September 2005 - Boost-users

[BGL] detect cycles in a directed graph
by Giulio Veronesi 22 Sep '05

22 Sep '05

Hi, This is my second post for the same problem. Excuse me! I think that the problema can be resolved using the EventVisitor Concept but I don't understand how use it. I would be extremely grateful for any assistance in the following problem: "get all the edges of a cycle". Thanks a lot in advance, Giulio

2 2

[Boost.Signals] One-shot slot?
by Simon Buchan 22 Sep '05

22 Sep '05

What would be the simplest way to make a slot that, once signaled, would disconnect itself? I thought that boost::signals::trackable & 'delete this' would do it, but it seems to be only for boost::bind functors... I am trying: my_slot* foo = new my_slot(...); connection c = sig.connect(bind(foo)); foo->my_connection = c; (vastly simplified, that probably wouldn't compile) and have my_slot 'delete this' in it's op() and 'my_connection.disconnect()' in it's d'tor, but this seems very inelegant, particlarly as this is in a mutli-threaded environment (I know signal isn't thread- safe, but some boost::mutex's normally do the trick) Should I try using a Combiner to weed out slots marked 'once only' or something? Many thanks in advance!

1 0

Boost.Graph can a graph that is a directedS be safely converted to a bidirectionalS?
by Elisha Berns 22 Sep '05

22 Sep '05

Hi, I have a Boost.Graph object that needs to model as closely as possible a DAG in order to be a dependency graph. As such it is an adjacency_list with a directedS policy. All works fine, except that it also needs to work as the inverse of a dependency graph: it needs to find the vertices for a given vertex V that are the 'dependents' (?) of V, that is to say that all the found vertices depend upon V, not that V depends upon them. So I assume that to search for the dependents of vertex V I need to get all the in-edges of V, and the in-edges of all those in-edges recursively? This is apparently feasible by using a bidirectionalS policy for the graph. But my questions are the following: 1) Is there some other better way to do this, such as another graph traversal algorithm? 2) What's the computational expense, in general, of using a bidirectionalS graph over a directedS graph in terms of speed and memory? 3) In the event that there is no better way and a bidirectionalS graph is much fatter and slower than a directedS graph does it make sense to construct the bidirectionalS graph only when needed? Since I only need this when its requested, not as a permanent feature of the application, I could copy the entire directedS graph to create some other graph for this purpose. But here the graph size can vary from a few hundred vertices to several tens of thousands of vertices, so if there is a way to avoid this its probably best. Thanks for any insight here, Elisha Berns e.berns(a)computer.org tel. (310) 556 - 8332 fax (310) 556 - 2839

2 3

Boost.Tuple extension *style* question
by Alan M. Carroll 21 Sep '05

21 Sep '05

2 1

ICU vs Boost Regex
by louis 21 Sep '05

21 Sep '05

I was just looking at the ICU code from IBM and noticed they have a regex engine. Does anyone have experience with this regex engine and how does it compare to the one in the boost library? Any feedback would be greatly appreciated. Thanks Louis

1 0

list of optional dependencies for building boost?
by Benjamin Rutt 21 Sep '05

21 Sep '05

I am new to boost and have a question on building it from source. I built it once using the command: bjam --with-python-root=/home/rutt/python-2.4 -sPYTHON_VERSION=2.4 --prefix=/home/rutt/boost install So you can see that I at least resolved the optional dependency on Python. I also noticed that it warned me that it could not find the bzip2 or gzip packages (I can't remember which). At the end of the build, it mentioned something about a number of errors it had in building, yet it did install the boost libraries. So it seems that the build process continues on past failures (something like 'make -k'). I guess those "failures" must have been unmet optional dependencies? My intention is to build a fully capable boost library and have zero errors and zero unmet dependencies. Is there some documentation or method to obtain information about all optional dependencies boost can pick up at build time? I would like to install them all, to ensure a clean, zero-error build and know that I have a fully functional boost. Thank you, -- Benjamin Rutt

2 3

Re: [Boost-users] inheriting from boost::thread
by Simmons, Aaron 21 Sep '05

21 Sep '05

I write a lot of cross-platform code that's used from JNI. In JNI, you should tell the JVM about native threads that are running. Because Java threads are named, when you attach a native thread to the JVM you have to specify a name. It occurred to me that I could write a "jnithread" class that inherits from boost::thread. It would take a thread-name argument. Before the threadfunc runs it would attach the new thread to the JVM, and when the threadfunc ends it would detach from the JVM. It's relatively easy to write such a class, and would make life easier for thread creation within JNI. However, I hesitate because of the lack of a virtual destructor... -----Original Message----- From: boost-users-bounces(a)lists.boost.org [mailto:boost-users-bounces@lists.boost.org] On Behalf Of me22 Sent: Wednesday, September 21, 2005 1:27 PM To: boost-users(a)lists.boost.org Subject: Re: [Boost-users] inheriting from boost::thread On 21/09/05, Simmons, Aaron <a-simmons1(a)ti.com> wrote: > > I've been thinking about writing a class that inherits from boost::thread. > However, I noticed that the destructor isn't virtual. Shouldn't it be? > boost::thread is designed neither as a base class nor to be used polymorphically. Why do you want to inherit from a thread? You still couldn't get at the private thread handles and there are no virtual functions to modify behavior. - Scott _______________________________________________ Boost-users mailing list Boost-users(a)lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

2 1

boost::test 1.33 compile problem on VC7.0+STLport (1.32 fine)
by Steve Streeting 21 Sep '05

21 Sep '05

Hi, I just upgraded to Boost 1.33 and a set of my unit test routines written against boost::test 1.32 have suddenly started reporting compile errors. I can trigger the issue very simply: --- #include <boost/test/included/unit_test_framework.hpp> #include <boost/test/unit_test.hpp> using namespace boost::unit_test; test_suite* init_unit_test_suite( int argc, char* argv[] ) { test_suite* test = BOOST_TEST_SUITE( "tests" ); } --- In Boost 1.32 this works fine, in Boost 1.33 it reports: --- c:\temp\boost_1_33_0\boost\test\impl\framework.ipp(305) : error C2440: 'initializing' : cannot convert from 'long' to 'boost::unit_test::for _each::static_any_t' Reason: cannot convert from 'long' to 'const boost::unit_test::for_each::static_any_base' No constructor could take the source type, or constructor overload resolution was ambiguous c:\temp\boost_1_33_0\boost\test\impl\framework.ipp(305) : error C2143: syntax error : missing ')' before '=' c:\temp\boost_1_33_0\boost\test\impl\framework.ipp(305) : error C2143: syntax error : missing ';' before '=' c:\temp\boost_1_33_0\boost\test\impl\framework.ipp(305) : error C2059: syntax error : ')' c:\temp\boost_1_33_0\boost\test\impl\framework.ipp(305) : error C2143: syntax error : missing ';' before '{' c:\temp\boost_1_33_0\boost\test\impl\framework.ipp(305) : error C2181: illegal else without matching if c:\temp\boost_1_33_0\boost\test\impl\framework.ipp(305) : error C2440: 'initializing' : cannot convert from 'long' to 'boost::unit_test::for_each::static_any_t' Reason: cannot convert from 'long' to 'const boost::unit_test::for_each::static_any_base' No constructor could take the source type, or constructor overload resolution was ambiguous --- .. and so on (hits upper error limit). I've just reverted back to 1.32 for now, but can you tell me why this happens? Many thanks Steveboost::test 1.33 compile problem on VC7.0+STLport

2 1

Re: [Boost-users] inheriting from boost::thread
by Sohail Somani 21 Sep '05

21 Sep '05

-----Original Message----- From: boost-users-bounces(a)lists.boost.org [mailto:boost-users-bounces@lists.boost.org] On Behalf Of Simmons, Aaron Sent: Wednesday, September 21, 2005 10:29 AM To: boost-users(a)lists.boost.org Subject: [Boost-users] inheriting from boost::thread I've been thinking about writing a class that inherits from boost::thread. However, I noticed that the destructor isn't virtual. Shouldn't it be? Your problem might be better solved using containment

1 0

inheriting from boost::thread
by Simmons, Aaron 21 Sep '05

21 Sep '05

I've been thinking about writing a class that inherits from boost::thread. However, I noticed that the destructor isn't virtual. Shouldn't it be? Thanks, aaron

2 1