December 2004 - Boost - lists.preview.boost.org

[Boost-bugs] [ boost-Support Requests-1091151 ] ublas extendability1
by SourceForge.net 25 Dec '04

25 Dec '04

Support Requests item #1091151, was opened at 2004-12-25 10:18 Message generated for change (Tracker Item Submitted) made by Item Submitter You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=207586&aid=1091151&group_… Category: None Group: None Status: Open Resolution: None Priority: 5 Submitted By: Nobody/Anonymous (nobody) Assigned to: Nobody/Anonymous (nobody) Summary: ublas extendability1 Initial Comment: Hi. I am using ublas with types other then built in scalars (like "double" or "int"). Typical jobs: prod(matrix<double>,vector<v3d<double> >) or triangular_solve(matrix<double>,vector<v3d<double> >). In order to be able to do it, I have made partial specialization of "promote_traits" template in namespace "boost::numeric::ublas" for my classes. This works. Is this technique good ? Or uBLAS intended only for built in types ? See attached files. Is it good to intervene to ublas namespace, like done in file "vnd.h" at the end. ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=207586&aid=1091151&group_… ------------------------------------------------------- SF email is sponsored by - The IT Product Guide Read honest & candid reviews on hundreds of IT Products from real users. Discover which products truly live up to the hype. Start reading now. http://productguide.itmanagersjournal.com/ _______________________________________________ Boost-bugs mailing list Boost-bugs(a)lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/boost-bugs

1 0

Lambda Question
by Brian Braatz 24 Dec '04

24 Dec '04

Is is possible to, given a functor generated as a result of a lambda bind expression, pull the return type from the lambda functor? (still digging through the lambda bind code and have not found the answer to this yet. I sent a previous post with sample code that shows what I am trying to do) Any help or direction would be greatly appreciated. Thanks Brian

4 3

RE: [boost] Re: Lambda Question
by Brian Braatz 24 Dec '04

24 Dec '04

> -----Original Message----- > On Behalf Of Arkadiy Vertleyb > "Brian Braatz" <brianb(a)rmtg.com> wrote > > > Is is possible to, given a functor generated as a result of a lambda > > bind expression, pull the return type from the lambda functor? > > I am wondering if "result_of" may be something that you are looking for. > OTOH, I don't really know what the status of this library is. > > Regards Thank you for your thoughts- result_of looks like what I want - but I am having difficultly figuring out what the lamda factor LOOKS like after it is compile time generated. (I am (very very) slowly dissecting lambda bind to try to figure this out). IF ANYONE knows the answer to how you would GET TO result_of syntax wise help would be appreciated. (previous mail with sample code below) -----Original Message----- From: boost-bounces(a)lists.boost.org [mailto:boost-bounces@lists.boost.org] On Behalf Of Brian Braatz Sent: Wednesday, December 22, 2004 8:03 AM To: boost(a)lists.boost.org Subject: [boost] Help \ Advice requested on getting the return type from aLambda fctor I wrote up the simple example that shows what I am trying to figure out Basically, I am trying to pull the deduced return type (or explicitly passed as a template param in lambda bind()) in my template. If I have a function that is receiving a functor, I would like to be able to take that functor and pull the RETURN TYPE out of it. Lets say I have a function that returns a string: std::string ReturnAString() { return "this is a string"; } And then here is my template function (code below does not compile) template < typename TFUNCTOR> TFUNCTOR::result_type CallFctorDeduceRet( TFUNCTOR & functor) { return functor(); } Now here is my bind call std::cout << ( bind<std::string>(&ReturnAString)); Is there a way to GET to the return type in the Lambda fctor that was generated with a BIND call? I would like to be able to pull the <std::string> from the bind call and use it in the template function. It is not a problem for my template function to be dependant upon lambda bind. I have been going through the lambda code. And while it is brilliant, I am having a bit of difficulty figuring out what I am looking at (grin). Any help or direction is greatly appreciated FULL SAMPLE CODE HERE: template <typename CURRIEDTYPE> // Id class for currying type without creating the object class Id { public: typedef CURRIEDTYPE TYPE; }; // Explictly pass IN a return type template < typename TFUNCTOR, typename TRET> TRET CallFctorPassInRet ( TFUNCTOR & functor, Id<TRET> & ) { Id<TRET>::TYPE TRet; TRet = functor(); return TRet; } std::string ReturnAString() { return "this is a string"; } // try to get return type from the lambda fctor // Below does not compile- but it is along the lines of what I // WANT to do- note that I am trying to get from the Lambda functor the deduced return type template < typename TFUNCTOR> TFUNCTOR::result_type CallFctorDeduceRet( TFUNCTOR & functor) { return functor(); } void Test_LambdaRetDeduction() { std::cout << "CallFctorPassInRet returns [" << CallFctorPassInRet( bind(&ReturnAString), Id<std::string>()) << "]" << std::endl; // below is what I want to make work std::cout << "CallFctorDeduceRet returns [" << CallFctorDeduceRet( bind<std::string>(&ReturnAString)) << "]" << std::endl; }

1 0

[logo] Happy Holydays...
by Rene Rivera 24 Dec '04

24 Dec '04

Boosters, Thought I'd give a tiny present for all... http://tinyurl.com/5kne3 http://boost-sandbox.sourceforge.net/vault/files/Boost%20Logos/Rene%20River… -- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org - grafik/redshift-software.com - 102708583/icq

1 0

Problems compiling Boost.Test example with gcc under Linux
by Noel Llopis 24 Dec '04

24 Dec '04

I suspect the answer to this is going to be really simple, but I've been trying to figure it out for several hours and I'm ready to give up (I couldn't even find a mention of the problem in the mailing list archives). All I want to do is use the Boost Unit Test library with gcc under Linux. I have Boost 1.32 correctly installed (I've used other libraries successfully without any problems). Now I simply try to compile the simplest possible example: http://www.boost.org/libs/test/example/unit_test_example1.cpp This is what I get: gcc -o test boost_test.cpp boost_test.cpp:11: error: `boost::unit_test' has not been declared boost_test.cpp:11: error: expected nested-name-specifier before "test_suite" boost_test.cpp:11: error: `test_suite' has not been declared boost_test.cpp:23: error: expected constructor, destructor, or type conversion before '*' token boost_test.cpp:23: error: expected `,' or `;' before '*' token boost::unit_test has not been declared??? How is that possible? The compiler is correctly including the boost/test/unit_test.hpp file from the first line of the program. What am I missing? It must be something really, really stupid. I can tell :-) Happy holidays everyone! --Noel Games from Within http://www.gamesfromwithin.com

3 3

filesystem lib patch
by François Dumont 24 Dec '04

24 Dec '04

Find attached a patch for the filesystem library. remove and rename symbols are injected within the std namespace before being defined by the inclusion of cstdio header. The patch make the code less Standard library implementation dependent. Francois Dumont

2 1

Some observation re libraries as DLLS
by Robert Ramey 24 Dec '04

24 Dec '04

I'm making adjustments in the build and test Jamfiles to take into account changes made to permit the serialization library to be built and used as a DLL. This as raised a couple of questions in my mind. Re: the Jamfile for test a) It would seem to me that if the test is to be built and run with <runtime-link>dynamic - indicating the usage of C++ runtime dll, that the default build for the test should be to use the dll versions of all the libraries. I've managed to do this in sort of a kludgy way and it might be working. But I'm still working on it. b) the above raises the issue of how one should specifiy a target be built to use the dll vs the static library. <runtime-link>dynamic does something else. I was looking for a variant but I don't find it. I'm thinking that the way its done in the Program Options library is the way to go. c) when using the dll versions of boost libraries, the dll has to be on the path (at least with windows). It seems that bjam takes care of this automatically. So its only a problem when I want to debug the app with my VC IDE. It might be a good idea if the document mentioned this. d) The section on Testing Auto-linking in .../more\separate_compilation.html leaves me extremely confused. If I try to build the test without auto-linking by including the BOOST_???_NO_LIB then the libraries names searched for don't match the onces created by the build Jamfile. That its the above define doesn't get propogated to the build Jamfile and I don't know how it would. If I alter my Jamfile so that it doesn't include common-variant-tag as a target. The when the bjam stage is invoked, the libraries won't have the version specfic stuff. The situation is made even more compilicated by the fact that I have one dll (boost_wserialization.dll) which calls into another one (boost_serialization.dll). I've made it work by using the variant tags all the time - that is never using BOOST_???_NO_LIB. But now I'm concerned the the section at the end of the reference which seems to suggest that this won't work on platforms which don't suppor t auto-link. This seemed very simple when I started, but now I'm trying to get the details right and I'm having lots of nagging issues like the above. Its actually harder to figure out than the serialization library itself. Robert Ramey

1 0

mpl::inherit_linearly reuse to produce loki's tuple?
by Larry Evans 24 Dec '04

24 Dec '04

Could inherit_linearly be used to create something like loki's tuple? The name, inherit_linearly, and a very brief look at the code suggests it does something like loki's GenScatterHeirarchy, which was used to create the loki tuples as shown in: http://cvs.sourceforge.net/viewcvs.py/loki-lib/loki/Reference/HierarchyGene… If so, then I was wondering if it would be worth it to reuse inherit_linearly to implement boost's tuple?

3 14

typeof: what to register?
by Arkadiy Vertleyb 23 Dec '04

23 Dec '04

Hi all, Here is a question: besides primitive types, which types/templates should be registered as a part of typeof library? The first candidate is STL. The containers, functors, binders, etc. The STL iterators is a separate question. In most STL implementations they are typedefs, and therefore their template names are non-standard and different. In Dinkumware they are nested classes, and so are the subject to non-deduced context, which prevents them from being registered in a trivial way. Peder does have a solution for this, based on a bug in VC7.1 that allows to work-around non-deduced context (I am personally against using this, although I have to admit that it does the job). So, if we decide to put #ifdefs, based on STL implementation, and *if* we decide to use Peder's solution, it looks like we can take care of all the standard library (including iterators) on the typeof lib level. I have to note here that, if we decide against iterator registration, iterator instantiations can be always registered as needed, by the end user: BOOST_TYPEOF_REGISTER_TYPE(MyClass) BOOST_TYPEOF_REGISTER_TYPE(std::vector<MyClass>::iterator) I think if other libraries want to be supported by typeof (like Spirit, for example), those libraries should register their types/templates themselves (just because of synchronisation issue). Any help would be provided, of course. I would greatly appreciate any opinions on the subject. Regards, Arkadiy

2 2

boost::noncopyable and DLL usage
by Roland Schwarz 23 Dec '04

23 Dec '04

The thread class derives from boost::noncopyable. While this works for static library builds, there are some ugly warnings when compiling user programs that use the DLL version of the threading lib. I suspect noncopyable not having the appropriate __declspec(export) in a DLL being the culprit. I suspect any library will show this behaviour when doing a DLL build. Is this effect known? Are the warnings avoidable? Thank you, Roland

2 5

Status of VC++.NET 2005 Beta
by Loïc Joly 23 Dec '04

23 Dec '04

Hello all, I would like to know if anybody has successfully built boost with VC++.NET 2005 beta version ? I tried to look the regression tests, but the links from boost web site are broken (This site is defaced!!! NeverEverNoSanity WebWorm generation 25.). By the way, is there any document that explains how to create a new toolset for boost::bjam ? Regards, -- Loïc

3 3

Re: [boost] [gui] rectangles and the CSS box model
by brangdon＠cix.compulink.co.uk 23 Dec '04

23 Dec '04

In-Reply-To: <cqbhda$lq7$1(a)sea.gmane.org> msclrhd(a)hotmail.com (Reece Dunn) wrote (abridged): > However, for efficiency, it is a good idea to expose the > members of point and size. Is that because your compiler cannot inline member functions? -- Dave Harris, Nottingham, UK

1 0

X-Platforming Boost GUI
by Alan Gutierrez 22 Dec '04

22 Dec '04

I've spent quite a bit of time getting this Boost GUI code to build anywhere, but now it builds on both OS X and W2K. I'm using XCode to build on OS X, Jamboost is working on W2K. I'm feeling like I might as well add Linux to the mix, and work toward having an early integeration across these three platforms. It would teach me a lot about Boost and bjam. -- Alan Gutierrez - alan(a)engrm.com

3 3

[gui] rectangles and the CSS box model
by Reece Dunn 22 Dec '04

22 Dec '04

Hi All, In the "thoughts on a GUI component library" thread there have been various discussions about the layout of a rectangle and come relating to CSS. Note that point and size are stable. The CSS3 Values and Units module (http://www.w3.org/TR/2001/WD-css3-values-20010713/) defines various units (e.g. inches, picas and em). This is a good basis for a GUI library metric type. The type should store the values as a floating point to allow things like 2.71cm and also to support platforms that store coordinates as floating points (e.g. Cocoa). The CSS3 relative values (em and ex) are dependant on the inherited font-size property, so it is difficult to use these units as there is no associated font-size property. Another issue relating to this is how to resolve metric values between units that are not directly convertable. I have suggested using a device::resolve method that takes metric, point, size and rect types, but it may not always be possible to get access to a device. A further complication is the location of the origin. If we treat the origin as being in the top left corner (which is true for Windows, HTML/CSS and other document models such as text editors) this will simplify the design and for platforms that have a different model (e.g. Mac) then the implementation on that platform will need to map between the different representations. I personally like the CSS box model (http://www.w3.org/TR/css3-box/) which can be defined as: struct css::box_model { rect margin; rect border; rect padding; size dimensions; // width and height // T content; }; The problem with supporting it by default is that we would need: class component: public css::box_model; class graphics::object: public css::box_model; This clearly violates the don't pay for what you don't need rule of C++, so for a css_block_layout, it needs to contain the box model properties and a pointer to the component/graphical object. Obviously some level of support for the CSS box model should be provided, allowing an implementation of it to be written. There has been some discussion and disagreement as to what representation a rectangle takes. It is a good idea to hide the representation of a rectangle and provide accessors to the various values. This is because different people will want different things from it. However, for efficiency, it is a good idea to expose the members of point and size. In supporting the CSS box model and adoption of top-left as (0, 0) the representation becomes: class rect { point tl; point br; }; with: top() = tl.y; left() = tl.x; bottom() = br.y; right() = br.x; width() = br.x - tl.x; width( w ): br.x = tl.x + w; size() = ::size( width(), height()) center() = tl + ( size() / 2 ) An issue with this is: do we keep the width of the rectangle when setting left/right? In order to keep things simple, the answer has to be no. Construction of a rect should have the following variants: rect() = rect( 0, 0, 0, 0 ) rect( n ) = rect( n, n, n, n ) // This models the CSS "margin: 2px;" behaviour. rect( t, r, b, l ): tl( t, l ), br( b, r ) // This models "margin: 1pt 2pt 3pt 4pt;" rect( point p, size s ) = rect( p.x, p.y, p.x + s.dx, p.y + s.dy ) // This allows for support for traditional GUI representations Note that this model will change the nature of a program that uses the current model. We also have inflate/deflate operations that should take size *and* rect variants. This will allow easier box-model calculations: rect disp( point( 0, 0 ), size( 500, 700 )); rect margin( 1, 2, 3, 4 ); rect box = disp; box.deflate( margin ); // apply margin adjustments // ... Regards, Reece

3 8

Re: [boost] mailing lists archives stuck
by JOAQUIN LOPEZ MU?Z 22 Dec '04

22 Dec '04

----- Mensaje original ----- De: Douglas Gregor <dgregor(a)cs.indiana.edu> Fecha: Miércoles, Diciembre 22, 2004 9:02 pm Asunto: Re: [boost] mailing lists archives stuck > On Dec 22, 2004, at 7:22 AM, Joaquín Mª López Muñoz wrote: > > To whom it may concern, the archives of the mailing lists, > > (for instance, http://lists.boost.org/MailArchives/boost/bydate.php) > > have stopped collecting messages since almost a week. > > The archives should be working again now. Thanks for the report! > Thanks for fixing the thing :) BTW, the *users* mailing list is also stuck :( As for the announce list, ist' hard to know, because its volume is so low. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

1 0

mailing lists archives stuck
by Joaquín Mª López Muñoz 22 Dec '04

22 Dec '04

To whom it may concern, the archives of the mailing lists, (for instance, http://lists.boost.org/MailArchives/boost/bydate.php) have stopped collecting messages since almost a week. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

2 1

Add conversion op doc to scoped_ptr
by cliffg＠codewrangler.net 22 Dec '04

22 Dec '04

This may already be known / in the queue / in CVS, but in case not ... The scoped_ptr synopsis and members does not document the unspecified-bool-type conversion operator that scoped_ptr provides (I didn't check scoped_array to see if the same issue is present or not.) I found this when a couple members of my team converted code from shared_ptr to scoped_ptr, with the original code relying on the conversion operator, and I was surprised when everything compiled cleanly (without adding the appropriate get() calls). Cliff

1 0

Help \ Advice requested on getting the return type from a Lambda fctor
by Brian Braatz 22 Dec '04

22 Dec '04

I wrote up the simple example that shows what I am trying to figure out Basically, I am trying to pull the deduced return type (or explicitly passed as a template param in lambda bind()) in my template. If I have a function that is receiving a functor, I would like to be able to take that functor and pull the RETURN TYPE out of it. Lets say I have a function that returns a string: std::string ReturnAString() { return "this is a string"; } And then here is my template function (code below does not compile) template < typename TFUNCTOR> TFUNCTOR::result_type CallFctorDeduceRet( TFUNCTOR & functor) { return functor(); } Now here is my bind call std::cout << ( bind<std::string>(&ReturnAString)); Is there a way to GET to the return type in the Lambda fctor that was generated with a BIND call? I would like to be able to pull the <std::string> from the bind call and use it in the template function. It is not a problem for my template function to be dependant upon lambda bind. I have been going through the lambda code. And while it is brilliant, I am having a bit of difficulty figuring out what I am looking at (grin). Any help or direction is greatly appreciated FULL SAMPLE CODE HERE: template <typename CURRIEDTYPE> // Id class for currying type without creating the object class Id { public: typedef CURRIEDTYPE TYPE; }; // Explictly pass IN a return type template < typename TFUNCTOR, typename TRET> TRET CallFctorPassInRet ( TFUNCTOR & functor, Id<TRET> & ) { Id<TRET>::TYPE TRet; TRet = functor(); return TRet; } std::string ReturnAString() { return "this is a string"; } // try to get return type from the lambda fctor // Below does not compile- but it is along the lines of what I // WANT to do- note that I am trying to get from the Lambda functor the deduced return type template < typename TFUNCTOR> TFUNCTOR::result_type CallFctorDeduceRet( TFUNCTOR & functor) { return functor(); } void Test_LambdaRetDeduction() { std::cout << "CallFctorPassInRet returns [" << CallFctorPassInRet( bind(&ReturnAString), Id<std::string>()) << "]" << std::endl; // below is what I want to make work std::cout << "CallFctorDeduceRet returns [" << CallFctorDeduceRet( bind<std::string>(&ReturnAString)) << "]" << std::endl; }

1 0

boost.sourceforge.net hacked?
by Joaquín Mª López Muñoz 22 Dec '04

22 Dec '04

Please take a look at http://boost.sourceforge.net/regression-logs/ ??? Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

7 6

RE: [boost] Unzipping the Boost distro
by Stefan Seefeld 22 Dec '04

22 Dec '04

> From: Richard Peters [mailto:r.a.peters@student.tue.nl] > If some boost authority certifies the released packages, and > users correctly > verify that the certificate is issued by the boost authority, and the > certificate is valid on the package, then certified executables can be > trusted as well. You are right, providing certificates is one thing. Whether or not to distribute executable self-extracting packages is different question. Regards, Stefan

1 0

RE: [boost] Unzipping the Boost distro
by Stefan Seefeld 22 Dec '04

22 Dec '04

> From: Richard Peters [mailto:r.a.peters@student.tue.nl] > Another argument that I thought of this morning: suppose we > do not publish a > self-extracting executable. What is going to stop an attacker from not > uploading his own self-extracting look-alike? If he can > change existing > archives, he probably can add other archives as well. right, but if some boost authority certifies its released packages, everybody is free to ignore such look-alikes. Isn't that the whole point of certification ? Regards, Stefan

2 1

automatic garbage collecting algorithm and its implementation in c++
by yilin yang 22 Dec '04

22 Dec '04

hi, everybody i've just designed and coded a garbage collecting infrastructure named u_canforget. it was uploaded onto the news group at boost.yahoo. the main idea is that we count the reference to track the life-cycle of the given object. as the new operator was overloaded to statisfy this requirement. the other important concept was the reference object. with the syntax represented by the redefined macros, the codes may be written in a very simple way with relationships automatically constructed, objects are freed without delete operater can be realized as well. i would keep tracking this thread, if there are any thought about this algorithm. thx for your advices. your's yilin. --------------------------------- Do You Yahoo!? 注册世界一流品质的雅虎免费电邮

1 0

[gui] Representation of graphical objects
by Reece Dunn 22 Dec '04

22 Dec '04

Hi All, A GUI library can be thought of as having two distinct, but related, units: components (GUI widgets) and graphical objects/operations. A GUI component consists of data and events (e.g. a push button has a title and an onpressed() event). The question is: how do we represent the graphics unit? There are clearly two views: high level and low level. You can represent a graphics library as a series of objects (lines, circles, polygons, images, text, etc.) each of which having a set of properties (color, background/fill color, etc.). This is a high-level view and is based on the vector graphics model. Consider you are writing an application that draws a fractal object on the screen such as a Hilbert curve. You decide to represent it as a list of line objects. What happens when you redraw the curve? For large numbers of lines (easy when you have a fractal depth of 5+), the redrawing process is going to be slowed down. This is because you will be changing the colour on each line drawing operation. This becomes more apparent if you are using images to fill polygons and are using the same fill pattern across several objects. There are various graphics primitives (font, pen, fill (i.e. brush), etc.). Each primitive operates on a canvas and some can be used to query information on a device (e.g. the area occupied by a text string). In Windows a device and canvas are represented the same (as a HDC), but it makes sense to distinguish between them: you draw to a canvas and query information about a device. These graphical primitives should form the core of the graphics unit, with the high-level graphical objects being viewed as an extended unit. With the discussion on using CSS and Andy Little's comment about what difference is there between GUI components and graphical objects in terms of them both being objects that have position information and are drawn onto a canvas, it makes sense to allow both types to share a common interface. The main question is really what happens on the redraw phase? Ideally, components should not be drawn on the ondraw event, but rather be controlled by the behaviour of the operating system. For owner drawn components, do we handle the drawing event here or on the request of the OS (I think the latter is best). What about lightweight components (i.e. components that do not take up OS resources w.r.t. component representation, but that do support an event model)? Should we treat lightweight controls as graphical objects with an event model? Do graphical objects support an event model? Should graphical objects be treated as lightweight components? Regards, Reece

1 0

Unzipping the Boost distro
by David Abrahams 22 Dec '04

22 Dec '04

I just downloaded boost-1.32.0 and started to unzip it with the built-in facilities in Windows XP; the OS reported it was going to take 50 minutes and it was *crawling*. Using cygwin's unzip tool it was done in under a minute. I'm thinking we should stop distributing raw .zip files without an extractor... -- Dave Abrahams Boost Consulting http://www.boost-consulting.com

11 24

regex library design (was: any interest in finite automata-based regex engine?)
by Vladimir Pozdyayev 22 Dec '04

22 Dec '04

John Maddock wrote: > This is an area I want to explore though, if I can get this next lot of code > out the door, then I'll create a cvs branch to experiment with this, if you > want to suggest / experiment with a design for the abstract creator in the > meantime, then go for it. Finally I got down to refactoring the ANFA code. Take a look at http://groups.yahoo.com/group/boost/files/anfa-regex/anfa091.zip (once again, this is not a fullscale regex library... yet) The DESIGN file content is appended to this message. -- Best regards, Vladimir Pozdyayev. ---------------------------------------------------------------------- -= Regex Design Issues =- The core classes. * charset Provides "bool operator( character )". Nothing much to say apart from that, but do see the Charset Issues section. * charset creator Supports arbitrary charset expressions (within the limits of a given set of possible operations). Implementations, however, are not required to provide _all_ the declared functionality; calls for unsupported features should result in appropriate exceptions. Also provides the "void create( charset & )" function which is used to initialize a charset newly created by "matcher creator". The "abstract_charset_creator" class provides stubs for all expression elements possibly to be requested by regex parsers. Implementations with limited functionality can inherit them and redefine only those functions that should actually do something useful. * matcher Provides the low-level matching functionality, say, finding the first occurrence of bla-bla-bla. On the other hand, replacing all occurrences is a high-level action, for it consists of (1) finding them and (2) creating a modified string---so it should go into the "regex" class. (On the other other hand, if it is possible to do replacement on the fly while searching, this becomes a low-level action. I don't know if it can be done in a sufficiently general way, however.) * matcher creator Like charset creator, only for matchers. * parser The syntax parser. Takes the input string in the form of begin-end iterators, and issues a sequence of charset/matcher creator calls ending with "matcher_creator::create" (or an exception). In essence, simply provides the function "void parse( matcher &m, iterator begin, iterator end )". A parser must be consistent with the properties of "creator" classes. * regex A wrapper for the "matcher" class. Provides the high-level creation & utilization routines. How they are connected. A sample from "regex.cpp": typedef basic_regex< basic_simple_parser< basic_charset_creator< wchar_t >, basic_anfa_matcher_creator< basic_charset< wchar_t > >, basic_anfa_matcher< basic_charset< wchar_t > > >, basic_anfa_matcher< basic_charset< wchar_t > > > regex; On "creators" and "create" functions. The name is somewhat misleading, since they fill target objects with compiled data rather than create them. Still, "charset compiler" sounds a bit weird... or does it? Anyway. The "creators" are subject to the following uses and requirements. They must be able to destruct themselves gracefully even if the expression they are being feeded with is only halfway done (in case someone has thrown an "unsupported" exception). The "create" function must do an implicit "pop" from the expression stack, so that the "creator" object could be reused. The "create" function can assume there's only one top-level expression tree node on the stack. -= Charset Issues =- (Should I rename them to "character sets" for consistency with the full-names style?) All the above templates have quite a freedom in intermixing different character types, let alone different character encodings. E.g., the sample program has all regex templates instantiated with wide characters, but the regex string itself is char-typed. This clearly needs to be controlled.

1 0