Path: csiph.com!weretis.net!feeder8.news.weretis.net!eternal-september.org!reader01.eternal-september.org!.POSTED!not-for-mail From: Keith Thompson Newsgroups: comp.lang.c++,comp.lang.c Subject: Re: lvalue assignment difference between C and C++ Date: Mon, 06 Feb 2023 12:59:18 -0800 Organization: None to speak of Lines: 84 Message-ID: <87v8keecih.fsf@nosuchdomain.example.com> References: <0mvDL.68810$b7Kc.34465@fx39.iad> <87h6vzupzf.fsf@nosuchdomain.example.com> MIME-Version: 1.0 Content-Type: text/plain Injection-Info: reader01.eternal-september.org; posting-host="181ea74528ed435ea473c00f282e1c24"; logging-data="3366995"; mail-complaints-to="abuse@eternal-september.org"; posting-account="U2FsdGVkX1/icDT+Ivfnx+I7Sg8NcFuc" User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/27.2 (gnu/linux) Cancel-Lock: sha1:jI4mSSD20G3QORkWXm9gPDVPwag= sha1:d6YrgvZKwPKdNSqIck7YTSOGu9k= Xref: csiph.com comp.lang.c++:88993 comp.lang.c:169206 Richard Damon writes: > On 2/6/23 4:01 AM, Keith Thompson wrote: >> Paavo Helde writes: >>> 04.02.2023 18:59 Muttley@dastardlyhq.com kirjutas: >> [...] >>>> Ok, that makes sense. Though I wonder - given the main C compilers >>>> used now >>>> are also C++ compilers - why they don't just silently allow it in C? Would >>>> there be any wierd or unfortunate side effects in C if it was? >>> >>> This would basically introduce a limited notion of 'reference' in the >>> C language, which would make the language more complex to teach and >>> learn. That's the opposite to the goals C currently has. >>> >>> And yes, having a simpler language in general means the programs >>> written in it need to be more complex. Here it just means a couple of >>> extra asterisks and ampersands, so not a big deal. >> I agree it's not a big deal, and I've already commented on the idea >> that most C compilers are also C++ compilers (they aren't). >> But I don't think that allowing a conditional expression to be an >> lvalue would require introducing references into the language. >> An lvalue is defined as "an expression that potentially designated >> an object" ("potentially" because *ptr is still an lvalue even if >> ptr==NULL). This would just mean that `x ? y : z` would designate an >> object, and the identify of that object depends on the value of `x`. >> It's not all that much different from `a[i]`, where the identify >> of the object it designates depends on the values of `a` and `i` >> -- or `*p` for that matter. >> If it were introduced in a future edition of the C standard, I don't >> think it would cause many problems. (And the generated machine code >> would probably use pointers anyway.) But I don't think there's enough >> demand to justify it. > > I think the limitation is that to parse: > > int a, b, c, d; > ... > (a ? b : c) = d; > > The "value" used from b, for parsing the expression isn't the integral > value of b, but something like a int& that will convert to the value > if needed. The whole design phylosophy of C is that while parse the > statement, you can evalutate the pieces largly ignoring remote > context. The remembering that you have an l-value never has to survive > going through an "expression" but always (currently) used just by the > immediate operator that consumes it. > > To implement this, b's l-valueness (if c is also an l-value) through > the conditional operator to see how that is being used. Since > conditionals can be nested, this requires a somewhat arbitrary level > of delay. > > C++ get around that by b evaluating not to the value of b, but a > reference to b, that switches to the value if used as an r-value or > stays an l-value if used as an assignment target. > > Thus the ability comes straight out of having a "reference" type. You make a good point that I hadn't thought of: that a conditional expression would have to be an lvalue or not depending whether b and c are both lvalues (and there would have to be tighter constraints on the types of b and c in that case). And that would violate the (probably not quite universal) rule that every C (sub-)expression is evaluated without regard to its context. The C++ standard does define this in terms of "lvalue references". But I think this could be defined in a future edition of C without using C++-style references, probably by saying something like: A conditional expression is an lvalue if and only if the second and third operands are lvalues of compatible types. (it would probably require substantially more words than that). I don't think it would be worth the effort, but I wouldn't complain if the C committee and implementers decided to do it. (I'll note that a parenthesized lvalue is an lvalue, but of course that's much more trival than a conditional expression.) -- Keith Thompson (The_Other_Keith) Keith.S.Thompson+u@gmail.com Working, but not speaking, for XCOM Labs void Void(void) { Void(); } /* The recursive call of the void */