[om] Matlab phrasebook?

Richard Fateman fateman at cs.berkeley.edu
Mon Oct 22 16:44:45 CEST 2001

Andrew Solomon wrote:

> Dear Richard,
> Thank you for your cheery email.
> On Sun, Oct 21, 2001 at 06:53:52PM -0700, Richard Fateman wrote:
>>At the risk of pointing out the obvious,
>>if Javamath wishes to communicate with a Matlab session,
>>then the simple and obvious method is to use character
>>strings that look like Matlab commands. 
> This is certainly enough for some purposes and JavaMath 
> allows you to install a compute server which communicates in 
> its native language or OpenMath or both. 
>>That is certainly
>>the way I did it in Macsyma, circa 1979. The return
>>journey is via arrays of floats.
> Although you may not be aware of this, some mathematical objects
> are neither easily nor naturally represented as arrays of floats.

Sarcasm accepted. I agree entirely.  I thought we were talking about
Matlab, not "mathematical objects".  While I have not kept
track of the latest and greatest Matlab, it is my impression
that the vast majority of happy Matlab users think they
are computing with arrays of floats.  They probably refer
to these objects as "numerical data".  While we might
like to think that when Matlab computes roots of a
polynomial, this is a mathematical object.  But polynomials
are arrays of floats.

>>Converting to an openmath XML object in between is
>>pointless, I think, unless maybe array of float
>>is going to benefit from being copied over a few
>>extra times. 
> OpenMath is helpful in the case you're using Matlab to process the
> output of another computer algebra system which also speaks
> OpenMath. This is just one example.

This suggests that you believe that Matlab is "another computer
algebra system".  It is a numerical computing environment.
It has string functions, file handling too. But fundamentally
does numbers.

Yes, there is a link to access Maple, and from what I've seen
the major method of communication is  still character string
commands, and arrays -- not of floats -- but of Maple data.
Perhaps the Maple data is  -- character strings--?

>>Or if someone is paying you to do
>>openmath, regardless of its utility.
> No-one is paying me to "do OpenMath" as you put it.
> Frankly, I find your remark offensive.

Actually, I was thinking it would be logical
that you (or whoever wanted a Matlab CD)
would be offering money to someone to write it. That
would provide a motivation, and leave it to you or
someone who supplies the money, to come up with whatever
rationale is needed.  I actually did not mean to insult
you by suggesting you were doing OpenMath "for the money"!

>>There are, I suspect, entirely other issues that make
>>Matlab semantics unwelcome in the rest of the world, at
>>least if it is still the way I remember it:  the meaning of a
>>variable Z depends upon whether there happens to be a file in some
>>file directory with that name.  Is Z a function or an array?
>>Whose file directory are you talking about?
> No-one is forcing you to use MATLAB but some people apparently find it
> useful. 

In fact I don't use Matlab much, but my university uses it
heavily for instruction in courses introducing numerical
computation to engineering undergraduates, and occasionally
in other areas (making use of various Toolboxes, which
substantially enlarge the scope of Matlab from being a
basic programming language).  Visualization, signal
processing, even artificial intelligence.
I have thought about (and even co-written a proposal with
Alan Edelman at MIT) to
implement some ideas on parallel computing
based on the Matlab language.  This idea in itself
has been worked over a few times; we thought we had some
novel approaches. The funding agency we sent it to apparently
did not care for it, so we did not pursue it.

> Honestly, Richard, you must be the only living computer scientist 
> who doesn't acknowledge the utility of establishing standards to ease the
> path to interoperability and accessibility.

Who was it who said "standards are wonderful --- that's why there are
so many of them!" ?

> If the idea offends you so much
> I suggest you unsubscribe from this list since it is our basic premise.

Actually, setting up small groups of people to establish standards that
do not correspond to clear requirements specifications, and that
are outside any standardization bodies (like ISO or IEEE), and which,
additionally discourage critics, seems to me like a clear route
to irrelevance.
   Standards setting has, not unreasonably, been compared to religious
wars.   And I have participated in a few standards committees 
personally. (IEEE floating point, and early stages of Common Lisp, and
  some C language numerics).


om at openmath.org  -  general discussion on OpenMath
Post public announcements to om-announce at openmath.org
Automatic list maintenance software at majordomo at openmath.org
Mail om-owner at openmath.org for assistance with any problems

More information about the Om mailing list