From: Seebs Date: 2009-11-09T22:35:07+09:00 Subject: Re: Ruby doesn't implement x++ for Fixnum's because ??? On 2009-11-09, Tony Arcieri wrote: > [Note: parts of this message were removed to make it a legal post.] > On Sun, Nov 8, 2009 at 9:54 PM, Joel VanderWerf wrote: >> The only sane answer is to do what C does, as far as parsing. > You will find the answer to these edge cases varies. No, you won't. The parsing is 100% consistent and has been, so far as I know, in every compiler released since the mid-70s. > Ruby is a pure expression-based language (and I see you acknowledge that > later in this message). There are no statements which don't return a value. But "expressions" can be extremely large. In C, "x++" yields the current value of x, and at some point between the prior sequence point and the next sequence point, x gets incremented. The reason that "i + i++" is *undefined behavior*, not just an unspecified value, is that Deep Magic may occur to implement or support the increment operators. There's no particular guarantee -- none at all -- that such code even runs, let alone does anything specific. That's fine. C has "sequence points", and because expressions in C are pretty simple and limited, you always have plenty. Consider: for (i = 0; i < 10; ++i) { printf("%d\n", i); } The evaluation of each of the control expressions in the for loop is separated from everything else by sequence points. There's a sequence point between the evaluation of i for the printf and the actual execution of printf, and another when printf is done, and so on. Let's look at a while loop to simplify a bit: i = 0; while (i++ < 10) { printf("%d\n", i); } In C, that should consistently yield the numbers 1 through 10. (When i is 0, i++ is < 10, but increments i, so printf sees 1. When i is 9, i++ is 9, thus still < 10, but increments i, so printf sees 10.) In Ruby, if you did something similar to this, you might write: i = 0 while i++ < 10 puts i end When, exactly, do we expect that the increment happens? This whole thing is an expression. I'm not aware of anything particularly equivalent to sequence points. Do we ensure that i is incremented before puts is called? Consider that this whole block (while ... end) could be a single argument to a function, because it's just an expression. In short, the fact that control statements in Ruby are also expressions makes it harder to run with the simple "by the end of the expression" heuristic most people use. Think about: puts i++ || i In C, you'd have a nice solid guarantee that the increment happens before the || (because &&/|| are sequence points). Would you in Ruby? I have no clue. > Correct, it's ++x that would have the same effect as x += 1, not that I find > the preincrementation vs postincrementation distinction particularly useful > in Ruby, which is why I've advocated picking the canonical one and going > with that, even though preincrementation has semantics which would make more > sense in Ruby. There's nothing particularly more canonical about postincrement. It is the one that's harder to replace by just writing something else. I would rather see both, personally. In particular, it looks very much as though postincrement would be expensive, so I'd want preincrement available so I could actually use it. :) -s -- Copyright 2009, all wrongs reversed. Peter Seebach / usenet-nospam@seebs.net http://www.seebs.net/log/ <-- lawsuits, religion, and funny pictures http://en.wikipedia.org/wiki/Fair_Game_(Scientology) <-- get educated!