September 2002 - Boost-users - lists.preview.boost.org

RE: [Boost-Users] Re: [BGL] Best way to abort an algorithm from w ithi n a visitor?
by Garriss, Michael 19 Sep '02

19 Sep '02

I agree. I hate to see the return -1 (for example) on some "exceptional" case in a function when an exception makes more sense. It's not necessarily an "oops" case, it's an "exception." Either way the caller has to check the special case. There is a good reason that exceptions are named exceptions and not errors. Mike Garriss -----Original Message----- From: Peter Dimov [mailto:pdimov@mmltd.net] Sent: Thursday, September 19, 2002 10:20 AM To: Boost-Users(a)yahoogroups.com Subject: Re: [Boost-Users] Re: [BGL] Best way to abort an algorithm from withi n a visitor? From: "Mark Storer" <mstorer(a)cardiff.com> > Using exceptions in this way just bugs me. Religious reasons I guess. As > the Esteemed Mr. Hylsop ;) mentioned earlier, exceptions are for "oops". > This seems alot like the whole "do{}while(0)" with various 'break' > statements sprinkled in to avoid multiple returns/nested ifs. Useful, but > Just Not Right. A common misconception. Many people try to project their own ideas/expectations on language features, rather than simply using them as tools where appropriate. There is nothing "oops"-y about exceptions. They are a control flow mechanism; one that has interesting properties, but a control flow mechanism nonetheless. If it works and leads to a better code, it's foolish to avoid it without a technical reason. A similar example is when people don't use "struct" for C++ classes because it "doesn't feel right", or when trying to choose between "typename" and "class" for template parameters. The language features simply do not carry this kind of religious/emotional payload. It's all in your mind. Yahoo! Groups Sponsor ADVERTISEMENT <http://rd.yahoo.com/M=234081.2377872.3808768.2225242/D=egroupweb/S=17050067 88:HM/A=1225247/R=0/*http://ad.doubleclick.net/clk;4547223;7007303;y?http:// www.ameriquestmortgage.com/welcome.html?ad=Yahoo01> <http://us.adserver.yahoo.com/l?M=234081.2377872.3808768.2225242/D=egroupmai l/S=:HM/A=1225247/rand=147617524> Info: < http://www.boost.org <http://www.boost.org> > Wiki: < http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl <http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> > Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service <http://docs.yahoo.com/info/terms/> . [Non-text portions of this message have been removed]

1 0

lambda lib & mingw gcc 2.95.2?
by Mynhard Rudolph 19 Sep '02

19 Sep '02

Hi all Does the lambda library compile under mingw with gcc 2.95.2? I'm a newbie and I can't figure out what's wrong with the following test example: #include <boost/lambda/bind.hpp> #include <boost/lambda/lambda.hpp> using namespace boost; using namespace boost::lambda; struct A { int d; }; int main (int argc, int argv[]) { A* a = new A(); int b = (a ->* &A::d); int c = (_1 ->* &A::d)(a); return c - b; } g++ test.cpp -Wall -Ix:/boost_1_28_0/ -D__PROCESSOR_X86__ -D__WIN32__ -D__GCC__ -o test.o test.cpp: In function `int main(int, int *)': test.cpp:12: no match for `const boost::lambda::lambda_functor<boost::lambda::pl aceholder<1> > & ->* int A::*' make: *** [test.exe] Error 1 Would appreciate any clues. ;-) Mynhard

2 3

RE: [Boost-Users] [BGL] Best way to abort an algorithm from withi n a visitor?
by Mark Storer 18 Sep '02

18 Sep '02

If efficiency is your goal, exceptions are not the answer. Just enabling exceptions at compile time makes a noticable performance hit. Actually throwing an exception can gobble up a 1000 cycles, give or take. That's per throw. If you catch and throw again, you'll pay again. Now if you've got more than a hundred or so entries left to go, exceptions really are the greener grass. What's wrong with 'return', perhaps along-side some sort of 'done' status variable? I'm only vaguely familiar with the vistor pattern, but these would seem to be the obvious choice. --Mark Storer Software Engineer Cardiff Software #include <disclaimer> typedef std::disclaimer<Cardiff> Discard;

2 1

[BGL] Best way to abort an algorithm from within a visitor?
by Chris Russell 18 Sep '02

18 Sep '02

Is there a way to abort an algorithm while it's processing? Can I throw an exception out of a visitor method? I have a case where I need the algorithm to crunch on a large graph until some condition dependent on property data is reached. If and when the condition is met, I would like to skip the processing of the remaining vertices/edges and save the cycles. Does using exceptions sound like a reasonable approach? - Thanks Chris P.S. Thanks for the help with properties and property maps the other day. I cleaned up my code and eventually tracked the source of my considerable confusion to a bug in an operator= defined on the object that I'm using as an internal property. This bug manifested itself by re-initializing previously set data seemingly at random >8|-~ Arg... My bad.

2 2

RE: [Boost-Users] [BGL] Best way to abort an algorithm from withi n a visitor?
by Jim.Hyslop 18 Sep '02

18 Sep '02

Steve M. Robbins [mailto:steven.robbins@videotron.ca] wrote: > I use that technique to abort the Dijkstra shortest path > search when I find my > goal node. It works fine, as far as I can tell, but I'd be > interested to > learn if there is a better way. It seems anti-idiomatic to me. Exceptions are intended to signify exceptional circumstances that you would not normally expect to encounter (at least according to one school of thought). An exception tells me "Oops, something went wrong here" not "we succeeded, here's the answer". Many of the exceptions I catch are called "oops": try { /* ... */ } catch (std::exception &oops) { /* ... */ } One possible snag might be if your code interacts with someone else's code, which will catch the exception and not pass it on to you. In the context we're talking about, though, I don't see that as a significant problem. An exception is certainly an easy way to abort the search, and since I don't have an alternative to offer, I'll shut up now ;=) -- Jim

2 1

Levenshtein Distance algorithm
by Philippe Lalande 18 Sep '02

18 Sep '02

Hi, I was wondering if anyone could tell me if Boost contains a function that could perform a 'Levenshtein distance' calculation on a sequence. The Levenshtein distance is a measure of the similarity between two strings calculated as the number of deletions, insertions, or substitutions required to transforms the source string to the target string. IMO, it is an interesting algorithm that could apply to any kind of sequence. I'd be happy to discuss interface and implementation on the dev list if such algorithm does not exist yet. Thank you Philippe

2 1

RE: [Boost-Users] Re: Newbie needs help w/VC6 & filesystem (take 2)
by Jim.Hyslop 18 Sep '02

18 Sep '02

John B. Williston [mailto:yg-boost-users@gmane.org] wrote: > On Tue, 17 Sep 2002 12:40:53 -0400, Douglas Gregor <gregod(a)cs.rpi.edu> > wrote: > > >Could you copy 'n' paste some of the link errors you are > getting? It seems > >like you did the right thing (at least, I don't know of a > better way), but > >perhaps knowing what symbols are being redefined would help > isolate the > >problem. > > I sure can. The following is the compile output: > > --------------------Configuration: TestFs - Win32 > Debug-------------------- > Compiling... > Main.cpp > Linking... > LINK : warning LNK4075: ignoring /INCREMENTAL due to /FORCE > specification > msvcprtd.lib(MSVCP60D.dll) : warning LNK4006: "public: __thiscall > [...] already > defined in Main.obj; second definition ignored I've typically found these errors when you have mixed the run-time library usage, or MFC usage. One part of your project wants to link in the C run-time library as a DLL, but another part of your project wants to link it in as a static library. Make sure that all the projects, including the ones that build any libraries you're linking with, use the same "Use runtime library" option (Project | Settings, "C/C++" tab, "Code Generation" category), and also that all projects use the same "Microsoft Foundation Class" option (on the General tab). -- Jim

2 1

checked_delete with pointer template parameter
by Craig Henderson 18 Sep '02

18 Sep '02

I need to free memory from a container with elements of pointers, where the container value_type is a pointer type. I cannot use the checked_delete template as it is, as the template parameter is defined without the '*' pointer. For example, the following fails (obviously): typedef std::vector<char *> X; X x; ... std::for_each(x.begin(), x.end(), boost::checked_delete<X::value_type>); I'm using VC7, so have no partial specialization available, so using the type traits library remove_pointer<> does not work: typedef boost::remove_pointer<typename X::value_type>::type type; std::for_each(x.begin(), x.end(), boost::checked_delete<type>); My existing solution uses a unary functor, as below, but this met with critism on the Boost Developers' list. What would be the 'boost' solution to this problem? template<typename T> class free_heap_ptr : public std::unary_function<T, void> { public: free_heap_ptr() { } ~free_heap_ptr() { } void operator() (T p) { delete p; } }; ... std::for_each(changes_.begin(), changes_.end(), free_heap_ptr<typename change_list::value_type>()); TIA -- Craig

2 1

Is there a Type2Type in boost ?
by Baptiste Lepilleur 17 Sep '02

17 Sep '02

Hi everyone, I've looked around and failed to find the Type2Type template, detailed by Andrei Alexandrescu in a CUJ article (http://cuj.com/experts/1810/alexandr.htm?topic=experts) in boost. While the implementation is trivial: template<typename Type> struct Type2Type { typedef Type original_type; }; I'd rather have it in a well known library. Has it been replaced by something else ? Baptiste.

2 1

Newbie needs help with VC6 & filesystem library
by John B. Williston 17 Sep '02

17 Sep '02

Greetings, Though I'm an "old hand" at C++, I'm a complete newbie when it comes to Boost. The piece that really caught my eye, in light of my current project, is the directory iterator. To date, however, I haven't had any success getting it to compile and function under Win32 with MSVC6, so I thought I would post here. I first have a couple of general questions about using Boost, and then one specific question about the problems with my project at hand. (1) Is there a "right" way to include Boost stuff with VC6? Adding the Boost base directory to my list of include directories seems to work for the official stuff, but it clearly doesn't work for the filesystem stuff because Directory.h is not in the base\boost directory. I can put it there myself, of course, but I'm wondering if there is some intended way to use the beta libraries before they become a part of the whole. (2) Is there a "right" way to link Boost stuff with VC6? The filesystem stuff, like the regex stuff and some other pieces, requires actual compilation and cannot simply be included as header files. I have already built all the official libraries using the bjam utility, but I cannot help noticing they're each stored in hideously long paths beneath the boost base directory. Is there a "right" directory under the Boost hierarchy into which the resulting libs should be copied for ease of inclusion in VC6? Or should I copy them into my project directories as needed? Or should I simply come up with my own system? To get to the more specific question, let me explain how I've addressed (1) and (2) for the time being. Regarding (1), I've copied Directory.h and Boost.h into the base\boost directory so that they may be included without changing the existing source code. That was simple enough. Regarding (2), I've decided to add Directory.cpp to my project and compile it as a module, rather than try to get the linking to work externally. Having said that, consider the following code: #include <windows.h> #include "boost/directory.h" int WINAPI WinMain(HINSTANCE hInstance, HINSTANCE hPrevInstance, LPSTR lpCmdLine, int nCmdShow) { using namespace boost::filesystem; std::string strFilename; for (dir_it it("./"); it != dir_it(); ++it) if (!get<is_hidden>(it)) strFilename = *it; return 0; } This is an extremely minimal test case of integrating the filesystem stuff into a Win32 application. When compiling the application, I get the following lengthy and inscrutable (to me, at least) error message: WinMain.obj : error LNK2001: unresolved external symbol "__declspec(dllimport) public: __thiscall boost::filesystem::get<struct boost::filesystem::is_hidden>::get<struct boost::filesystem::is_hidden>(class boost::filesystem::dir_it const &)" (__imp_ ??0?$get@Uis_hidden@filesystem@boost@@@filesystem@boost@@QAE@ABVdir_it@12@@Z) The error message seems to suggest that the filesystem stuff isn't available in my project. I find that confusing because the Directory.cpp module compiles without any errors, and I am thus inclined to think that, in fact, all of the filesystem stuff is in my project. Thus, I must also ask the third question: (3) Can anyone explain to me what I might be doing wrong and, more importantly, how I might correct it? Thanks much to all the Boost authors for an interesting set of tools. And thanks in advance to any who respond for all the help with my specific issue. John

6 6