[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: associative methods


Regarding my #more suggestion.  I guess you're right -- I was being too
conservative on this.  Our compiler (should we get the go-ahead to jump
into this game) will certainly be able to determine whether a #rest
sequence has dynamic extent or not.  It will then be able to generate
optimal non-consing code in most cases, and to warn the user if
unanticipated consing is going to occur.

I was more worried about people who want to write simple compilers on
itty-bitty computers, and about the anxiety of users if they're not sure
just what the compiler will do.  But maybe all Dylan compilers will be
smart enough to do this right, eliminating any need to tamper with #rest
args.  (Or at least, all Dylan compilers used by people who care AT ALL
about efficiency.)
I agree that we have to get sealing and inlining right to make this all
work well.
So, let's forget about my more-arg suggestion and just make sure that the
various Dylan compilers don't reinforce the myth that rest-args must be
avoided at all costs.

-- Scott
Scott E. Fahlman
School of Computer Science
Carnegie Mellon University
5000 Forbes Avenue
Pittsburgh, PA 15213

Internet: sef+@cs.cmu.edu