C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Modifying a sequence with std::for_each? Goto page 1, 2, 3  Next          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message Peter C. Chapin Guest Posted: Tue Aug 03, 2004 11:24 am    Post subject: Modifying a sequence with std::for_each? Is it well defined to pass std::for_each a function that modifies its argument (for example, a function that takes an argument by reference)? Both g++ v3.3.1 and MSVC++ v7.1 will compile such usage but the programs bomb when they run. For example, when the g++ produced program runs, I get a SIGSEGV the first time std::for_each attempts to use the given function. Peter [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top John Potter Guest Posted: Wed Aug 04, 2004 10:14 am    Post subject: Re: Modifying a sequence with std::for_each? On 3 Aug 2004 07:24:29 -0400, "Peter C. Chapin" <pchapin (AT) sover (DOT) net> wrote: Quote: Is it well defined to pass std::for_each a function that modifies its argument (for example, a function that takes an argument by reference)? Yes. Quote: Both g++ v3.3.1 and MSVC++ v7.1 will compile such usage but the programs bomb when they run. For example, when the g++ produced program runs, I get a SIGSEGV the first time std::for_each attempts to use the given function. G++ works fine. Post the bad code so that we can show you your error. John [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top john dibling Guest Posted: Wed Aug 04, 2004 12:11 pm    Post subject: Re: Modifying a sequence with std::for_each? On 3 Aug 2004 07:24:29 -0400, "Peter C. Chapin" <pchapin (AT) sover (DOT) net> wrote: Quote: Is it well defined to pass std::for_each a function that modifies its argument No, it is not. for_each() takes Input Iterators, which cannot be assigned through. Use transform() instead. - john dibling dibsatdiblingdotcom [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Peter C. Chapin Guest Posted: Wed Aug 04, 2004 1:50 pm    Post subject: Re: Modifying a sequence with std::for_each? John Potter <jpotter (AT) falcon (DOT) lhup.edu> wrote in news:FfLPc.24008$iK.11057 @newsread2.news.atl.earthlink.net: Quote: G++ works fine. Post the bad code so that we can show you your error. My program was attempting to modify a string literal (indirectly) and this was the problem. String literals contain constant characters despite the fact that you can put their address into a pointer to non-const character. Peter [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top red floyd Guest Posted: Thu Aug 05, 2004 3:17 am    Post subject: Re: Modifying a sequence with std::for_each? john dibling wrote: Quote: On 3 Aug 2004 07:24:29 -0400, "Peter C. Chapin" wrote: Is it well defined to pass std::for_each a function that modifies its argument No, it is not. for_each() takes Input Iterators, which cannot be assigned through. Use transform() instead. Josuttis lists for_each() as a modifying algorithm (pp 325,326), and notes that it's more efficient for modify in place than transform(), due to the lack of the assignment. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Ali Cehreli Guest Posted: Thu Aug 05, 2004 10:39 am    Post subject: Re: Modifying a sequence with std::for_each? On Wed, 04 Aug 2004 05:11:18 -0700, john dibling wrote: Quote: On 3 Aug 2004 07:24:29 -0400, "Peter C. Chapin" wrote: Is it well defined to pass std::for_each a function that modifies its argument No, it is not. for_each() takes Input Iterators, which cannot be assigned through. This is incorrect. for_each works with any iterator-like type that supports operator!= operator++ operator* As long as the operation (functor) passed to for_each accepts what operator* returns, it will work fine. Quote: Use transform() instead. That works too. Ali P.S. If I am not mistaken, transform is the only option when it is needed to be guaranteed that the functor is not copied. I can't remember that subtle difference between the two but I think it was about for_each not giving the guarantee that the functor will not be copied. And I think transform gives the guarantee that the functor will not be copied. I am confused now. :) [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Peter C. Chapin Guest Posted: Thu Aug 05, 2004 10:39 am    Post subject: Re: Modifying a sequence with std::for_each? john dibling <dibs (AT) dibling (DOT) com> wrote in news:i2hvg092j559hjsm9l4i3566slo4k6nmeb (AT) 4ax (DOT) com: Quote: Is it well defined to pass std::for_each a function that modifies its argument No, it is not. for_each() takes Input Iterators, which cannot be assigned through. Hmm. Isn't this more of a restriction on the implementation of for_each than it is on the program that uses for_each? The for_each template might only use the facilities of input iterators, but there's no reason I can't provide it with, say, random access iterators. In other words, if I *know* that I'm giving for_each a sequence that can be modified, can I expect for_each to behave itself if I also give it a function that tries to modify that sequence? Peter [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top llewelly Guest Posted: Thu Aug 05, 2004 10:51 am    Post subject: Re: Modifying a sequence with std::for_each? John Potter <jpotter (AT) falcon (DOT) lhup.edu> writes: Quote: On 3 Aug 2004 07:24:29 -0400, "Peter C. Chapin" wrote: Is it well defined to pass std::for_each a function that modifies its argument (for example, a function that takes an argument by reference)? Yes. But 25.1.1 says the arguments to for_each are InputIterators, and 24.1.1 says InputIterators are not assignable. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top James Hopkin Guest Posted: Thu Aug 05, 2004 10:54 am    Post subject: Re: Modifying a sequence with std::for_each? john dibling <dibs (AT) dibling (DOT) com> wrote Quote: On 3 Aug 2004 07:24:29 -0400, "Peter C. Chapin" wrote: Is it well defined to pass std::for_each a function that modifies its argument No, it is not. for_each() takes Input Iterators, which cannot be assigned through. Use transform() instead. Sorry, that's just wrong. InputIterator is the *minimum* requirement for for_each. True, if you pass input iterators which aren't forward iterators (e.g. istream_iterator), you can't use a modifying functor. However, there's no problem passing forward iterators and modifying elements in the range. Use transform if you *don't* want to modify in place. James [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Francis Glassborow Guest Posted: Thu Aug 05, 2004 2:08 pm    Post subject: Re: Modifying a sequence with std::for_each? In article <86smb241ls.fsf (AT) Zorthluthik (DOT) local.bar>, llewelly <llewelly.at (AT) xmission (DOT) dot.com> writes Quote: John Potter <jpotter (AT) falcon (DOT) lhup.edu> writes: On 3 Aug 2004 07:24:29 -0400, "Peter C. Chapin" wrote: Is it well defined to pass std::for_each a function that modifies its argument (for example, a function that takes an argument by reference)? Yes. But 25.1.1 says the arguments to for_each are InputIterators, and 24.1.1 says InputIterators are not assignable. No, it requires that they be at least InputIterators it does not (quite correctly) require that they are only that. What the Standard does is require that for_each() work even if the iterator you provide only meets the minimum requirements for and InputIterator. -- Francis Glassborow ACCU Author of 'You Can Do It!' see http://www.spellen.org/youcandoit For project ideas and contributions: http://www.spellen.org/youcandoit/projects [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top John Dibling Guest Posted: Fri Aug 06, 2004 10:22 am    Post subject: Re: Modifying a sequence with std::for_each? Quote: Josuttis lists for_each() as a modifying algorithm (pp 325,326), and The book you quote is 5 years old, and since that there has been a new C++ standard published. In that standard, for_each() is listed in section 25.1.1 under "Non-Modifying Sequence Operations." So, Josuttis is out-of-date with regards to for_each(). Even still, my Austern book (Generic Programming and the STL) is also copyright 1999, same as Josuttis, and Austern lists for_each() as non-mutating as well. In addidtion, Austern describes the non-mutating algorithms as a whole like this: "This chapter describes basic STL algorithms that operate on a range of iterators without changing the elements those iterators point to." (p.199) I can't explain or understand why the two books dissagree with each other regarding for_each(), becasue although I don't know the J. book well, I know it is respected. But whatever. The authoritative source is the Standard, and it says for_each() is a non-modigying sequence operation. Therefore any operation passed to for_each() which modifies the sequence is ill-formed. Quote: notes that it's more efficient for modify in place than transform(), due to the lack of the assignment. The Standard says that the complexity for : for_each(first, last, f) is "exactly last-first applications of f." Furthermore, it says that the complexity for: transform(first, last, result, op) is exactly last-first applications of op. So the complexity of both functions is the same. I don't see what extra assignment J. could be talking about, but I don't think he was right. - John Dibling dibsatdiblingdotcom - John Dibling dibsatdiblingdotcom [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top John Dibling Guest Posted: Fri Aug 06, 2004 10:25 am    Post subject: Re: Modifying a sequence with std::for_each? Quote: for_each works with any iterator-like type that supports [snip] This is true, and what you have just described is what the Standard defines an an InputIterator. (24.1.1) Quote: P.S. If I am not mistaken, transform is the only option when it is needed to be guaranteed that the functor is not copied. I can't remember that subtle difference between the two but I think it was about for_each not giving the guarantee that the functor will not be copied. And I think transform gives the guarantee that the functor will not be copied. I am confused now. Instead of focusing on just the minutiae technical details of for_each()'s and transform()'s requirements, focus instead on the big-picture intent of the two algorithms. for_each() is intended to pass each element in a collection to a function, f. transform() is intended to create new elements in a collection based on the elements that are already in some collection. The differences are not subtle, they are gross. Its possible to use for_each to do some kind of accumulation, by repeatadly modifying the functor that is passed to for_each(). If this is what you intend to do however, you might consider using accumulate() instead, if for no other reason that for the sake of semantic clarity. But you have to pick your spots - accumulate() is right sometimes, for_each() other times. Note however that it is *not* possible to use for_each() to modify the original sequence. This is what the OP's original post was all about (at least, that's what the sublect line indicates). And this is my main point in my original reply to the OP. It is also possible to use transform() to not just go from one collection to another -- you can also do an in-place change in a source collection. This functionality is guaranteed by the standard: [25.2.3] : transform(InputIterator first, InputIterator last, OutputIterator result, UnaryOperation op) : : NOTE: result may be equal to first [...] For example, suppose you have a collection of ints. If you wanted to run throught the collection and multiply each value by 3, putting the new value back in the same collection in place of the original, you can use transform, like this: (Note: TestVals could also be a vector, list, or what have you) [FILE: main.cpp] #include <cstdlib> #include <algorithm> #include <functional> #include <iostream> using namespace std; int TestVals [] = {1, 2, 3, 4, 5}; static const size_t NumTestVals = sizeof(TestVals)/sizeof(TestVals[0]); template<class T> struct dump { void operator()(const T&t) const { cout << t << endl; } }; int main() { cout << "original collection:" << endl; for_each(&TestVals[0],&TestVals[NumTestVals],dump transform(&TestVals[0],&TestVals[NumTestVals],&TestVals[0],bind1st(multiplies<int>(),3)); cout << "resulting collection:" << endl; for_each(&TestVals[0],&TestVals[NumTestVals],dump return 0; } [END FILE: main.cpp] - John Dibling dibsatdiblingdotcom [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top John Dibling Guest Posted: Fri Aug 06, 2004 10:26 am    Post subject: Re: Modifying a sequence with std::for_each? Quote: Isn't this more of a restriction on the implementation of for_each than it is on the program that uses for_each? My claim is that it is both. Quote: if I *know* that I'm giving for_each a sequence that can be modified, can I expect for_each to behave itself if I also give it a function that tries to modify that sequence? In order for your assertion to be correct, you must know something about the implementation of for_each(). It goes against the idea of even having a Standard at all to make the programmers know implementation details of the interfaces defined in the Standard. In other words, these things are supposed to be black boxes. The interface defines the iterators as InputIterators. In order for your code to work, you must know that for_each() doesn't use them as just InputIterators, becasue if it did, they could be const. For instance, the Standard says that for_each() must iterators that can be used as InputIterators. Well, InputIterators can't be assigned through, so you could pass in a const T*, and that is an InputIterator. Similarly, there is nothing in the standard that says for_each() can't explicitly convert whatever iterator you pass in to a const version of that iterator. Of course you can't assign through a const pointer,so any attempt to use that implemenation of for_each to modify the collection would not compile. Now of course, you aren't going to win any awards for wrinting a for_each() like that, but that's wuite beside the point. The point is it is standard compliant to do so. - John Dibling dibsatdiblingdotcom [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top John Dibling Guest Posted: Fri Aug 06, 2004 10:26 am    Post subject: Re: Modifying a sequence with std::for_each? Quote: InputIterator is the *minimum* requirement for for_each. Not only is it the minumum requitrement, it is the only requirement regarding the iterators. Quote: However, there's no problem passing forward iterators and modifying elements in the range. From a conceptual point of view, the problem is that the Standard defines for_each() as a non-Mutating sequence operation, but you have used it to mutate the sequence. It is true that in most cases you can contort your code to get for_each to mutate the sequence just fine. But the subject of the OP was: "Modifying a sequence with for_each()?". According to the Standard, this is an ill-defined use of for_each(). Here's some sample code that illustrates what I believe the OP is trying to do: [FILE: main.cpp] #include <cstdlib> #include <algorithm> #include <functional> #include <iostream> using namespace std; int TestVals [] = {1, 2, 3, 4, 5}; static const size_t NumTestVals = sizeof(TestVals)/sizeof(TestVals[0]); template<class T> struct dump { void operator()(const T&t) const { cout << t << endl; } }; template struct multiply_assign : public unary_function<T,void> { multiply_assign(T mul) : m_mul(mul) {}; void operator()(T& r) const { r *= m_mul; } private: T m_mul; }; int main() { cout << "original collection:" << endl; for_each(&TestVals[0],&TestVals[NumTestVals],dump // TEST LINE int * pFirst = &TestVals[0], * pLast = &TestVals[NumTestVals]; for_each(pFirst,pLast,multiply_assign<int>(3)); cout << "resulting collection:" << endl; for_each(&TestVals[0],&TestVals[NumTestVals],dump return 0; } [END FILE: main.cpp] The multiply_assign functor is used by for_each to modify the sequence. My claim is that this is not a Standard-friendly way to use for_each, and here's why. Since for_each takes InputIterators, it should be possible to send const pointers to for_each. If the way for_each is being used is Standard-compliant, the new code should still compile. But if you change the test line from: // TEST LINE int * pFirst = &TestVals[0], * pLast = &TestVals[NumTestVals]; ....to... // TEST LINE const int * pFirst = &TestVals[0], * pLast = &TestVals[NumTestVals]; .... the code no longer compiles. Quote: Use transform if you *don't* want to modify in place. You can use transform to modify in place. This functionality is guaranteed by the Standard. I discuss this in greater depth in my reply to another post, and there is runnable sample code there as well. Now, I must also point out that in re-reading the OPs original post, "Is it well defined to pass std::for_each a function that modifies its argument " This question is different from the subject line. I t*is* well formed to modify the 3rd paramater to for_each(). You would do this to sum up a vector of ints, for instance, and this is fine. But what you cannot do is pass a functor to for_each() that changes the values in the sequence. I hope I haven't answered a question that the OP didn't intend to ask... - John Dibling dibsatdiblingdotcom [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Francis Glassborow Guest Posted: Fri Aug 06, 2004 3:16 pm    Post subject: Re: Modifying a sequence with std::for_each? In article <thi4h0h8cu0mhdv3uv7sbb17he79sp8fn6 (AT) 4ax (DOT) com>, John Dibling <jdibling (AT) yahoo (DOT) com> writes Quote: The book you quote is 5 years old, and since that there has been a new C++ standard published. In that standard, for_each() is listed in section 25.1.1 under "Non-Modifying Sequence Operations." So, Josuttis is out-of-date with regards to for_each(). Even still, my Austern book (Generic Programming and the STL) is also copyright 1999, same as Josuttis, and Austern lists for_each() as non-mutating as well. In addidtion, Austern describes the non-mutating algorithms as a whole like this: "This chapter describes basic STL algorithms that operate on a range of iterators without changing the elements those iterators point to." (p.199) I can't explain or understand why the two books dissagree with each other regarding for_each(), becasue although I don't know the J. book well, I know it is respected. But whatever. The authoritative source is the Standard, and it says for_each() is a non-modigying sequence operation. Therefore any operation passed to for_each() which modifies the sequence is ill-formed. One (Nico's) is written by a programmer (albeit one who is an active and knowledgeable member of WG21) who knew from experience that for_each() usually can modify the elements. Matt Austern comes from the implementation side of the C++ community and as such is far more aware of the exact wording of the C++ Standard. In theory using for-each() with a modifying predicate is an error, in practice it usually works because making it always fail requires, I think, an inefficient implementation of for_each(). -- Francis Glassborow ACCU Author of 'You Can Do It!' see http://www.spellen.org/youcandoit For project ideas and contributions: http://www.spellen.org/youcandoit/projects [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Display posts from previous: All Posts1 Day7 Days2 Weeks1 Month3 Months6 Months1 Year Oldest FirstNewest First         C++Talk.NET Forum Index -> C++ Language (Moderated) All times are GMT Goto page 1, 2, 3  Next Page 1 of 3   Jump to: Select a forum C/C++----------------C LanguageC Language (Moderated)C++ Language (Moderated)C++ language (comp.lang.c++)C++ language, library and standards Non-English C/C++ Newsgroups----------------C++ (German)C++ (French)  You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum Powered by phpBB © 2001, 2006 phpBB Group SEO toolkit © 2004-2006 webmedic.