APL in its ghetto (?) - Programming Languages

Paul Mansour a écrit :
[color=blue]
> Jim,
>
> Interesting question. Couple of observations:
>
> What about A.B.C.D <-5, where A exists, but not B C and D?[/color]

Autovivification of all of them, non ?

Except of course is the question asked to the system is whether A.B.C.D
/exists/ ;-)

[color=blue]
> It may be that asking the interpreter to create a namespace for you as
> you try to stuff something in non-existing space is equivalent to
> allowing the expression A[1]<-5 to work even if A does not exist.[/color]

A nice thing, indeed, because in that case there is nothing more to
distinguish an *integer(s)-indexed* array and a *string(s)-indexed*
array. In fact, both could in such a case be mixed with the same array
name, the problem being the following one, with a non-existing A :

A[2 3 5]<- 11 22 33

do we have 5=rho A (seems reasonable) or 3=rho A ?

A['Paris' 'Rome']<- 'France' 'Italy'

What is rho A in such a case ?

And what if apply the second operation on the first A rather than on an
empty name ? Interesting questions, and the difficulty is probably not
to solve them, but to solve them with elegance.

The scripting Dictionary object, usable with APL+Win, APL/W, and APX,
already provides a mechanism for creating a collection of name/value pairs
where name is a string and value is either a leteral or numeric.

On Thu, 29 Dec 2005 07:55:51 -0600, Paul Mansour wrote (in article
<1135864551.834839.298880@g47g2000cwa.googlegroups.com>):
[color=blue]
> Jim,
>
> Interesting question. Couple of observations:
>
> What about A.B.C.D <-5, where A exists, but not B C and D?[/color]

A, in that case, would have to be a namespace. But if both A and B were
already defined as namespaces then, as I would like things, C would become a
namespace and D a variable.
[color=blue]
> It may be that asking the interpreter to create a namespace for you as you
> try to stuff something in non-existing space is equivalent to allowing the
> expression A[1]<-5 to work even if A does not exist.[/color]

In my own world, and unfortunately not that of the seeming remainder of the
community of APL implementors (shame on me!), the bracket notation describes
a qualification on an existing object, in this case A.
[color=blue]
> Would certainly be nice if we could create new namespaces without the []NS
> '' though.[/color]

Well, we can create a variable or a function without a system command, so why
not a namespace?

--
James L. Ryan -- TaliesinSoft

On Thu, 29 Dec 2005 08:39:51 -0600, FDA wrote
(in article <43b3f45b$0$3785$79c14f64@nan-newsreader-07.noos.net>):

[in response to my stating the following]
[color=blue][color=green]
>> In Dyalog APL I can do the following
>>
>> a) Create a namespace CC with the appropriate system command or system
>> function
>>
>> b) Assign variables within CC
>>
>> CC.US <- 1
>> CC.CA <- 1
>> CC.AU <- 61
>> CC.BR <- 55
>>
>> c) look up country code for a given country
>>
>> CC.US
>> 1
>> CC.BR
>> 55[/color]
>
> Fine, Jim, but what do you do if you want to index with plain strings,
> without restrictions ? I guess
>
> CC.'United States of America' will not do :-(
>
> We just *cannot* ask people to code their information as we had to to in
> the 60s and 70s, mostly because a good part of the information they want
> to deal with comes directly from the Internet.[/color]

If we wish to index with plain strings we can create a matrix where, say, the
first column contains the indexing strings and the second column contains the
associated value. It is then a simple task to perform an index search on that
first column and use the resulting value of that search to extract the value
in the second column.

If this is to be a commonly performed action it is a trivial matter to write
APL functions that act as surrogates for the missing primitive capability.

One of the joys of APL is the ease with which one can create defined
functions that in effect extend the language beyond the built-in
capabilities.

--
James L. Ryan -- TaliesinSoft

> On Wed, 28 Dec 2005 10:57:24 -0600, Christopher Browne wrote[color=blue]
> (in article <m3k6dpnoqz.fsf@mobile.int.cbbrowne.com>):
>[color=green]
>> In Perl, I can specify an associative array of country telephone
>> prefixes, thus:
>>
>> $CC{"US"} = "1";
>> $CC{"CA"} = "1";
>> $CC{"AU"} = "61";
>> $CC{"BR"} = "55";
>> $CC{"FR"} = "33";
>> $CC{"HK"} = "852";
>> $CC{"IN"} = "91";
>> $CC{"IE"} = "972";
>> $CC{"JP"} = "81";
>> $CC{"UK"} = "44";
>>
>>
>> And later, look up the country code for a given country
>>
>> $x = $CC{$some_country};[/color]
>
> In Dyalog APL I can do the following
>
> a) Create a namespace CC with the appropriate system command or system
> function
>
> b) Assign variables within CC
>
> CC.US <- 1
> CC.CA <- 1
> CC.AU <- 61
> CC.BR <- 55
>
> c) look up country code for a given country
>
> CC.US
> 1
> CC.BR
> 55[/color]

What you're describing seems to require that the country name be
referenced in the text of the program.

How do I evaluate the country code if I just have the string...

C = 'US'

In the various other languages, I can use something like:
$area_code = $CC{$country}
--
(format nil "~S@~S" "cbbrowne" "acm.org")
[url]http://linuxdatabases.info/info/x.html[/url]
Rules of the Evil Overlord #67. "No matter how many shorts we have in
the system, my guards will be instructed to treat every surveillance
camera malfunction as a full-scale emergency."
<http://www.eviloverlord.com/>

On Thu, 29 Dec 2005 08:48:36 -0600, FDA wrote
(in article <43b3f667$0$3791$79c14f64@nan-newsreader-07.noos.net>):
[color=blue]
> Why use the dot when the bracket already exists ?
>
> CC['US' 'CA' 'AU' 'BR'] <- 1 1 61 55
>
> It seems *much* more user-friendly, as the concept is exactly parallel
> to what we already have with regular APL indexing, and so does not need
> to learn a syntax addition. It is just extended with backwards
> compatibility (except that what caused one a "syntax error" does not
> anymore. Remember the "NONCE ERROR" at the beginning of APL\360?[/color]

My reply was in the context of what is currently available in Dyalog APL.

And, speaking of "NONCE ERROR", I'm reminded of the dreaded "IMPLICIT ERROR"
that at one time existed in a number of APLs, emerging when one used a
localized system variable prior to that variable having been the recipient of
an assigned value post localization. We took a different approach with
Burroughs APL\700 and automatically set the value of a localized system
variable to the value it had in the calling environment, thus eliminating the
case where the system variables were without a value. Interestingly this
notion was initially met with strong resistance by the implementors of other
APL systems, but it prevailed in the end.

--
James L. Ryan -- TaliesinSoft

That is exactly right!

a{comes from}90

creates a variable in the DEFAULT namespace #. The statement is exactly
equivalent to:

#.a{comes from}90

When you are attempting to execute a.a{comes from}90, you are NOT using
the default namespace which is implicitly referred to but referring to a
namespace named <a>: if it does not exist, the response should be VALUE
ERROR, just as when you reference a variable/function that does not
exist.

Omission of a named nameshace defaults the scope to the DEFAULT namespace:
if you specify a namespace and it does not exist, what do you expect if not
VALUE ERROR?

What would you expect when saving a document to f:\xx\mydoc.txt if f: does
not exist, if xx does not exist or if neither of them can possibly exist?
With a.a{comes from}90, it is conceivable that <a> cannot POSSIBLY exist
either because there is no memory []wa to create it or because it is
already in use as another object in the active session.

TaliesinSoft a écrit :
[color=blue]
> If we wish to index with plain strings we can create a matrix where, say, the
> first column contains the indexing strings and the second column contains the
> associated value. It is then a simple task to perform an index search on that
> first column and use the resulting value of that search to extract the value
> in the second column.[/color]

Why not, indeed ? We could also ask users to program in assembly
language using appropriate subroutines.

Let us be serious, please. If I have 4000 references to manage, I
certainly do *not* want to search through them sequentially, much less
to keep them sorted at a great expense to apply a clumsy dichotomic
method. *Direct access is what is needed* , no awkward /substitute/ . It
is also, by the way, what database give us once we go from the
prototyping to the real application phase, a thing the Perl conceptor(s)
understood while allowing to "bless" an associative array in order to
associate it with any database.

Last but not least, one database <=> one name if we want our programs to
be readable.

[color=blue]
> One of the joys of APL is the ease with which one can create defined
> functions that in effect extend the language beyond the built-in
> capabilities.[/color]

One of the lesser joys is that when you try to merge different APL
programs that have used different "easy" ways to extend the language,
you realize you have to duplicate the associated (different, of course)
functions, to say nothing about the predictible conflict of names used
to perform similar functions.

No substitute to introducing basic functions of the mind as basic
functions of the language too, period. :-(

TaliesinSoft a écrit :

[color=blue]
> And, speaking of "NONCE ERROR", I'm reminded of the dreaded "IMPLICIT ERROR"[/color]


One of my favorites to pull a prank on young engineers (already knowing
APL, of course) was to write in one of their workspaces when they were
not present :


quadLX <- '''quadIO<-''1'''''


Then watch. It was a *very* good way to test the way they were reasoning :-)