On 31 Mar 2004, campbell@xxxxxxxxxxxxxxxxxxxxxxxxxxx wrote: > It's not quite as clear-cut as you make it seem. Olin Shivers stuck > with the convention because there were no inherent benefits _for_those_ > sorting_algorithms_. However, there _are_ inherent benefits for > _searching_ (and for a few sorting algorithms). Olin planned to have versions of sort functions with suffix "3". So, maybe the compromise could be the same for this SRFI-43: to have both, vector-binary-search and vector-binary-search3 > > My other problem with three-way comparators is the fact that there > > is no three-way boolean to represent their results; symbols and > > -1, 0, 1 are equally bad in this respect, more so because of > > lack of precedent (no standard Scheme function uses symbols > > as flags or enumerated options). > > > > P.S. Symbols are actually worse than -1, 0, 1 because with > > the latter, one can use, say, * as a substitute for three-way > > boolean operations; with symbols all operations have to be > > written explicitly. > > I'm sticking with negative, zero, and positive. That convention is > used all over the place, it's efficient, and it's convenient. Agreed. (This was exactly Olin's reply to my (naive) proposal to use 3 symbols for representing comparison results.) Greetings Sven
Attachment:
pgpmhz5pSC257.pgp
Description: PGP signature