April 2006 - Boost - lists.boost.org

Re: [boost] property tree & registry approach
by Alexander 28 Apr '06

28 Apr '06

"Marcin Kalicinski" <kalita(a)poczta.onet.pl> wrote in message news:<e2t2kb$6ja$1(a)sea.gmane.org>... > My answer is quite short, but I discovered a major flaw with the approach > you propose, and it renders most of it unimplementable. Please see below: > > > May be it will be interesting to compare another approach to the proposed > > Property Tree Library. > > > class basic_registry_proxy > > { > > public: > > ... > > template<class T> virtual bool get(const path, T&) const = 0; > > No can do. Virtual functions cannot be templates! > > This rules out automatic type conversions. It would have to rely on > something else to provide it. You are absolutely right. At my implementation basic_registry_proxy has only simple get/set functionality: virtual bool get(const path, string&) const = 0; Template function belongs to wrapper which contains pointer to basic_registry_proxy. I was trying to simplify my class hierarchy and produce silly indeed interface. I am sorry. Regards, Alexander

1 0

Testing Property Tree xml parser
by Jose 28 Apr '06

28 Apr '06

Hi, I am trying to finish my review of the Property Tree. I find the scope of the library for program configuration very useful. I am currently testing only the read_xml parsing, and although it is only meant for very simple xml files i find its xml support very very sketchy. I've performed multiple experiments with rss feed files, which are very simple xml files. This is the loop I am using to test different xml paths using a command line utility: BOOST_FOREACH(ptree::value_type &v, pt.get_child( argv[2] )) cout << "value: " << v.second.data() << endl; Below is a list of my experiments: 1. parsing the artima.com spotlight feed Result: FAILED The path is rdf:RDF.item.title and I get invalid character entitly. I think the parser should support the semicolon within the tag name, given that in many cases the config files might be generated by real xml programs which use namespaces and it should be able to read them even if it does not support save. 2. parsing the MSDN visual c++ feed Result: FAILED The path is rss.channel.item.title and I get an "xml parse error". Is there a posibility of getting more meaningful errors ? 3. parsing the main CNN feed Result: FAILED The path is rss.channel.item.title. This query fails with no error but if the path is shortened to rss.channel.item it dumps all the values within item, but there is no value at that level (only nested tags) 4. Parsing the Google News RSS feed Result: FAILED The path is rss.channel.item.title. I get "Invalid character entity error". A more meaningful error should be possible with the position in the file where the entity occurs. 5. Parsing the Google News Atom feed Result: FAILED The path is feed.entry.title. I get "Invalid character entity error". regards jose

3 11

[SoC] Boost not listed?
by Joaquín Mª López Muñoz 28 Apr '06

28 Apr '06

As of now, the following page http://code.google.com/soc/ does not list Boost amongst the mentor organizations. Maybe we missed some deadline or were rejected, is it transitory? Maybe a ping to soc2006support(a)google.com is in order? Thank you, Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

4 3

[Property tree review] Comments & layout information
by Marcin Kalicinski 28 Apr '06

28 Apr '06

I have an interesting idea how to preserve comments & layout information. The relevant structures should be included in the tree under a special meta key: Config file: key1 data1 ;comment before key2 key2 data2 ;comment after key2, on the same line The resulting tree might look like that: key1 data1 key2 data2 _metadata_ { key1 { row 1 column 1 } comment "comment before key2" { row 2 column 1 } key2 { row 3 column 1 } comment "comment after key2, on the same line" { row 3 column 15 } } This way it is easily accessible using ptree interface, it is readable, editable, and it is possible to manipulate it programatically. And above all, it does not require any extra support from ptree, only from the parsers. With some effort it also allows 100% reproduction of original layout. --Marcin

2 2

property tree & registry approach
by Alexander Belyakov 28 Apr '06

28 Apr '06

May be it will be interesting to compare another approach to the proposed Property Tree Library. I was highly pleased to see this library but here I focused at differences which could offer more flexible and scalable solution at some aspects of registry behavior. Of course, this flexibility has some cost and functionality of libraries not overlapped fully. But the registry functionality is the most interesting part for me. class path; class basic_registry_proxy { public: ... template<class T> virtual bool get(const path, T&) const = 0; template<class T> virtual bool set(const path, T&) = 0; ... } //basic_registry_proxy - represents the whole regestry tree. //Not any key but whole tree. //path - identify actual key at basic_registry_proxy. class simple_registry : public basic_registry_proxy; // uses win API class win_registry : public basic_registry_proxy; // uses win API class thread_safe_registry : public basic_registry_proxy; // uses Boost.Thread class dispersed_registry : public basic_registry_proxy; // stores data at multiple files class basic_parser { public: virtual load(basic_registry_proxy*, const char* PathatAtProxy = 0) const = 0; virtual save(basic_registry_proxy*, const char* PathatAtProxy = 0) const = 0; } class win_registry_parser : public basic_parser; class XML_registry_parser : public basic_parser; usage example: // copy windows registry to simple_registry (1) win_registry_parser wregparser("HKEY_LOCAL_MACHINE/SOFTWARE/"); simple_registry sregistry; wregparser.load(&sregistry); // convert windows registry to simple_registry (2) win_registry wregistry; simple_registry sregistry; copy(&wregistry, path("/"), &sregistry, path(/)); Assume that at windows registry we have such structure: somekey1 defaulttext1 REG_SZ somekey2 somekey3 defaulttext2 REG_SZ somevalue1 somedata REG_SZ somevalue2 1234 REG_DWORD Lets choose prefix "#" for subsidiary attribute information. So we could get new representation of registry. somekey1 defaulttext1 { #role key {} #type REG_SZ {} somekey2 { somekey3 defaulttext2 { #role key{} #type REG_SZ {} } somevalue1 somedata { #role value {} #type REG_SZ {} } somevalue2 1234 { #role value{} #type REG_DWORD {} } } } At this case the new registry syntaxes is richer than windows registry one. Regards, Alexander Belyakov

2 1

Compilation problem :Regex
by mahesh karanth 28 Apr '06

28 Apr '06

Hello All, When I try to use regex with boost.multiindex I got the following compilation error. But when I indipendently compile multiindex package and regex package both compiles without any error. What is the problem? Can anyone please help me to getout of this Error dump: D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/regbase.hpp(125) : error C2955: 'collate' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\locale(67) : see declaration of 'collate' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/perl_matcher.hpp(32) : error C2955: 'map' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\map(140) : see declaration of 'map' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/perl_matcher.hpp(36) : error C2955: 'map' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\map(140) : see declaration of 'map' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/perl_matcher.hpp(40) : error C2955: 'map' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\map(140) : see declaration of 'map' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/perl_matcher.hpp(44) : error C2955: 'map' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\map(140) : see declaration of 'map' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/perl_matcher.hpp(48) : error C2955: 'map' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\map(140) : see declaration of 'map' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/regbase.hpp(125) : error C2955: 'collate' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\locale(67) : see declaration of 'collate' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/perl_matcher.hpp(32) : error C2955: 'map' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\map(140) : see declaration of 'map' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/perl_matcher.hpp(36) : error C2955: 'map' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\map(140) : see declaration of 'map' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/perl_matcher.hpp(40) : error C2955: 'map' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\map(140) : see declaration of 'map' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/perl_matcher.hpp(44) : error C2955: 'map' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\map(140) : see declaration of 'map' D:\BOOST\INCLUDE\BOOST-1_33_1\boost/regex/v4/perl_matcher.hpp(48) : error C2955: 'map' : use of class template requires template argument list C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\map(140) : see declaration of 'map' D:\MYPROJECTS\BRIGHTCLOUD AGENT\URLNORMALIZATION\CDNList.h(18) : warning C4251: 'CDNVector' : class 'std::vector<class std::basic_string<char,struct std::char_traits<char>,class std::allocator<char> >,class std::allocator<class std::basic_string<cha Best Regards Mahesh Karanth --------------------------------- Yahoo! Messenger with Voice. PC-to-Phone calls for ridiculously low rates.

3 12

[bind] bind_visit_test failure on GCC 4.1.0 and HP/CXX
by Markus Schöpflin 28 Apr '06

28 Apr '06

The test fails with the following message: bind_visit_test.cpp(61): test 'vis.hash == 34' failed in function 'int main()' It has been failing on HP/CXX only for quite some time, but now I noted it also fails on GCC 4.1.0 Linux and Tru64. Is this an issue with the test? Markus

8 21

[mpl] error compiling with evc4
by Michael Fink 28 Apr '06

28 Apr '06

Hello, while porting Boost to evc4 I encountered the following error in boost\mpl\aux_\preprocessed\msvc60\and.hpp (and the same in or.hpp). Note that evc4 is very similar to msvc6, so it uses most of the defines set for this compiler. The error messages are: {path}\boost\mpl\aux_\preprocessed\msvc60\and.hpp(42) : error C2908: explicit specialization; 'and_impl<1>' has already been instantiated from the primary template {path}\boost\mpl\aux_\preprocessed\msvc60\and.hpp(42) : error C2039: 'result_' : is not a member of '`global namespace'' {path}\boost\mpl\aux_\preprocessed\msvc60\and.hpp(42) : error C2143: syntax error : missing ';' before '<' {path}\boost\mpl\aux_\preprocessed\msvc60\and.hpp(42) : error C2059: syntax error : '<' {path}\boost\mpl\aux_\preprocessed\msvc60\and.hpp(44) : error C2143: syntax error : missing ';' before '{' {path}\boost\mpl\aux_\preprocessed\msvc60\and.hpp(44) : error C2447: missing function header (old-style formal list?) [...] The code in and.hpp looks as follows: namespace boost { namespace mpl { namespace aux { template< bool C_ > struct and_impl { template< typename T1, typename T2, typename T3, typename T4 > struct result_ : false_ { }; }; template<> struct and_impl<true> { template< typename T1, typename T2, typename T3, typename T4 > struct result_ : and_impl< BOOST_MPL_AUX_NESTED_TYPE_WKND(T1)::value >::template result_< T2,T3,T4,true_ > { }; }; template<> struct and_impl<true> ::result_< true_,true_,true_,true_ > : true_ { }; } // namespace aux [...] The compiler doesn't recognize the out-of-class template instanciation. I can comment out the last struct definition, but then the compiler gives errors in other places where the template is used. How can I solve that with the evc4 compiler? Thanks, Michael

1 0

rational.hpp requires working unary minus
by Yuriy Koblents-Mishke 28 Apr '06

28 Apr '06

Test Code: std::numeric_limits<long> lm; boost::rational<long> mn( lm.min() ); std::cout << mn << '=' << mn / 1 << std::endl; Output: -2147483648/1=1/0 Reason: lm.min() == -2147483648 lm.max() == 2147483647 and unary minus is not well defined for lm.min(); on my system -lm.min() == -2147483648 The problem is expected with any computer representing negative integers using binary two's compliment.

1 0

evc4 port of Boost
by Michael Fink 28 Apr '06

28 Apr '06

Hello, I'm currently trying to port Boost for the Microsoft eMbedded Visual C++ 4.0 (evc4) compiler that targets PocketPC 2003 and other devices based on Windows CE 4.2. The compiler's frontend (parser, etc.) is based on Visual C++ 6.0, with little to no changes. As Boost (at present) supports this compiler, porting should be rather easy (or so I thought). For Standard C++ library support I use STLport 5, since evc4 doesn't have a working one. Before putting more work into porting, I have some questions: Is it desirable to port Boost for this platform? I read that at some version support for vc6 will be dropped. If possible, I would like to have at least some version of Boost that works with evc4. Where can I get help with more complex error messages that appear during porting? What I've done so far: I created a custom evc4-stlport-tools.jam file in boost\tools\build\v1 so that I can build different libraries. It's more of a hack-together than a cleanly integrated build file. I changed some Boost.Config files to support evc4 compiler. I fixed all issues for evc4 in libs\thread\src\* so that the Boost.Thread library compiles without warnings and errors. thanks, Michael

2 1