reverse and nreverse in nset performance review

Hi Stavros,

>>>>> "Stavros" == Stavros Macrakis <macrakis at alum.mit.edu> writes:

    Stavros> Andreas,
    Stavros> In reviewing the performance of the set functions, I noticed that there was
    Stavros> a call to "reverse" in do-merge-asym where an "nreverse" would make more
    Stavros> sense.  I wrote this code, so I was embarrassed to see this elementary
    Stavros> oversight.  But then I checked the revision log of nset, and discovered that
    Stavros> it was you who had made this change in nset.lisp 1.20 ->
    Stavros> 1.21<http://maxima.cvs.sourceforge.net/maxima/maxima/src/nset.lisp?r1=1.20&r2=1.21>;(March
    Stavros> 2007). Barton had also pointed this out to me last year, but I wasn't
    Stavros> focussed on nset at the time.

    Stavros> I had used nreverse in do-merge-asym because the list structure here is
    Stavros> supposed to be guaranteed to be unshared.  I assume you changed it to
    Stavros> reverse in the belief that there are cases where it might be shared. If that
    Stavros> is true, then there is a bug in do-merge-asym which should be corrected at
    Stavros> the source, not symptomatically.  If it is not true, the only effect will be
    Stavros> to increase the amount of cons'ing, which is generally considered a bad
    Stavros> thing. WHich is it?

Since it is about 1 1/2 years ago now I am no longer quite sure
what it was, but I think I was concerned about do-merge-asym might
share data.
If you are sure this is not so, you could surely change it back.

Sorry for any inconvenience that micht have given you.

Thanks,

Andreas



-- 
ceterum censeo redmondinem esse delendam.