Lisp vs. Java article

[Andrew Dalke]

Henrik Ekelund wrote:
>Probably the argument about shorter development time holds for Python
>as well as Lisp. But to be honest the argument about execution speed
>does not hold. Lisp has real compilers, Python is bytecode
>interpreted.

No "probably" about it.

The paper at
  http://www-aig.jpl.nasa.gov/public/home/gat/lisp-study.html
is based off of a paper by Lutz Prechelt, (see [jccpp] at
http://wwwipd.ira.uka.de/~prechelt/Biblio/ ).  Prechelt did
a followup titled "An empirical comparison of C, C++, Java, Perl,
Python, Rexx, and Tcl" and named [jccpprt] on that Biblio page.

He finds the development times for the "scripting" languages
are about half as long as the times for the "contentional"
languages.  Python programmers on average took about 4 hours
of development time, which can be compared to the Lisp dev.
time of 5 hours.  (See Prechelt's paper for why this isn't
a strictly good comparison.)

It's doubly hard to compare the runtimes.  In the Prechelt paper,
figure 1, C takes under 4 seconds to finish, Python is about
25 seconds, and C++ is about 60 seconds.  In the Gat report,
the C and C++ times are merged together.  Prechelt says that
   "If we aggregate the languages into only three groups
    (one with C/C++, one with Java, and one with scripts), we
    find that C/C++ is faster than Java (p = 0:074) and tends
    to be faster than scripts (p = 0:15).  There is no
    significant difference between average Java and Script
    run times."

This is different than what Gat reports, perhaps because he
is basing his work off of Prechelt's original paper or perhaps
because he doesn't toss some of the outliers.

So an hour less development time in Python for a program that
runs a minute slower than the Lisp code - I can live with that :)

                    Andrew
                    dalke at dalkescientific.com