If you want to do the work, a C interface to DFSORT and/or SYNCSORT on
the model of the IBM-supplied COBOL and PL/I interfaces would be very
useful.
The relative merits of C as a language for implementing sort
algorithms are not really of much interest. Controversy about these
merits can be avoided. Every non-trivial sort is I/O bound. The
efficiency of its specialized I/O operations makes or breaks a sort,
and they are implemented in assembly language by specialists. Novices
and even people with years of experience in other kinds of development
cannot expect to be useful.
Perhaps also worth noting is that in your discussions of
sort-algorithm performance you have underestimated the merits of the
bubble sort for already fully [or almost fully] ordered data and, of
course, for this case only. A bubble sort is optimal, does just N - 1
verificatory comparisons and no exchanges for the N keys in the
multiset
K = {k(1) <= k(2) <= . . . <= k(i) <= . . . <= k(N)}.
John Gilmore, Ashland, MA 01721 - USA
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
