World: r3wp

It's highly annoying to have paths bomb out instead of return NONE

We don't have that anymore since a while, both in R2 and R3.

>> b: []
== []
>> b/1
== none
>> b/a
** Script Error: Invalid path value: a
** Near: b/a

Was talking about the first form only (with the index)

but yeah

R3 is more lenient

Hasn't the first form always been like that?

Don't think so, but my memory is not that good :-)

Don't think so, but my memory is not that good :-)

I Remember something like: 
Hey ! You past the end !

At one point in R2, it was something like this:
>> X: [1 2]
== [1 2]
>> X/2
== 2
>> X/3
== none
>> X/4
!Error

(...well, for some reason I remember it that way, at least.  There's 
a chance I have it confused with something else...)

Steeve, you are right, I missed that :(

Gregg, my answer to your question is that writing micro-optimal code 
in the first place is usually not premature optimization; it is instead 
just avoiding having to rewrite as much of the code later. So it's 
still optimization, just not premature. Making macro-optimizations 
before you have the overall design done is usually premature optimization, 
but if your code style is consistent then you can make them later 
without having to rewrite too much, in my experience :)

Re "premature optimization" - I do not think I ever did have any 
feeling I used that. For example, I would never want to write O(n 
*n) code when O(n) is available, no matter how would anybody want 
to call my approach.

I posted a note on Geany as a possible linux rebol tool in IDE as 
Carl's Rebol Blog is no place for running notes - but altme cannot 
tag a topic ?  And trying to select a group here on linux as ALTme 
1.2.25 is loading is just a crap shoot - highlight and click and 
close yore eyes or is it the reverse?

Probably because AltME resorts the groups as you are trying to click 
the one you want, thus getting the wrong one.

We might want to change groups on the opimization topic. How about 
Performance?

Do we have NaN (Not A Number) defined in R3?

I don't think so. Is it needed?

This page suggests that the "internals" of R3 does have NaN defined 
http://www.rebol.net/wiki/Decimals-64but it doesn't appear to be 
accessible.

Related topic, Jerry: INFs (see comment from Carl, which perhaps 
offers some hope that the situation will be normalised one day):
    http://www.curecode.org/rebol3/ticket.rsp?id=1717

This page suggests that the 

internals" of R3 does have NaN defined" - how does the text suggest 
it? I, as a coauthor will try to correct the text, if it does.

Certainly, IEEE754 defines NaN, but Rebol does not need to use/define 
it in any way.

So, for Rebol decimals, NaN is not defined/supported at present, 
nor are there any plans to use it in the future.

Thanks for the answer, Ladislave.

(but, of course, if such a wish exists, there is always the possibility 
to put it as a wish to CureCode)

yes, put it in curecode. if it's rejected, at least there would be 
documentation of why it's rejected.

Do we know a reason why it is desirable to have decimal! NaNs?

Ladislav - The following sentence implies that there is an internal 
NaN:
 

The exponent value 2047 is reserved for overflow and NaN (Not a Number)

I wrote that, I think. I got it from a IEEE 754 definition, like
http://en.wikipedia.org/wiki/IEEE_754-1985

Think of that part of the text as a description of the floating point 
standard used by CPUs.

Maybe you can try
to-decimal #{ ... 16 digits ... }

If that still works, you can use it to test, what you get, when you 
construct numbers like NaNs. To see how REBOL3 handle it.

Just checked. That sentence is in the section named "IEEE754 standard", 
so it is OK.

The IEEE754 standard reserves such a value for NaNs, overflow, etc. 
But, that does not mean, Rebol has to implement those.

BTW, "reserved for X" does not mean "X is implemented", it just means, 
that it should not be used for other purposes.

Nevertheless, I would like to see some coherent reasons why to support 
NaNs in Rebol.

As far as I am concerned, I guess, that Carl felt that:

>> 1 + (square-root -1) / 2
** Math error: positive number required
** Where: square-root
** Near: square-root -1

is more comfortable, than

>> 1 + (square-root -1) / 2
== #[NaN]

could be

The primary reason for supporting NaNs would be for easy of interaction 
with systems that do support NaN, eg Oracle.


Right now, any REBOL system that was trying to trade values with 
an Oracle system that supported NaN and +/-INF would need to code 
for special cases.


However, I do not know of anyone who has such a need -- so time for 
some to make the busines case!

Business case for NaNs???? Zero probability.

It depends on the business :)

I am sure, that my estimate is accurate

perhaps it's better to look at why other languages implement NaN.

Any particular language you do want to imitate?

If it's part of a standard, then I guess not. I'm guessing that languages 
like javascript implement it to make error handling of bad numbers 
easier in ways that may not be necessary for REBOL.

IIRC, the NaN stuff is mostly necessary in assembler and on the hardware 
level to trigger an exception and somehow report back a problem. 
If any layer now handles this exception it's not necessary to further 
bubble it upwards to interpreters, user scripts etc.

IMO In JS the NaN stuff is just annoying 'feature' which makes debugging 
harder.  Would you like to deal with the NaN or better just get an 
immediate error?

Many languages are implemented using C. If you don't do anything 
particular regarding NaNs, you get outputs like from this C program:

#include <math.h>
#include <stdio.h>

int main()
{
	printf ("%lf\n", sqrt (1.0));
	printf ("%lf\n", sqrt (-1.0));
	printf ("%lf\n", sqrt (2.0));
}

Output:

1.000000
nan
1.414214


So it may actually take more effort to grab the NaN output and make 
e.g. an error output. Like REBOL does:

>> square-root -1
** Math Error: Positive number required
** Near: square-root -1

(By REBOL example is by using R2. Not sure how R3 does this.)

*My*

I've never felt a need for NaN, but I also haven't done symbolic 
stuff or things where I think it would help to have it as a missing 
value. I'm open to it if there's a valid case though.