I really would prefer no changes to the definition of the Snowball language
itself at this stage. In Grapeshot I compile Snowball scripts in quite a
different way, and keeping Snowball compatible with this new work is
important to me.
Besides, I did think through the consequences of not having locals (or
parameters in routine calls) and was satified with the design as it stood.
Certainly the design we have has stood the test of time.
I have done several tests in the past comparing the deferencing of variables
in structures as against using static variables (or the Porter stemmer and
so on), and the overhead for the dereferencing has always been either
invisible or quite negligible. I don't think one can use speed improvements
to justify a change here.
This archive was generated by hypermail 2.1.3 : Thu Sep 20 2007 - 12:02:48 BST