April 2007 - Boost-users - lists.preview.boost.org

math toolkit review
by Arnaldur Gylfason 25 Apr '07

25 Apr '07

Hi, This is my contribution to the boost::math review. This is a partial review of components 1) and 2) mentioned in the formal review announcement, i.e. 1) the statistical distributions and 2) special functions. ? What is your evaluation of the design? Very good in my opinion. I think using global functions is very appropriate in libraries such as this one. ? What is your evaluation of the implementation? I did not look at the implementation but I compared it with another implementation and it compares favorably. More on this below. ? What is your evaluation of the documentation? Very clear. I only browsed it, but that was enough to get the feel for the library and start using it right away. ? What is your evaluation of the potential usefulness of the library? Very useful in applications that deal with anything related to statistics or probability. ? Did you try to use the library? With what compiler? Did you have any problems? Yes I used it on a Linux box using g++ 4.0.2. No problems, except I was using version 1.33.0 of boost so I had to change all use of tuple to be in namespace tuples. Took a short time though. ? How much effort did you put into your evaluation? A glance? A quick reading? In-depth study? I guess a quick reading plus I wrote 2 small test programs that compare accuracy and speed with implementations from the R-project. ? Are you knowledgeable about the problem domain? I have been working in a development group in a Statistics department of a genetics company for several years so I guess I'm fairly knowledgeable. ? Do you think the library should be accepted as a Boost library? Yes, definitely. The test programs are compareAccuracy.cc and compareSpeed.cc. The programs were compiled with g++ 4.0.2 using -O3 and run on a Linux box with Intel Xeon 2.4 GHz. I can give more info on this if anyone needs more info. comparAccuracy calls several statistical distribution functions over a range of values, comparing results with what similar routines from the R-project give. The output gives the maximum abs difference and maximum of relative absolute difference between boost::math and R. The results were: norm pdf max diff: 5.55112e-17 norm pdf max rel diff: 2.20594e-16 t pdf max diff: 1.11022e-16 t pdf max rel diff: 1.17145e-15 norm cdf max diff: 1.11022e-16 norm cdf max rel diff: 2.55269e-15 t cdf max diff: 1.59317e-14 t cdf max rel diff: 3.18634e-14 F pdf max diff: 1.77636e-15 F pdf max rel diff: 7.05565e-14 F cdf max diff: 5.55112e-16 F cdf max rel diff: 1.56512e-14 chi2 pdf max diff: 5.55112e-17 chi2 pdf max rel diff: 4.26471e-14 chi2 cdf max diff: 3.46945e-16 chi2 cdf max rel diff: 6.27547e-14 exp pdf max diff: 0 exp pdf max rel diff: 0 exp cdf max diff: 0 exp cdf max rel diff: 0 gamma pdf max diff: 2.77556e-17 gamma pdf max rel diff: 3.84987e-15 gamma cdf max diff: 1.11022e-16 gamma cdf max rel diff: 1.85721e-15 lognorm pdf max diff: 1.11022e-16 lognorm pdf max rel diff: 2.21914e-16 lognorm cdf max diff: 1.11022e-16 lognorm cdf max rel diff: 1.43305e-15 cauchy pdf max diff: 0 cauchy pdf max rel diff: 0 cauchy cdf max diff: 1.11022e-16 cauchy cdf max rel diff: 2.2064e-16 weibull pdf max diff: 1.11022e-16 weibull pdf max rel diff: 2.22042e-16 weibull cdf max diff: 0 weibull cdf max rel diff: 0 norm quantile max diff: 8.88178e-16 norm quantile max rel diff: 4.00185e-16 t quantile max diff: 4.07907e-14 t quantile max rel diff: inf F quantile max diff: 1.11022e-15 F quantile max rel diff: 1.08231e-15 F quantile max diff: 1.11022e-15 F quantile max rel diff: 1.08231e-15 chi2 quantile max diff: 4.26326e-14 chi2 quantile max rel diff: 3.92646e-16 exp quantile max diff: 5.55112e-17 exp quantile max rel diff: 2.11093e-16 gamma quantile max diff: 3.55271e-15 gamma quantile max rel diff: 3.56265e-16 lognorm quantile max diff: 1.42109e-14 lognorm quantile max rel diff: 9.10487e-16 cauchy quantile max diff: 2.17426e-12 cauchy quantile max rel diff: 0.104537 weibull quantile max diff: 1.11022e-16 weibull quantile max rel diff: 2.0992e-16 uniform pdf max diff: 0 uniform pdf max rel diff: 0 uniform cdf max diff: 0 uniform cdf max rel diff: 0 uniform quantile max diff: 0 uniform quantile max rel diff: 0 beta pdf max diff: 1.11022e-15 beta pdf max rel diff: 3.14562e-15 beta cdf max diff: 4.44089e-16 beta cdf max rel diff: 2.9679e-15 beta quantile max diff: 2.22045e-16 beta quantile max rel diff: 8.38281e-16 binom pdf max diff: 2.08167e-17 binom pdf max rel diff: 0.00011212 binom cdf max diff: 9.4369e-16 binom cdf max rel diff: 3.46145e-14 poisson pdf max diff: 1.04083e-17 poisson pdf max rel diff: 1.05393e-13 poisson cdf max diff: 0 poisson cdf max rel diff: 0 nbinom pdf max diff: 1.38778e-17 nbinom pdf max rel diff: 1.16396e-13 nbinom cdf max diff: 8.88178e-16 nbinom cdf max rel diff: 2.37872e-15 lgamma max diff: 1.13687e-13 lgamma max rel diff: 0.211576 tgamma max diff: 7.43092e+142 tgamma max rel diff: 9.15759e-14 A few results need comment. ->t quantile max diff: 4.07907e-14 ->t quantile max rel diff: inf Here division by 0 gives inf. max diff is quite acceptable though. Some guards should be placed to handle 0 or near 0 but I haven't done it. I'm not worried about the accuracy but anyone may feel free to remedy this. ->cauchy quantile max diff: 2.17426e-12 ->cauchy quantile max rel diff: 0.104537 Here we are almost surely again dividing by something near 0. Similar comments as above apply. ->lgamma max diff: 1.13687e-13 ->lgamma max rel diff: 0.211576 Same old story ->tgamma max diff: 7.43092e+142 ->tgamma max rel diff: 9.15759e-14 At the higher end of the range tgamma gives such high numbers, that max diff doesn't tell us much. max rel diff gives acceptable results though. No quantile comparisons are given for the discrete distributions since R always returns integers but boost::math gives fractional values. It depends on application what you want obviously but giving results within the range (discrete in this case) is in some sense more natural. All in all I am quite happy with the results. Obviously R could be wrong but it's use is widespread in the Statistics community so it should be very well tested. comparSpeed calls a few statistical distribution functions repeatedly (see code), timing the loop, repeating the same thing for the R version. The output is the time in second: N repeats = 5000000 R norm pdf: 1.31 boost::math norm pdf: 1.1 R norm cdf: 1.49 boost::math norm cdf: 0.27 R norm quantile: 0.75 boost::math norm quantile: 1.83 R lgamma: 4.02 boost::math lgamma: 7.4 N repeats = 1000000 R t pdf: 0.91 boost::math t pdf: 4.85 R t cdf: 0.17 boost::math t cdf: 0.03 R t quantile: 12.88 boost::math t quantile: 56.15 R F pdf: 0.06 boost::math F pdf: 0.05 R F cdf: 0.07 boost::math F cdf: 0.85 R F quantile: 18.59 boost::math F quantile: 54.54 The results vary somewhat but this would be acceptable for my applications. quantile functions seem to be somewhat slower in boost::math than in R. Maybe they're more accurate. More detailed analysis would be called for if people are concerned about this. The 2 test programs are attached to this message. If they're somehow scrambled, anyone can write to me and ask for them. Last but certainly not the least, my compliments to the authors of boost::math ;-) cheers Arnaldur Gylfason, deCODE genetics

2 1

Regex match problem
by Kalin Nakov 25 Apr '07

25 Apr '07

Hi, I experience big troubles with regex matching. The code below should obviously match 'abc', but does not, because it tries to match the whole string. If r='abcdefg', then match is successfull, but if it is only 'abc' and input is 'abcdef', the match always fails as described below. boost::match_results<std::string::const_iterator> m; std::string inp = "abcdefg"; boost::regex r = "abc"; if (!boost::regex_match(inp, m, r, boost::match_default)) // always fails, don't know why How can I successfully match 'abc' agains input 'abcdefg'? Thank you very much in advance, Kalin

2 1

Re: [Boost-users] Regex match problem
by KLINIK Markus SDI2-G (AREVA NP GmbH) 25 Apr '07

25 Apr '07

You are right, regex_match tries to match the whole input string. What you want is regex_search. Best Regards Markus

1 0

Serialization and filesystem::path
by Hughes, James 25 Apr '07

25 Apr '07

Hello all, Am I right in thinking that I need to provide serialization support for filesystem::path myself because it is not built in to the serialization libraries? If that is the case, what is the best way to go about it? I had thought it was built in originally, but not been able to get things to compile - have I missed something? James This message (including any attachments) contains confidential and/or proprietary information intended only for the addressee. Any unauthorized disclosure, copying, distribution or reliance on the contents of this information is strictly prohibited and may constitute a violation of law. If you are not the intended recipient, please notify the sender immediately by responding to this e-mail, and delete the message from your system. If you have any questions about this e-mail please notify the sender immediately.

2 1

[std][msvc++][vs2005sp1] unable to open iostream with greek letters in path name
by TOL Tsagarakis Manos 25 Apr '07

25 Apr '07

Hello everyone, I know this is not a boost related issue, but I hope someone has a solution. The question is very simple and should not even be asked: “How can I open a file using std::iostreams, if the files full path, contains Greek letters (or probably any other language except English)?” The test case below shows part of the problem as it successfully opens the file using std::oftream and writes some Greek in it. Then I can read back its contents using std::ifstream. Now the file name in windows explorer or command line should be "ελληνικό όνομα αρχείου.txt" (in case you can't read this, it is "greek file name.txt"). But instead it becomes "åëëçíéêü üíïìá áñ÷åßïõ.txt" Actually when this is done by a full MFC application it is impossible to open the file! I get assertion failure! I use MS Visual Studio 2005 sp1 and std lib and CRT that comes with it. Has anyone faced this problem before? Do I have to do something special in order to be able to use standard iostreams library with Greek file names? Thank you in advanced, Emmanuil Tsagarakis // test_greek_filenames.cpp compile with: /EHsc #include <iostream> #include <fstream> #include <iterator> #include <algorithm> #include <string> #include <vector> #include <assert.h> using namespace std; namespace { struct string_by_line : std::string { }; std::istream & operator >> (std::istream & is, string_by_line & strLine) { getline(is, strLine); return is; } const string s_strFileName("ελληνικό όνομα αρχείου.txt"); } int main(int argc, char **argv) { ofstream ofs; ofs.open(s_strFileName.c_str()); assert(ofs.good()); ofs << "Writing to file ..." << endl; ofs << "Ελληνικά περιεχόμενα στο αρχείο (greek file contents) ..." << endl; ofs.close(); ifstream ifs; ifs.open(s_strFileName.c_str()); assert(ifs.good()); std::vector<std::string> vecContents; std::copy(std::istream_iterator<string_by_line>(ifs ), std::istream_iterator<string_by_line>(), std::back_inserter(vecContents)); for (size_t i=0; i<vecContents.size(); ++i) cout << vecContents.at(i) << endl; ifs.close(); }

2 2

Re: [Boost-users] [SPAM] - Re: [bind, regex] strange MSVC 7.1 compiler effect: Members ofunnamednamespace are put in boost::re_detail when stdafx.hincludesbind.hpp and regex.hpp - E-Mail-Adresse wurde in Betreff-Zeile gefunden.
by Stefan Felkel 25 Apr '07

25 Apr '07

> -----Ursprüngliche Nachricht----- > Von: boost-users-bounces(a)lists.boost.org > [mailto:boost-users-bounces@lists.boost.org]Im Auftrag von > John Maddock > Gesendet: Mittwoch, 25. April 2007 11:38 > An: boost-users(a)lists.boost.org > Betreff: [SPAM] - Re: [Boost-users] [bind,regex] strange MSVC 7.1 > compiler effect: Members ofunnamednamespace are put in > boost::re_detail > when stdafx.hincludesbind.hpp and regex.hpp - E-Mail-Adresse wurde in > Betreff-Zeile gefunden. > > > Stefan Felkel wrote: > > Hi there, > > > > I have a strange compiler effect with Microsoft Visual Studio.NET > > 2003 and boost 1.33.1 in a project that uses bind and regex. All > > Symbols defined in an unnamed namespace are part of the namespace > > boost::re_detail::'anonymous-namespace'. The effect occurres when I > > use precompiled headers and bind.hpp is included before regex.hpp. > > > > This is my stdafx.h which is used to generate the > precompiled header: > > > > #pragma once > > > > #include <boost/bind.hpp> > > #include <boost/regex.hpp> > > > > This is the main source file: > > > > #include "stdafx.h" > > #include "boost/current_function.hpp" > > > > namespace { > > void foo( void ) > > { > > printf( "%s\n", BOOST_CURRENT_FUNCTION ); > > } > > } > > > > int main( int argc, char * argv[] ) > > { > > foo(); > > return 1; > > } > > > > The program generates this output: > > > > void __cdecl boost::re_detail::`anonymous-namespace'::foo(void) > > > > instead of the expected output: > > > > void __cdecl `anonymous-namespace'::foo(void) > > > > When I change the include order in stdafx.h so that regex.hpp is > > included before bind.hpp the output is as expected. > > Has someone else came across such a problem? Is it a compiler bug? > > Weird indeed, I don't believe it's our problem as such: > except that both > regex and bind use unnamed namespaces. Looks like the poor > compiler is > getting confused :-( > > John. > > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users > I think you're right - Microsoft Visual Studio.NET 2005 doesn't have this problem. Stefan

1 0

Adapter to object instance
by MainBrain 25 Apr '07

25 Apr '07

Hello, im looking through http://www.boost.org/libs/functional/index.html, there are ways to access a class method but i cant find a way to access a method of an object. How can i write an adapter for "instance.bar()" in this example: class foo { int _i; public: bar(); // manipulates _i } instance; Greetings, Yelve Yakut -- View this message in context: http://www.nabble.com/Adapter-to-object-instance-tf3632838.html#a10144039 Sent from the Boost - Users mailing list archive at Nabble.com.

2 2

[bind, regex] strange MSVC 7.1 compiler effect: Members of unnamed namespace are put in boost::re_detail when stdafx.h includes bind.hpp and regex.hpp
by Stefan Felkel 25 Apr '07

25 Apr '07

Hi there, I have a strange compiler effect with Microsoft Visual Studio.NET 2003 and boost 1.33.1 in a project that uses bind and regex. All Symbols defined in an unnamed namespace are part of the namespace boost::re_detail::'anonymous-namespace'. The effect occurres when I use precompiled headers and bind.hpp is included before regex.hpp. This is my stdafx.h which is used to generate the precompiled header: #pragma once #include <boost/bind.hpp> #include <boost/regex.hpp> This is the main source file: #include "stdafx.h" #include "boost/current_function.hpp" namespace { void foo( void ) { printf( "%s\n", BOOST_CURRENT_FUNCTION ); } } int main( int argc, char * argv[] ) { foo(); return 1; } The program generates this output: void __cdecl boost::re_detail::`anonymous-namespace'::foo(void) instead of the expected output: void __cdecl `anonymous-namespace'::foo(void) When I change the include order in stdafx.h so that regex.hpp is included before bind.hpp the output is as expected. Has someone else came across such a problem? Is it a compiler bug? Regards, Stefan

2 1

Re: [Boost-users] [program_options] missing typed_value
by Romulus 24 Apr '07

24 Apr '07

>> Hi everybody, [...] >> Here is the error message: >> option_groups.cpp: In function 'int main(int, char**)': >> option_groups.cpp:42: error: 'class >> boost::program_options::typed_value<std::basic_string<char, >> std::char_traits<char>, std::allocator<char> >, char>' has no member named >> 'implicit' >Can you please post some sample here, how you are trying to use it? I >assume there is a problem >with implicit function operator () call. But it would be clearer if you post >some usage sample >here. >With Kind Regards, >Ovanes Markarian The problem occurs with the sample file (attached) : /usr/share/doc/libboost-doc/examples/libs/program_options/example/option_groups.cpp Regards

2 1

boost::thread::thread() crashes in pthread_create
by Venikov, Andy 24 Apr '07

24 Apr '07

Hi all, recently I had to move from Mandrake to RedHat linux. On the new system, a simple thing like this crashes: <------- test.cpp #include <boost/thread/thread.hpp> void * Thread1() { //sleep some time return 0; } int main(int, char**) { boost::thread thread1(Thread1); thread1.join(); return 0; } --------> The crash happens in the create_pthread call in boost::thread constructor. Here's the build command: g++ -c test.cpp g++ test.o ../boost/stage/lib/libboost_thread-gcc-mt.a -lpthread Outuput of g++ --version: g++ (GCC) 3.4.5 20051201 (Red Hat 3.4.5-2) Output of uname -a: Linux joe.rps.local 2.6.9-34.ELsmp #1 SMP Fri Feb 24 16:54:53 EST 2006 i686 i686 i386 GNU/Linux Everythink worked fine on Mandrake. I followed exactly the same steps to build boost on the new system.... Thanks, Andy.

2 1