OSNews

That's either genius or the dumbest thing I've ever seen.

Edited 2012-12-20 13:20 UTC

You just gotta get used to it, that's all. It's simpler for the interpreter to handle.

It's a pattern that people can't see the difference. :-) Otherwise, it wouldn't be outstanding.

I understand why you'd do that, but... I'd have the reverse problem. I'd have to constantly be rearranging any REBOL code I wrote because I would have written it expecting precedence (which comes naturally to me because I have a strong background in math.)

No, you can just add parentheses, as I showed.

So your justification that Rebol has a strange behaviour because C has also a strange (different) behaviour??

Both are wrong IMHO: the best behaviour is: associativity as in math for + - * / and for all the other operators Rebol behaviour.

That's an interesting idea, which would indeed fit best with common people's experience. The nice thing about REBOL is that it's a meta language: you have control over how the language is evaluated. REBOL makes it easy to write an expression evaluator that would use the rules you stated. I would indeed implement such an evaluator if I would allow users to enter their own expressions, for example in a calculator application.

I absolutely disagree. There are 2 choices to my mind:

1. Use standard precedences for arithmetic, comparison and logic operators as defined in common mathematical usage.

2. Have everything use standard precedences.

C's implementation is actually flawed - Ritchie himself acknowledges so (http://cm.bell-labs.com/cm/cs/who/dmr/chist.html), so blindly copying that seems the wrong approach.

So you never learned any of the mnemonics for operator precedence? Or indeed had to learn in in school, or where it is still very much needed in real life?

If you start with the assumption that unary operators come first, and comparison and logical operators come at the end, then the old saw of My Dear Aunt Sally (multiplication, division, addition, subtraction) holds.



Or like how people do with real math, or have to do in REBOL to make sure the intent is clear?

I am not impresses with this argument. I have met many die hard proponents of Smalltalk, APL, Lisp, and Forth — and with the exception of Forth — none of them have actually claimed that the precedence in their language was somehow better. Instead they would say something like "well the operator precedence is a bit funky, but that is a consequence of how the language works, you will get used to it"

I will end as a general thought: It's ok to admit the flaws of something you love. It isn't somehow admitting defeat. Nothing that is useable in the real world is perfect. Nor in promoting your thing, do you need to talk down something else.

There are many flaws in REBOL, but this is not one of them. We're fixing many of the other flaws in Red, but we're pleased with the default expression precedence as it is.

As I noted before, REBOL makes it easy to implement your own expression evaluator with different precedence when you want.