- Boost - lists.boost.org

Re: [boost] Interest in dimension/unit-checking library?
by Keith Nicewarner 08 Jan '06

08 Jan '06

Geoffrey Romer wrote: > Would there be any interest in a library for attaching units (physical > or otherwise) to numeric variables, so that the type system could > enforce dimension- and unit-consistency in all calculations, and > convert semi-automatically between compatible units? I'm fairly new to boost, so forgive my late entry into this discussion. I too recognized the need for this kind of library back in the late 90s and wrote this library available on SourceForge: http://tuoml.sourceforge.net/ It's pretty stable and used on several projects (there's a military simulation tool that has standardized on it). I've recently discovered the meta-programming method for dimensional quantities by Scott Meyers: http://se.inf.ethz.ch/~meyer/publications/OTHERS/scott_meyers/dimensions.pdf This inspired me to revisit my library and it is turning out to be a fairly easy conversion. The old library had to explicitly represent each unit combination of interest (e.g., Speed = Length/Time), which is ugly and doesn't allow arbitrary combinations. The new method fixes all this (in fact, it simplifies things to the point where there is no library object file any more -- just headers). A key concept I tried to enforce is that any time the programmer sets or gets a numeric value, the specific units (i.e., Feet, Meters, etc.) must be specified. I don't want any units to be implicit. My library provides ostream formatting for handy printing. I also provide scalar, 2D, and 3D types. For the 2D/3D types, I provide various transformation operators like rotation. For 3D rotations, I provide the option of cosine matrix or unit quaternion. It's been suggested to me that I submit my library to boost. Being new to boost, I'm not sure how complete and tested the code needs to be. Any feedback or suggestions are welcome. Keith.

11 37

Re: [boost] Performance comparison between ptr_vectorandvector<cow_ptr<Shape> >, vector<copy_ptr<Shape> >
by David Maisonave 08 Jan '06

08 Jan '06

"Thorsten Ottosen" <tottosen(a)dezide.com> wrote in message news:<dpmc2d$f75$1(a)sea.gmane.org>... > David Maisonave wrote: > > "Thorsten Ottosen" <tottosen(a)dezide.com> wrote in message > > >>I'm not sure I fully understand how the detection mechanism would > >>work in your cow_ptr. > > > > > > Example code: > > template<typename T_obj> > > cow_ptr(T_obj* type): m_ref_count(new ref_count(type, > > get_alloc_func(type))) > > { > > assert(typeid(*type) == typeid(T_obj)); > > } > > how can that assertion ever fail when T_obj is deduced? The only way that cow_ptr and copy_ptr can get slice is if the wrong type is pass to the constructor. Example code: Derivedtype* p = new DerivedDerivedType; cow_ptr<BaseType> pBase(p); //This will cause an assertion since p type does not match the pointee > > > Since the cow_ptr works by capturing the type on the constructor, it > > can easily determine if the right type is being passed in, and if it > > matches the pointee type. That can't be done with the boost pointer > > containers because they don't catch the type on the constructor, nor > > is it part of the required logic for using a clone function method. > > > > > >>Nevertheless, as soon as the object is outside > >>a cow_ptr might be sliced. That's reason enough not to allow it. > > > > > > I'm not sure what you're referring to here. Can you post an > > example? > > *ptr = *ptr2; Can you please give a more complete example? Which is the smart pointer, here, and exactly what type of pointers?

2 1

[variant] Feature request
by Ion Gaztañaga 08 Jan '06

08 Jan '06

Hi, These days I was trying to use boost::variant to create a read-write lock with many priority policies (reader priority, writer priority, alternating...). Each of this policies need different members (some need one mutex and two condition variables, other two mutexes and one condition variable...). These can be expensive OS resources. Trying to optimize resources I wanted to use boost::variant. The constructor of the read-write lock takes an argument with the wanted policy (just like boost::read_write_mutex) and I need to initialize resources according to this run-time policy. The problem is that I can't directly call the correct constructor, because boost::variant requires a convertible type to possible internal types to call that constructor. What I would propose is the possibility of selecting via integer value which member of boost::variant will be constructed, just like which() returns the index of constructed member. This way, we can efficiently compact members of classes. class MyRWLock { public: enum { r_prioriy, w_prioriy, a_prioriy }; MyRWLock(int run_time_policy); private: struct r_prioriry_members { //... }; struct w_prioriry_members { //... }; struct a_prioriry_members { //... }; //Union of needed members boost::variant<r_prioriry_members, w_prioriry_members, a_prioriry_members> m_data; }; inline MyRWLock::MyRWLock(int run_time_policy) : m_data(/*what do I have to do to select a constructor?*/) { } Since an integer can be a convertible parameter of one variant type constructor, maybe we can use a wrapper holding the integer as variant argument. namespace boost { struct int_selection { explicit int_selection(int selection) : m_selection(selection){} int get_selection const { return m_selection; } private: int m_selection; } } I don't know if this feature is hard to achieve generically (that would require big meta-programming implementation, surely), but this bounded factory-like interface (constructing one of selected types from an integer) would be in my opinion a big win. One alternative is to have the first variant type as a dummy type whose default constructor is trivial and after that use operator=. But that can require dynamic memory (according to variant documentation). Do you see this as an unneeded feature or it's just too hard to implement? Regards, Ion

2 1

[iostream] possible bug
by Nicola Musatti 08 Jan '06

08 Jan '06

Hallo, when trying to build up a configuration for the new Borland BCB 2006 compiler I got an error message complaining about line 49 in file boost/iostreams/positioning.hpp: inline stream_offset fpos_t_to_offset(fpos_t pos) ^^^^^^ Shouldn't the above read std::fpos_t ? This solves the problem for my compiler. Cheers, Nicola Musatti

1 0

[typeof] function references and gcc 3.3.5
by Alexander Nasonov 08 Jan '06

08 Jan '06

The following code doesn't compile on gcc 3.3.5 typedef int (fun)(bool); fun& foo(); BOOST_STATIC_ASSERT((boost::is_same<BOOST_TYPEOF(foo()), fun>::value)); The error is error: no matching function for call to `ensure_obj(int(&)(bool))' This can be fixed by introducing more specialized ensure_obj: template<class T> T& ensure_obj(const T&, int); template<class R, class T0> R(& ensure_obj(R (&)(T0), ...) )(T0); template<class R, class T0, class T1> R(& ensure_obj(R (&)(T0,T1), ...) )(T0,T1); // ... ( second argument is not required for gcc 3.3.5, it's there to resolve ambuiguity on other compilers ) This fix has been tested on three different compilers: > ~/pkgsrc-root/gcc3/bin/g++ -v Reading specs from /home/alnsn/pkgsrc-root/gcc3/lib/gcc-lib/i386-pc-freebsd6/3.3.5/specs Configured with: ./configure --prefix=/home/alnsn/pkgsrc-root/gcc3 --host=i386-pc-freebsd6 --enable-shared --enable-languages=c++ Thread model: posix gcc version 3.3.5 > g++ -v Using built-in specs. Configured with: FreeBSD/i386 system compiler Thread model: posix gcc version 3.4.4 [FreeBSD] 20050518 > /usr/local/intel_cc_80/bin/icpc -v -V Intel(R) C++ Compiler for 32-bit applications, Version 8.1 Build 20050207Z Package ID: l_cc_pc_8.1.028 Copyright (C) 1985-2005 Intel Corporation. All rights reserved. FOR NON-COMMERCIAL USE ONLY Version 8.1 Also, other contexts mentioned by Arkadiy in other thread should be modified. BTW, why test/function_ref.cpp checks only test_wrapper type, not a function reference? -- Alexander Nasonov

3 8

m4 macro for boost?
by Thomas Porschberg 07 Jan '06

07 Jan '06

Hi, I want integrate some of the boost libraries in a project which has the GNU autoconf/automake tools as build system. I have to cope some questions like: Where is your boost library installed ? Is the version of your boost-lib recent enough ? Can I compile a small program using the boost libraries ? Unfortunately I'm not an experienced GNU auto[conf/make] user. But I note that a lot of software components come with a m4 macro to support the configuration process. I found a boost.m4 file at http://alien.cern.ch/cache/org.glite/project/boost.m4. I tried it to include to my project but it needs some revision. Can someone from the boost people give some advice ? I know that boost does use bjam. But GNU auto[conf/make] are very common(due to cygwin/mingw even under windows) so that I hope to find an answer. best regards, Thomas --

4 9

Iostreams 'drain' algorithm proposal
by Andrew Walrond 07 Jan '06

07 Jan '06

Hello list. New here, but a Boost user so thought it was time I started lurking with a view to contributing. When using boost::iostreams, I find I frequently need/implement a 'drain' algorithm. This is identical to boost::iostreams::copy, except that it doesn't close the Sink. This allows mutiple sources to be 'drained' into a single sink. Unless there is a generic algorithm for doing this that I have missed, I'd like to propose the addition of the 'drain' algorithm to boost::iostreams. I have a patch, should there be any interest. Andrew Walrond

2 2

main cvs anonymous login problem
by Bjørn Roald 07 Jan '06

07 Jan '06

Hi, I am trying to update my cvs working directory, but trying to log on to the server failes. Anybody have a clue what is going on?? cvs -d:pserver:anonymous@cvs.sourceforge.net:/cvsroot/boost login Logging in to :pserver:anonymous@cvs.sourceforge.net:2401/cvsroot/boost CVS password: cvs login: authorization failed: server cvs.sourceforge.net rejected access to /cvsroot/boost for user anonymous Thanks Bjørn

4 4

How do I use bjam?
by Daryle Walker 07 Jan '06

07 Jan '06

I want to try out the documentation run described by the message archived at <http://article.gmane.org/gmane.comp.lib.boost.documentation/1472>. I downloaded bjam for Mac OS X from our SourceForge download area. I placed it in my grand Boost directory in my user directory. (I have separate sub-directories for Boost-1.33.1-from-a-ZIP, Boost CVS, and the sandbox CVS.) I made a link from bjam to ~/bin. When I ran bjam, I got: Unable to load Boost.Build: could not find "boost-build.jam" --------------------------------------------------------------- Attempted search from /Users/daryle up to the root and in these directories from BOOST_BUILD_PATH and BOOST_ROOT: /usr/share/SOMETHING-I-FORGOT. Please consult the documentation at 'http://www.boost.org'. Now, I said I have the Boost files within my user directory, _not_ in a system level directory like Boost.Build assumes. I tried exporting a value for BOOST_BUILD_PATH to my CVS version of Boost, and it didn't do anything: Unable to load Boost.Build: could not find "boost-build.jam" --------------------------------------------------------------- Attempted search from /Users/daryle up to the root and in these directories from BOOST_BUILD_PATH and BOOST_ROOT: ~/Documents/Programming/Boost/main-cvs/boost/. Please consult the documentation at 'http://www.boost.org'. It doesn't matter if I put "bjam --v2" instead, nor any other command line option. Why can't it lock on to my Boost files? When I say "ran bjam," I just entered "bjam" as a command within a random directory, like my home directory, with no special Boost files within it. I also ran it like the Gmane message, but it still couldn't start up. A related question: is the regular Boost.Build set-up useful for someone that creates Boost libraries using the regular and sandbox CVSes? After all, any installed headers and (static and/or dynamic) libraries would get stale after any CVS update. -- Daryle Walker Mac, Internet, and Video Game Junkie darylew AT hotmail DOT com

4 9

[range] why not range_result_iterator remove_cv?
by MB 07 Jan '06

07 Jan '06

Hi, I think non-intrusive customization way needs four specializations; 'X', 'const X', 'volatile X' and 'const volatile X'. At last, twenty specializations are required for five metafunctions. I view that's something like the pitfall of public virtual interfaces. Only range_result_iterator happens to work around by using remove_cv. See: I added followings to <libs/range/doc/example.cpp>: template< class Range > typename boost::range_result_iterator<Range>::type my_begin(Range& rng) { return boost::begin(rng); } // then... my_begin(cpair); // error Why not something like: template< class T > struct range_result_iterator : mpl::eval_if< is_const<T>, range_const_iterator< typename remove_cv<T>::type >, range_iterator< typename remove_cv<T>::type > > { }; Regards, MB p-stade.sourceforge.net

2 2