[MLton] Welcome to David Kuestler

[Henry Cejtin]

I use MLton a lot at work to process what I think of as very large volumes of
data.  My standard test run consists of an input  file  which  is  about  1.5
gigabytes,  consisting  of  1/2  trillion floating point numbers.  All of the
data has to fit in RAM during the lengthy computation,  but  there  are  many
duplicates.

This  all  works  fine in MLton (by building hash tables of values on input).
At one point the GC has to switch to the mark-and-sweep mode, but I have  had
no  problems  with  it.  (We really are looking forward to MLton on x86-64 so
that we can cross the 4-gig barrier.)

At a previous job  I  used  MLton  to  convert  data  for  insertion  into  a
relational  database.   The  data  there  was  much  smaller,  only  about 50
megabytes, although the resulting tables took  up  about  10  times  that  in
Postgres.

The current stuff is very floating point intensive, while the older stuff was
more character tweaking.  During all of it MLton has performed perfectly.