Space efficient sorting of arrays 
Author Message
 Space efficient sorting of arrays


> (1) Those who share Arthur's once-avowed attitude of too-many-cooks-
>     spoil-the-broth in the extreme.  For such people, every additional
>     developer means a sacrifice in product quality.  I can think of very
>     few core parts of A or A+ that Arthur did not write.  Anecdotally,
>     I was present when Arthur suggested to John Mack, then Morgan Stanley's
>     Fixed Income Division head, that all of FID's applications ought
>     to be developed by no more than 6 programmers.

> (2) Those who would not accept the risk that a lone developer would
>     be hit by a truck (sorry to be morbid), lose interest, or otherwise
>     not be available for continued development and support.  Availability
>     of source code may mitigate this to a degree.

Both very relevant points. You obviously _know_ much more about the actual
history of K (and A+) than I do, which I guess increases `my surprise' at
`your surprise', but since---as you rightly point out---my surprise isn't
a matter of any practical consequence, that point can rest.

And I like surprises, anyway.

As to the substance of your comments, if you represent Arthur's view correctly
(and it does `sound' right to me) I'd wholeheartedly concur with it. IME,
programmers have the same productivity performance as do policemen: you get
1/n work out of a collection of n workers, i.e. 2 people do half the work of one
(they spend a lot of time talking with one another---you know, nose of one
police car sniffing tail of another down in the town park that is _least_ subject
to crime).

And to the `hit by a truck' argument, I am sympathetic to it, but unmoved. Multiple
`developers' is no guarrantee that loss of a major design force could be survived.
Strikes me there is just as much chance that the code could be passed to someone
good that was an outsider as that someone good might rise out of a group
already working on some project.

Wed, 26 May 2004 03:17:48 GMT  
 Whither APL or wither APL ??


>> I did not say that one could not do numerical analysis in APL, or that one
>> could not gain a great deal of understanding by doing so.  I said that it
>> would defeat the pedagogical objectives of a specific course which I took
>> and you didn't.

> Just what were the "pedagogical objectives" that were better served by
> using fortran in a Physics course?  I could see it in a FORTRAN
> programming course but you implied that the object was to learn some
> numerical analysis.

The teaching of specific techniques that are already implemented as
primitives in APL.

>> > Examples:
>> > ...

>> All of this is well and good.  But I'm not sure I see your point.

> My point is that I don not believe APL in any way hides the concepts of
> numerical analysis except in three (well really, one) cases:  Solving
> simultaneous equations, matrix inversion and least squares fitting.

And we were doing all of those.

> I
> could see the point in outlawing the use of {quad-divide} at least until
> the students had written their own version if these were part of the
> curiculum.

Telling a student not to use a particular feature of the language is like
telling a cat not to scratch the couch.

>> > And I don't agree that "it would mean writing really bad code ."  In
>> > fact, it leads to a good understanding of where loops are fundamental
>> > and where they are artifacts of programming languages.

>> If you consider writing Fortan in APL to be good code, be my guest.

> I definitely don't.  Nor do I think it's a good idea forcing the use of
> looping simply because the prescribed language can't handle arrays in a
> reasonable way.

And I don't see any purpose to be served by letting students think that
arrays are processed by magic.

>> Personally I don't see that as a "chore"--if you understand the algorithms
>> then implementing them is not all that difficult in any language.

> So you contend that once you understand the principles of some
> algorithm, it could be coded, tested and, perhaps, modified as quickly
> in FORTRAN as in APL?

Depends on the algorithm.  Some yes, some no.  Really depends on how well
the problem fits the language.

>> I'm sorry, but "where computers were available or assumed"?  Why would
>> they _not_ be "available or assumed?"

> Because the instructor might require the use of pencil an paper or, for
> that matter an abacus.  Numerical analysis was around before PCs.

Yes it was.  "Computer" used to be a job description.  But I don't think
that even the most {*filter*}professor would advocate going back to the era
when calculations were performed by rooms full of ladies in green
eyeshades.  Numerical analysis, as performed in the final quarter of the
20th century by working physicists and engineers almost invariably
involved the use of a computer.

> Ted

Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net)

Wed, 21 Sep 2005 14:30:50 GMT  
 Whither APL or wither APL ??


> And I don't see any purpose to be served by letting students think that
> arrays are processed by magic.

I've worked on a machine where (,C){is}(,A)+(,B) is a single hardware
instruction.  What would you do there?


Thu, 22 Sep 2005 01:33:45 GMT  
 Whither APL or wither APL ??
On Sat, 5 Apr 2003 11:33:45 -0600, Ted Edwards wrote

> I've worked on a machine where (,C){is}(,A)+(,B) is a single hardware
> instruction.  What would you do there?

You worked on a Burroughs B6700?   ;-)

-- James L. Ryan -- TaliesinSoft

Thu, 22 Sep 2005 04:39:24 GMT  
 Whither APL or wither APL ??


> You worked on a Burroughs B6700?   ;-)


Yes, and the B5500/5700

Great machines. And the ALGOL was was also VERY nice.


Thu, 22 Sep 2005 11:28:33 GMT  
 [ 132 post ]  Go to page: [1] [2] [3] [4] [5] [6] [7] [8] [9]

 Relevant Pages 

Powered by phpBB® Forum Software