evaluation

On Sunday 12 October 2008 01:00:46 pm Stavros Macrakis wrote:

> > 2a) The lisp Table implementation is the same speed when
> > compiled as create_list.  It is interesting that makelist is
> > about 10 times slower. makelist should probably be rewritten
>
> It sounds like that would be a good idea.

  I only tested them on large lists (10^5--10^6), using
  showtime:true . Repeating a test yields times that vary by
  100% or more, but a factor of 10 is pretty clear. I don't
  yet know how to time them more carefully and didn't bother
  making a loop to accumulate times for shorter lists. makelist
  appears to be older than create_list. But I suspect create_list
  would work as a drop in replacement for the majority of the
  cases in share.

     > grep -r makelist  src share | wc
       602    2834   54515
     > grep -r create_list  src share | wc
       17      78    1252


> Could you post a specification of Table?

 http://reference.wolfram.com/mathematica/ref/Table.html

 http://documents.wolfram.com/mathematica/functions/Table


Below is link to the Maxima implementation.
The files are:

 mmacompat_lisp.lisp -- only Table is in here. This code
        does not depend on any other files.

 rtest_mmacompat.mac -- the first block of tests are for Table,
       which consist of documented examples as well as edge
       cases. I need to find a way test that it handles incorrect
       input more or less gracefully.

 mmacompat_mac.mac -- To run the test above, this must be loaded.
       (the Table code in here is unused)


They are at:

  http://johnlapeyre.com/mmacompat/new/

So

  load("mmacompat_lisp.lisp")
  load("mmacompat_mac.mac")
  batch("rtest_mmacompat.mac",test)