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

Re: [Boost-users] Threading a Shared Model with pointers
by pcunite＠fsmail.net 05 Mar '11

05 Mar '11

I explained this wrong. What is really happening is that each thread has a pointer to the object (forget about the vector for the moment) and it uses this as a key many hours later. When it is given serialized access to the "model/database cache", the thread wants to do work using that key, of course now it is no longer valid. This is why using map works, but I did not want to search every time, that is why weak_ptr works. Just wanted to know the best approach for this type of problem. All access to everything is thread safe. Any thoughts now?

1 0

newbie: installation question
by Gunnar Salentin 05 Mar '11

05 Mar '11

Hi folks I just started with boost and have run into the first problem: the example program (see http://www.boost.org/doc/libs/1_46_0/more/getting_started/unix-variants.html) does not run though everything should be in place: The error is 'example.cpp:1:35: error: boost/lambda/lambda.hpp: No such file or directory" yes, I did check my file structure (see attached extract from Terminal). Installation of boost was done via sudo port... Anyone an idea what the issue might be? Thx GS User:~ SUserHG$ ls Desktop Downloads Library Music Public QuantLib-1.0.1.tar.gz example.cpp Documents Install-QuantLib.txt Movies Pictures QuantLib-1.0.1 Sites SUser:~ SUserHG$ cd /opt/local/include/boost/lambda SUser:lambda SUserHG$ ls algorithm.hpp casts.hpp construct.hpp core.hpp exceptions.hpp lambda.hpp numeric.hpp bind.hpp closures.hpp control_structures.hpp detail if.hpp loops.hpp switch.hpp SUser:lambda SUserHG$ cd SUser:~ SUserHG$ c++ -I /opt/local/include/boost example.cpp -o example example.cpp:1:35: error: boost/lambda/lambda.hpp: No such file or directory example.cpp: In function ‘int main()’: example.cpp:8: error: ‘boost’ has not been declared example.cpp:8: error: ‘lambda’ is not a namespace-name example.cpp:8: error: expected namespace-name before ‘;’ token example.cpp:12: error: ‘_1’ was not declared in this scope

3 3

Threading a Shared Model with pointers
by pcunite＠fsmail.net 04 Mar '11

04 Mar '11

Threading a Shared Model with pointers I have a vector of pointers to objects created with new. Multiple threads access this vector in a safe manner with get/set. However, a thread may delete one of the objects, in which case another thread's pointer to the object is no longer valid. How can a method know if the pointer is valid? I have to examples that work and would like your thoughts. Test for pointer validity: 1. Use integers instead of pointers. A hash (std::map) checks to see if the pointer is still valid. Public methods look like get(size_t iKey), etc. Insde the method it does: if((it = mMap.find(iKey)) != mMap.end()) { TMyType * pMyType = it->second; // do something with know good pointer. } 2. Have a vector of shared_ptr. Each thread tries to call lock() on its weak_ptr. If the returned shared_ptr is null we know someone deleted it while we were waiting. Public methods look like get(boost::weak_ptr<TMyType> pMyType), etc. Insde the method it does: boost::shared_ptr<TMyType> pGOOD = pMyType.lock(); if (pGOOD == NULL) { // null, bad pointer do something return; } 3. Test for null on plain raw pointers? Is this possible? Public methods look like get(TMyType * pMyType), etc. Option 1 and 2 actually seem to work well. I don't know how they will scale. What is the best approach? Is there a portable version 3?

1 0

[general] mingw.org site was hacked
by Jim Bell 04 Mar '11

04 Mar '11

Of interest to MinGW users: It seems that the mingw.org site was hacked a couple days ago: <http://sourceforge.net/mailarchive/message.php?msg_id=27147169>

1 0

Boost Concepts with iterators and ranges
by Robert Ramey 04 Mar '11

04 Mar '11

I've been experimenting with Boost Concepts in conjunction with Boost Range and by implication boost iterators. I have a few questions from the documentation: a) I see BOOST_CONCEPT_USAGE described in some detail in the section Creating Concept Checking Classes. But in the section titled "Reference" I only find in the subsection "macros" BOOST_CONCEPT_ASSERT and BOOST_CONCEPT_REQUIRES. Is BOOST_CONCEPT_USAGE meant to be used? b) "Deprecated Concept Checking Classes" "For each of the concepts documented here, the library includes an identical concept checking class whose name ends in "Concept" For example, in addition to RandomAccessIterator, the library defines a RandomAccessIteratorConcept class template." It sounds like this means that I shouldn't expect to find "WritableIteratorConcept" but in fact in "interator_concepts.hpp" find just that. I'm assuming that's an oversight. c) except for the above, the Range library uses things like ForwardTraversal and ForwardTraversalConcept the iterators library uses ForwardTraversal only. d) the iterators concepts are in the root namespace boost_concepts while the range concepts are in the root namespace boost. e) boost/range/concepts.hpp includes the file boost/iterator/iterator_concepts.hpp All this makes it very hard to keep all this straight in one's head. Is there any possibility that all this could be made simpler to follow by some combination of documentation improvements and/or namespace coordination? Robert Ramey

3 2

[program_options] escaping literal hash symbol for parse_config_file
by Rhys Ulerich 04 Mar '11

04 Mar '11

Hi all, With boost::program_options::parse_config_file, is there any way to escape a literal hash symbol ('#')? I want to set std::string-based option 'foo' to contain the var 'bar#baz', but the sensible-seeming foo=bar\#baz fails in that 'foo' only contains 'bar'. Other variants on quoting and escaping also fail. The source for common_config_file_iterator::get() in config_file.cpp seems to indicate this isn't possible. Can anyone confirm or deny that? Thanks, Rhys

1 0

[boost::property_tree] rapidxml get_index bug for UTF8 ?
by 乔志强 04 Mar '11

04 Mar '11

When use UTF8, non ASCII char is > 127, but char is signed, So get_index() return a big value. char c = -120; get_index(c) VC2010 say: boost::property_tree::detail::rapidxml::internal::get_index<char> returned 4294967168 unsigned int then internal::lookup_tables<0>::lookup_whitespace[internal::get_index(ch)] is error. My patch: inline size_t get_index(const Ch c) { // *** char c (ASCII / UTF8) and wchar_t c: 0 ~ 127 is ASCII char size_t r = c; //******** convert to unsigned // If not ASCII char, its sematic is same as plain 'z' // if (c > 255) //********* ASSCII is 0 to 127 if (r > 127) //******** check r, or check if(c < 0 || c > 127) { return 'z'; } return r; //******** return r } This is boost code: boost_1_46_0\boost\property_tree\detail\rapidxml.hpp template<class Ch> inline size_t get_index(const Ch c) { // If not ASCII char, its sematic is same as plain 'z' if (c > 255) { return 'z'; } return c; } // Detect whitespace character struct whitespace_pred { static unsigned char test(Ch ch) { return internal::lookup_tables<0>::lookup_whitespace[internal::get_index(ch)]; } }; // Detect node name character struct node_name_pred { static unsigned char test(Ch ch) { return internal::lookup_tables<0>::lookup_node_name[internal::get_index(ch)]; } };

2 1

[boost::property_tree] info_parser_read assert in vs2010 when use UTF8
by 乔志强 04 Mar '11

04 Mar '11

When use UTF8, non ASCII char is > 127, but char is signed, So if the value > 127, the char c is < 0, (unsigned)(c + 1) is a big value, read_info() assert in vs2010 when use UTF8 in debug build. std::isspace() call extern "C" int __cdecl _chvalidator( int c, int mask ) { _ASSERTE((unsigned)(c + 1) <= 256); return _chvalidator_l(NULL, c, mask); } My patch is us my_isspace(): template<class Ch> bool my_isspace(const Ch c) { if(c >= 0 && c <= 127) { return isspace(c); } return false; } boost_1_46_0\boost\property_tree\detail\info_parser_read.hpp // Advance pointer past whitespace template<class Ch> void skip_whitespace(const Ch *&text) { using namespace std; while (isspace(*text)) ++text; } // Extract word (whitespace delimited) and advance pointer accordingly template<class Ch> std::basic_string<Ch> read_word(const Ch *&text) { using namespace std; skip_whitespace(text); const Ch *start = text; while (!isspace(*text) && *text != Ch(';') && *text != Ch('\0')) ++text; return expand_escapes(start, text); }

1 0

[uBLAS] Matrix inversion
by Ryan 03 Mar '11

03 Mar '11

I haven't found anywhere in the documentation for a method that does matrix inversion. Does such a method exist but isn't well documented? Ryan

6 19

Re: [Boost-users] [MSM] test if state machine is in a given state /const get_state
by Christophe Henry 03 Mar '11

03 Mar '11

> I considered to use the concept of flags, but then rejected it, simply > because it adds another layer of abstraction, which is great in certain > cases, but also increases the overall complexity. > > In my case i need hardly any properties which span over more than one > state, so i prefer to keep thinks simple for now. It's the constant > struggle to have enough, but not to much abstraction... > > But thanks for the hint, i'll keep that in mind. Hmmm as you can provide every state with a specific tag, I don't see why to develop a less generic, less efficient concept which will be hard to make work for submachines. Am I missing something? >>> 2) Is a const version of "get_state" possible ? And how? >> >> Sure. My mistake. Will be fixed in 1.47. Would you mind opening a >> ticket so I don't forget it? > > done, ticket #5253. Now i'am looking forward to 1.47 ;-) Thanks. Regards, Christophe

1 0