[MLton-user] Re: [MLton] using MLBs for scripting

Raymond Racine rracine@adelphia.net
Wed, 15 Sep 2004 22:55:55 -0400


On Wed, 2004-09-15 at 17:16, Stephen Weeks wrote:
> > What's the real motivation? Just to take advantage of the byte-code
> > interpreter? 
> 
> The motivation is to be able to do scripting with MLton, and to be
> able to use a single file for a script.

Cannot really address the best hows of scripting support.

But...

I've used SML/NJ to write a script or two at work, as well as Python,
MzScheme, Guile, scsh, Bash, Ant, Groovy and Haskell in a work
environment.

Surprisingly, in my experience, the strong vs dynamic typing issue, was
not an issue.  Most would assume the light weightness of dynamically
typed language is "ideal" for scripts and is an advantage.  To the
contrary the REPL, fast compile, and static error checking of SML/NJ, I
found, enhanced the speed of scripting.  

What I found lacking was libraries, bindings, and bindings, and
libraries.  Lots of them.  Breadth, depth and scope. 

So this dove tails and segues into another potential sweet spot for
mlbs, libraries, and a structured system to accommodate comprehensive
suites of them.

Consider the smlnj-lib issue and the bundle with consideration.  The
smlnj-lib is a bag of hodge-podge code (from a context of the cohesion
of the various files, not disparaging the quality of the
implementations).  

Its needs structure, codification, segmentation and classification. An
analogous hierarchical java package system?  SML basis like nested
structures?  Is is worth codifying the namespace structure of these
nested structures?

How should large comprehensive suites of libraries and components be
organized in SML or MLton with mlbs in particular?  


Ray