C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Lightweight string class?          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message Alan McKenney Guest Posted: Thu Sep 15, 2005 9:13 am    Post subject: Lightweight string class? We are finding that the standard C++ string class, at least as implemented by Sun, is too expensive for what we do with it. We have switched from having std::string data members to using char [] members and using strncpy(), etc., and gotten a noticable (and needed) speed-up. Before I go out and reinvent the wheel, has anyone encapsulated C-style string handling in a "lightweight" string class? One thing is that we don't need a string object to be able to hold an arbitrarily large string; in most cases, when writing the code, we can deduce an upper bound on the size. This means you don't have to use "new" and "delete" for the data array. We also mainly just copy the strings around and compare them with other strings and with literal strings, so it doesn't need to be optimized for inserts and replaces. I checked Boost, but I didn't see anything like it there. Is this a frequently-enough encountered problem that there's a widely-used solution [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Maciej Sobczak Guest Posted: Thu Sep 15, 2005 10:18 am    Post subject: Re: Lightweight string class? Alan McKenney wrote: Quote: One thing is that we don't need a string object to be able to hold an arbitrarily large string; in most cases, when writing the code, we can deduce an upper bound on the size. I think then that what you are really looking for is not the "lightweight string class" (since general approaches to this use dynamic memory or at best the short string optimization), but rather "statically bounded string class". This could be a good idea. Quote: I checked Boost, but I didn't see anything like it there. I think that you could use boost::array<char, N> for this, with some helper functions for setting the value. Typedef is your friend, as well: typedef boost::array<char, 100> MyString; MyString s; set_string(s, "Some string literal"); What about book-keeping the current length of the string? Is this needed and would you take advantage of having it? -- Maciej Sobczak : http://www.msobczak.com/ Programming : http://www.msobczak.com/prog/ [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Alan McKenney Guest Posted: Thu Sep 15, 2005 11:00 pm    Post subject: Re: Lightweight string class? Maciej Sobczak wrote: Quote: Alan McKenney wrote: snip I think that you could use boost::array<char, N> for this, with some helper functions for setting the value. Typedef is your friend, as well: typedef boost::array<char, 100> MyString; MyString s; set_string(s, "Some string literal"); What about book-keeping the current length of the string? Is this needed and would you take advantage of having it? -- Maciej Sobczak : http://www.msobczak.com/ The small-string optimization would help us some, since many of our strings have max size 3 or 4. I'm not sure we want to require them to all have the same max size, since we also have strings with 32-char max. What I was hoping for was something like a template: LWString<3> mem1; LWString<32> mem2; etc., which could be used the way std::string is, at least for assignment, comparison, and conversion to and from const char *. I think I can "roll my own" in a few days. However, I thought that maybe someone has dealt with this in a more imaginative or just plain better way than I would. I thought that, before I reinvent the wheel (and a square wheel at that), I might consult the combined (amassed?) wisdom of the C++ community. -- Alan McKenne [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Old Wolf Guest Posted: Fri Sep 16, 2005 8:17 am    Post subject: Re: Lightweight string class? Alan McKenney wrote: Quote: We are finding that the standard C++ string class, at least as implemented by Sun, is too expensive for what we do with it. We have switched from having std::string data members to using char [] members and using strncpy(), etc., and gotten a noticable (and needed) speed-up. strncpy() is a poor choice for a string copying function, because it does not 0-terminate its output in some circumstances, and wastes time in most circumstances. For the amount of code required to use strncpy() safely, it turns out that just using memcpy() or some other such function is simpler. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Allan W Guest Posted: Fri Sep 16, 2005 8:19 am    Post subject: Re: Lightweight string class? Alan McKenney wrote: Quote: The small-string optimization would help us some, since many of our strings have max size 3 or 4. I'm not sure we want to require them to all have the same max size, since we also have strings with 32-char max. Maciej Sobczak's idea of a "statically bounded string class" would be like the "small-string optimization" carried to the max. Quote: What I was hoping for was something like a template: LWString<3> mem1; LWString<32> mem2; etc., which could be used the way std::string is, at least for assignment, comparison, and conversion to and from const char *. Exactly. Whatever your string size is, that's the size of the internal char array -- as you said, you wouldn't have to use new[] or delete EVER. As an implementation detail, you could derive LWString<int N> from LWStringBase, which would implement all of the operations that don't depend on N -- this would drastically reduce code bloat on some compilers with poor optimization. For instance, on some systems LWString<2>::operator char* would be distinct from LWString<3>::operator char*, even though they do exactly the same thing. The only disadvantage I see is that users always have to count the size that is needed. LWString<6> hello = "hello"; // Wastes 1 or 2 characters LWString<6> hello = "goodbye"; // Generates a run-time error The string length wouldn't be automatic, so it's a source of potential waste and/or errors... and I doubt it could be detected at compile time. Any method that involves counting the size of the string on assignment is going to end up calling operator new[], at least sometimes. Quote: I think I can "roll my own" in a few days. However, I thought that maybe someone has dealt with this in a more imaginative or just plain better way than I would. I thought that, before I reinvent the wheel (and a square wheel at that), I might consult the combined (amassed?) wisdom of the C++ community. When you've finished writing it, consider submitting it to Boost? [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top indranil Guest Posted: Fri Sep 16, 2005 12:28 pm    Post subject: Re: Lightweight string class? You can check out Andrei Alexandrescu's flex_string template. It has the same interface as std::string but he uses all his policy based design tricks to use different storage policies as template parameters. He supplies policies for small string optimisation, copy on string, vector of chars, amongst others. The site is down at the moment :- http://www.moderncppdesign.com "Alan McKenney" <alan_mckenney1 (AT) yahoo (DOT) com> wrote Quote: We are finding that the standard C++ string class, at least as implemented by Sun, is too expensive for what we do with it. We have switched from having std::string data members to using char [] members and using strncpy(), etc., and gotten a noticable (and needed) speed-up. Before I go out and reinvent the wheel, has anyone encapsulated C-style string handling in a "lightweight" string class? One thing is that we don't need a string object to be able to hold an arbitrarily large string; in most cases, when writing the code, we can deduce an upper bound on the size. This means you don't have to use "new" and "delete" for the data array. We also mainly just copy the strings around and compare them with other strings and with literal strings, so it doesn't need to be optimized for inserts and replaces. I checked Boost, but I didn't see anything like it there. Is this a frequently-enough encountered problem that there's a widely-used solution [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Tony Delroy Guest Posted: Fri Sep 16, 2005 12:29 pm    Post subject: Re: Lightweight string class? Old Wolf wrote: Quote: For the amount of code required to use strncpy() safely, it turns out that just using memcpy() or some other such function is simpler. Not so. For example, char buffer[1024]; std::strncpy(buffer, source, sizeof buffer - 1); buffer[sizeof buffer - 1] = ''; // note: just cater for unterminated case... vs size_t copy_len = std::max(std::strlen(source), sizeof buffer - 1); std::memcpy(buffer, source, copy_len); buffer[copy_len] = ''; The memcpy approach can be expected to be slower for having necessarily used strlen. Particularly bad is that strlen counts all the way to the end of the source string even though only the first (sizeof buffer - 1) characters are potentially copied. Cheers, Tony [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Allan W Guest Posted: Sat Sep 17, 2005 1:57 pm    Post subject: Re: Lightweight string class? Tony Delroy wrote: Quote: Old Wolf wrote: For the amount of code required to use strncpy() safely, it turns out that just using memcpy() or some other such function is simpler. Not so. For example, char buffer[1024]; std::strncpy(buffer, source, sizeof buffer - 1); buffer[sizeof buffer - 1] = ''; // note: just cater for // unterminated case... vs size_t copy_len = std::max(std::strlen(source), sizeof buffer - 1); std::memcpy(buffer, source, copy_len); buffer[copy_len] = ''; I think that Wolf was suggesting: std::memcpy(buffer, source, sizeof buffer); Depending on the definition of source, this could be perfectly safe. It might copy more bytes than the minimum required, but if it's quick enough, who cares? [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top kwikius Guest Posted: Sun Sep 18, 2005 2:45 pm    Post subject: Re: Lightweight string class? Allan W wrote: Quote: The only disadvantage I see is that users always have to count the size that is needed. LWString<6> hello = "hello"; // Wastes 1 or 2 characters LWString<6> hello = "goodbye"; // Generates a run-time error you could make use of non-type template parameters: #include <string> #include <iostream> template<char p[]> struct sstring{ size_t length()const{ return strlen(p);} operator char const*()const {return p;} }; #define STATIC_STRING(name,string_constant) namespace private_{ char name [] = string_constant; } sstring<private_::name > name; STATIC_STRING(s,"hello"); int main() { std::cout << s << " " << s.length(); } the only drawback is that it must always defned be in namespace scope regards Andy little [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Darko Miletic Guest Posted: Mon Sep 19, 2005 5:10 pm    Post subject: Re: Lightweight string class? indranil wrote: Quote: You can check out Andrei Alexandrescu's flex_string template. It has the same interface as std::string but he uses all his policy based design tricks to use different storage policies as template parameters. He supplies policies for small string optimisation, copy on string, vector of chars, amongst others. The site is down at the moment :- http://www.moderncppdesign.com It appears that site is moved to this location: http://erdani.org/ Darko [ 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 Page 1 of 1   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.