intro to ML

[Henry Cejtin]

I was trying to decide if I should just stick to the core language.  I gather
you think that that is wrong?  I guess either way you have to  say  something
about  the  module  language so that they can get to the standard basis.  The
problem is that then they really have to be able to produce types instead  of
just observing them.

Unfortunately  I  see  I  really am not going to have enough time to get very
far.  The whole notion of polymorphic types, since they have never  seen  any
thing  like  that,  is probably a session in and of itself, as is the idea of
using types to drive things, as is the whole aspect of variables not varying.

Still,  I have hooks to get them interested.  I had to write some simple hack
reading  through  a  largish  text  file  (30  meg)  and  extracting   fields
(`|'-separated).   I used the Substring structure a lot.  The result was most
excellent.  Very clean, very simple and  very  fast.   I  wrote  versions  in
Python  and using awk (awk has VERY fast I/O but isn't a real full language).
The result was that the sizes were all about the same  (MLton  less  than  5%
bigger than awk with Python being 35% bigger) but the speeds were:

    MLton     5.3 CPU seconds
    awk      15.3 CPU seconds
    Python  277.8 CPU seconds