January 2010 - Boost - lists.boost.org

Re: [boost] [Boost-users] GGL Review Results
by Christophe Henry 13 Jan '10

13 Jan '10

Hi all, I'm happy to hear this result! Thank you Dave for your support and for taking on the very hard job of managing the review. I also want to thank Franz Alt, David Bergman, Michael Caisse, Barend Gehrels, Darryl Greene, Juraj Ivancic, Erik Nelson and Kenny Riddile for taking the time to write the mentioned thorough reviews and all those who supported the library. The development will of course continue and I hope that many people will further provide me with more valuable nuggets. I used the time after the review to continue the development. You will find in the sandbox a few new features: - increased dispatch speed and other optimizations. - removed some SFINAE to help poor VC8 ;-) - new eUML syntax as discussed. You can now write: target == source + event [guard] /action or source + event [guard] /action == target. - base events: if eventA is the base of eventB, you can use eventB for transitions defined for eventA (doc not updated yet) - internal transitions (first syntax). More to come in the next days. Next, as required, I'll rewrite the doc, add the reference and some UML tutorial. So, as Dave wrote, a lot of work is waiting :) Thanks again to all of you!

1 0

Re: [boost] [msm] scalability
by Christophe Henry 13 Jan '10

13 Jan '10

>> If one is ready to give these ones up, I think we can have pretty >> easily something running without speed loss (I'd even bet on a speed >> gain). All it'd take would be the MPL book example cooked with the MSM >> sauce. >That'd be great. Let me know how I can help. > >Jeff First of all, what would really help would be if you (and any other interested) could list me the main UML features you use so that we can define what the mini backend must provide in addition to what the book example does.

1 0

[uuid] status?
by Vladimir Prus 13 Jan '10

13 Jan '10

I notice that uuid is still not merged to the release branch. What is missing and can the merge be performed in next 24 hours? - Volodya

4 9

[MSM] Meta State Machine Library Accepted
by David Abrahams 13 Jan '10

13 Jan '10

I'm pleased to announce the acceptance into Boost of the Meta State Machine library by Christophe Henry The library got seven reviews, all thoughtful and in-depth, and was unanimously accepted by all reviewers (conditionally, by three of them). Reading through the reviews (not to mention all the other commentary) was truly an overwhelming experience for me. The thoroughness and depth of knowledge of the reviewers seemed to be a cut well above what I'm used to seeing. That, along with the unanimity of their votes, and Christophe's good-natured responsiveness to questions and criticism during the review, make me very confident in accepting MSM. The reviews contain scads of feature requests, ideas, and other valuable nuggets for Christophe to work on, so many that nobody other than Christophe would find a full enumeration useful. So, sorry Christophe, I'm going to leave it to you to troll for details in the message archives. [I suggest taking a pass over all issues mentioned and turning them into Trac tickets. Maybe you can get the review authors to do that themselves for their own reviews.] However, I can summarize some broad themes and issues of general interest to Boosters (and a few things I feel strongly about but won't demand): * There was universal satisfaction with the performance of state machines generated with MSM. * Though the tutorial was generally praised, every conditional acceptance was predicated on improving the documentation, in particular on providing a library reference. This really must be done. There were also several requests to break the tutorial into multiple pages. * One reviewer had problems reading the tutorial because it relies so heavily on UML terms-of-art. I agree with him that people only familiar with FSMs ought to be able to read the tutorial as well. That may mean including documentation of some UML concepts in the library docs. * eUML is really cool but it is apparently still in flux; for now it should probably be labelled as an experimental feature subject to change. * SimpleStates do not support internal transitions. They should. * The name is a bit misleading. Now would be the time to come up with something better if you're going to do it. StateChart is called StateChart in part because of my insistence that any library called Boost.FSM should be able to implement small, tight state machines appropriate for embedded systems use. Even though this library has those properties, I do not know whether the name would be appropriate, for several other reasons I won't list but that you, dear reader are probably thinking of right now. Thank you to Christophe and all the reviewers for restoring my faith in the Boost review process. And of course, congratulations to Christophe. Now the fun is over and the hard work begins... :-) -- Dave Abrahams Meet me at BoostCon: http://www.boostcon.com BoostPro Computing http://www.boostpro.com

2 1

Release branch closed
by Vladimir Prus 13 Jan '10

13 Jan '10

Boosters, The release branch is now closed for all commits, except by explicit permission of release managers. -- The Release Managers, Beman Dawes Daniel James Eric Niebler Rene Rivera Vladimir Prus

5 12

Re: [boost] [msm] eUML guard/action location
by Christophe Henry 13 Jan '10

13 Jan '10

> Cristophe, I could definitely help out with macros and use of PP for MSM, if you need that help. Unfortunately, the next few weeks (at least) are a bit too hectic. >But, now you know that I might be willing (if not currently available) :-) > > /David I looked at your abstract fsm example and I easily believe you :) I happily and thankfully accept your offer. Just contact me when you have time and we can work this out. Thanks a lot! Christophe

1 0

Re: [boost] [msm] eUML guard/action location
by Christophe Henry 13 Jan '10

13 Jan '10

> IMO making it easy for the user to declare these things (maybe with a > macro) and documenting that as the normal approach, dropping all the > parens, could be a big improvement. Unfortunately, I'm pretty bad at macros (and try to avoid them as much as I can) and after spending quite some time at it, I don't come further. Ok, for states, it's trivial: BOOST_MSM_STATE((Empty_Entry(),Empty_Exit()),Empty_,empty) Where Empty_ is the state type name and empty the instance for use in the transition table. Implemented with: # define BOOST_MSM_STATE(expr,state_name,instance_name) \ typedef BOOST_TYPEOF(build_state(expr)) state_name;state_name const instance_name; Where it gets slightly more "fun" is when I want to also remove the parens for the actions/guards. For example, if I have an action called start_playback which is: struct start_playback : euml_action<start_playback> { template <class FSM,class EVT,class SourceState,class TargetState> void operator()(EVT const& ,FSM&,SourceState& ,TargetState& ) { cout << "player::start_playback" << endl; } }; I want to use the same principle. start_playback_ will be the type and start_playback the instance. I come to: BOOST_MSM_ACTION((start_playback_,start_playback, (({ template <class FSM,class EVT,class SourceState,class TargetState> void operator()(EVT const& ,FSM&,SourceState& ,TargetState& ) { cout << "player::start_playback" << endl; } })) )) I thought of using PP tuple: // the action type #define BOOST_MSM_ACTION_TYPE(x) \ BOOST_PP_TUPLE_ELEM(3,0,x) // the action instance for use in the stt #define BOOST_MSM_ACTION_INSTANCE(x) \ BOOST_PP_TUPLE_ELEM(3,1,x) // the action code #define BOOST_MSM_ACTION_CODE(x) \ BOOST_PP_TUPLE_REM_CTOR(1,BOOST_PP_TUPLE_ELEM(3,2,x)) # define BOOST_MSM_ACTION(tuple) \ struct BOOST_MSM_ACTION_TYPE(tuple): euml_action<BOOST_MSM_ACTION_TYPE(tuple)> BOOST_MSM_ACTION_CODE(tuple) ; BOOST_MSM_ACTION_TYPE(tuple) const BOOST_MSM_ACTION_INSTANCE(tuple) ; But I fail at getting the action correctly so I generate: struct start_playback_: euml_action<start_playback_> ({ template... // notice the extra parens I hoped that REM_CTOR would do the job but no. Any idea if it is possible?

2 1

[msm] scalability
by Christophe Henry 13 Jan '10

13 Jan '10

Hi Simon, > I think it would be worthwhile talking a little more about compile time performance in the documentation. You're right, I could explain more in the documentation. Unfortunately, the answer is not straightforward. The compilation time (and how far the compiler plays with us) depends on a number of factors: 1) the compiler 2) the compiler version 3) even your hardware (I get very different compile times on 2 different machines with the same processor) 4) the state machine structure and the chosen front-end 1) usually, g++ crashes later than VC and needed fewer workarounds 2) Surprisingly, g++4.2 > g++ 4.4. VC10 > VC9 > VC8 3) you will need RAM, especially with VC8 4) the compile-time is proportional to the number of transitions in the transition table and to the depth of the state machine (meaning submachine of submachine of...). EUML will cost you much more with VC than g++ as it implies a metaprogramming layer above the standard metaprogramming of the standard front-end and because the underlying decltype/typeof seems much better implemented with g++ (if you don't count the generated code size). A small test with VC9 in Release mode (and no other code) gave me: 20 transitions => 11s 30 transitions => 16s 40 transitions => 22s 50 transitions => 31s This is just to give you an idea as it can greatly depend on the factors named above. > Also I'd like to see some suggestions of how to improve compile times (and how factors including the number of entries in a transition table affect the compile time). There is just one solution: reduce the number of transitions. The best way to do it is using orthogonal regions. For example, if you see in your diagram many transitions to a single state based on the same event, you should factor them all into a second region. Adding submachines could help in some cases. On one hand, the compiler now has to generate code for a second machine, OTOH it reduces the number of transitions in the main one. This will only help if your submachine handles a small number of events, as Msm will add a row in the main table for every event of the submachines. If you can in exchange reduce the number of transitions, it can help. > Is there provision to split a machine across multiple compilation units (I assume not)? You unfortunately assume correctly, as what costs compile-time really is the heavy metaprogramming. > You include instructions for how to increase the maximum size of an MPL vector, but this only works up to 50 entries. What do we need to do to increase this further? This is pretty simple, we need to: - add a vector60/70/80....hpp in mpl/vector. - add a map60/70....hpp in mpl/map. - if more than 50 states, add vector60/70....hpp in fusion/vector This is pretty simple. If that helps, I can add them into the Msm sandbox as I already wrote them. We could also request a feature in the MPL. You will however not manage to increase the size of the transition table forever. At some point will almost every compiler crash in a big cloud of dark smoke. To give you an idea, I usually manage: - 60-80 transitions max with VC9. The exact point is variable on a number of different factors which are hard to understand. - probably less with VC8. VC8 will also quite fast need > 2GB RAM. - 80 transitions with g++4.2. It could probably manage more but I ran out of RAM. - 50-60 with g++4.4. Why 4.4 is in this case less good than 4.2 is a mystery to me. - In Debug mode, you'll see the compiler give up faster - Euml will also have an influence. > Euml looks really cool, but I'm getting an ICE when using it with 36 transitions in EumlSimple.cpp (12 and 24 work fine though) VC8 only has limited support for eUML for 3 reasons: - less than perfect metaprogramming capabilities - bad support for SFINAE - VC problems with decltype/typeof So yes, I would advise eUML for VC8 for only small state machines and without using complicated functors as actions/guards. VC9 would work much better, VC10 better than VC9 too (so Microsoft seems to be working on it). >One of my favourite features of boost::statechart is state local storage. It looks like this would have to be done manually in msm? Yes, this is not supported yet. However, Msm has always been developed very user-oriented and what users ask, they usually get it. Msm has from the beginning greatly gained from clever people suggesting new features and I will not stop the development after the review, whatever the outcome is. If you'd like a feature, I'd be happy to discuss a solution with you and implement it in the next version, thus making Msm more useful. > The library is looking really powerful and expressive (not to mention fast!) for small state machines. Thanks. This is the reason for the compile-times ;-) Christophe

5 5

Re: [boost] Meta State Machine (MSM) review
by Christophe Henry 12 Jan '10

12 Jan '10

Hi Barend, Sorry that I didn't have time to address earlier the questions in your review. >I would like to see a connection with Boost.Graph (I mean, see it in an >example or so, not within the kernel), to see what I'm doing with all >those transitions; I asked this on the list and this is possible, good. It is a good point and I'll try to provide this, at least for simple state machines. >1) the struct "Row" is written with a capital, this is unusual within >Boost. Words with a capital are normally template parameters. I >understand why it is done (there was already a row without a capital in >Msm 1.1), but I think it could have been solved differently (for >example: "transition", or if that is too long "tr" which could stand for >transition row or table row). I don't like to see Row where I expect to >see a normal classname. Conventions are always hard to define. I usually try to follow the following convention: - rows are usually not capitalized (a_row, g_row...) - except the row working with functors, like Row - eUML functors are also capitalized and finish with an underscore. - eUML functions are not capitalized and finish with an underscore. This tries to match the phoenix syntax. This allows differentiation between functors/non functors. I found this system practical but would not mind changing it if required. >The capitals are also used in the documentation and samples (struct >Empty) but there I've less problems with it. Maybe use of Capitals for >states is a state machine convention Yes, it is a common habit in UML to capitalize state names (though I am not aware of any written rules). >I noticed different conventions in >the source files. Many template parameters are completely capitalized. I >had done the same thing, earlier, but was hinted that this is unusual as >well. Just a matter of style (as long as library users normally don't >see it, it is less important) Yes, while the user doesn't see it, it is not very beautiful so I'll change this. >2) the class state_machine is quite long (in lines of code), containing >several sub-structs which maybe (don't know) could be moved to a >separate source file (detail). Yes, it became quite big, you're right. It's quite practical for me to have functions and helpers at the same place but some parts could be separated. >You have to call p.start() manually Yes. Calling start() will move the machine in its initial state and call on_entry on the machine and the initial state. You have to call it explicitly because sometimes you simply don't want on_entry to be called at the creation of the state machine. >Thanks for submitting this excellent library. I'm happy that you like it. Thanks for the great work you put on the review. Regards, Christophe

2 1

[Polygon] "Ambiguous use of partial specialization" in CodeWarrior
by Kef Schecter 12 Jan '10

12 Jan '10

I'm using a special version of CodeWarrior for an embedded system, but I suspect that other versions of the software will show the problem. The problem is, basically, the compiler hates polygon_traits<T>. Here's an example from polygon_90_set_traits.hpp: template <typename T> struct traits_by_concept<T, polygon_90_concept> { typedef polygon_traits<T> type; }; It chokes on the typedef, saying "Ambiguous use of partial specialization". This is the complete text of the error message; it provides no other information. Most of the surrounding definitions are very similar, but only the ones that use polygon_traits<T> -- instead of, say, rectangle_traits<T> -- raise errors. Here's another example from polygon_traits.hpp: template <typename T> struct view_of<rectangle_concept, T> { typedef typename polygon_traits<T>::coordinate_type coordinate_type; [snip] }; It chokes on the typedef and any code that tries to use the typedef. The same problem happens with all the other view_of functions and a couple of others, but all the problems seem to deal with the use of polygon_traits<T> in a template. Unfortunately, I know very little about partial specialization, traits classes, or any advanced features of generic programming in C++. My understanding of templates is limited to stuff like making basic containers, generalized sorting functions, etc. -- the easy stuff. So I find the definition of polygon_traits too complicated and am unable to fully comprehend the essence of the problem, let alone how to fix or work around it. Since I need to get this to work, if a proper solution is not obvious, I'm willing to accept any kind of clumsy hack, including disabling parts of the library that I might not need. (I use very little of the Polygon library, but I have a hard time finding an alternative to the part I do use -- namely, finding the union of polygons while allowing for holes. Alternative suggestions in that area are welcome, but I can't use GPLed code, or code that I would need to pay a license for to use commercially.) Any ideas? - Kef

4 16