommonTreeNodeStream and then make current
> one another name like UnbufferedTreeNodeStream?
>
> Later I will even make an IndexedTreeNodeStream for really fast rule-
> based translation...where are my grad students when I need them!  Oh,
> I don't have any...that's right.
>
> I know most people in western world are on vacation, but anybody from
> Asia or wherever wanna comment?  Should we buffer the tree node
> stream just like we do for the token streams by default?
>
> Ter
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.antlr.org/pipermail/antlr-interest/attachments/20061223/a3f797c9/attachment.html 


More information about the antlr-interest mailing list