Hello to be make it easy to read a table from a database and save the result with boost serialization (and to write back back to the database) I have devised a scheme to achieve C++ reflections that allows me to set a field value and get a field value using a string data member name. I am using the fact that in C++ functions can take a reference to a class instance and operate on the class data members, where the functions have hardcoded the datamember names. Since those functions can be declared as static, I simply remember the pointer to my getters and setters in static maps associated with every class instance So when a class instance gets created -- it updates the static maps of getters and setters. I do not like the fact that I have to use a mutex on the function maps. ANd in general wanted to ask if such approach has been implemented in one of the boost libraries or if not ... why not I am attaching a 'brief' of the mechanism. Thank you in advance for any feedback. Vlad ------------------------------------------------------------- struct tField { string name; //other DB-related attributes //order of declaration, isnullable, otltype } struct reflectable { //returns a string name of the field char* (*tF_get_field_name)(void), //assign a value to the field. The value is obtained from functor //that's why we call it 'set' typedef void (*tFset__field)(reflecatble& , CFFunctorAssignVal&) ; //send the field value (const) to somewhere //that's why we call it 'get' typedef void (*tFget__field)(const reflectable&,CFFunctorReceiveVal&) ; typedef map tName2Fields; typedef map tMFset__field; typedef map tMFget__field; //derived classes must have the "static" version of the maps //we inforce it by declaring abstract functions here //also each class will have to provide a static function //that gets returns a pointer to each map }; tempate < typename T, //cpp type tF_get_field_name fname, //function ptr to get name tFset__field assing_functor, //functor pointer to assing val to field tFget__field get_functor, //functor to read val from field tName2Fields* ptrFieldMap, tMFset__field* ptrSetterMap, tFget__field* ptrGetterMap struct rffield { //register a field name (obtained from fname() //with ptrFieldMap //register pointer to getter with ptrGetterMap //register pointer to setter with ptrSetMap } ---- separate header ----------------- struct myclass: public reflecatble { static allTheMaps...; //to simplify I have a macro that takes in field name, cpp type //and declares them and generates getter and setter functions //that accept functors static void field1__getter ( const reflectable&th,CFFunctorReceiveVal& f) { f(th.field1); } static void field1__setter ( const reflectable& th,CFFunctorAssignVal& f) { f(th.field1); } static const char* field1__name (void) { const char* res="field1"; return res; } //single declration causes all the registration to take place //since I am passing parameteres to rffield constructor //via template. It is almost like ability to invoke constructor // right in a header file. rffield field1; } ------------ cpp for my class has the static maps -- ----------- -- V S P toreason@fastmail.fm -- http://www.fastmail.fm - Faster than the air-speed velocity of an unladen european swallow

Some issues: 1. This fails to compile: #include #include #include int main() { } MSVC8, Windows XP, the beginning of output is: ------ Build started: Project: qq, Configuration: Debug Win32 ------ Compiling... qq.cpp c:\boost_1_38_0\include\boost-1_38\boost\range\adaptor\replaced.hpp(116) : error C2146: syntax error : missing ';' before identifier 'make_replaced_range' c:\boost_1_38_0\include\boost-1_38\boost\range\adaptor\replaced.hpp(116) : error C2433: 'replace_range' : 'inline' not permitted on data declarations c:\boost_1_38_0\include\boost-1_38\boost\range\adaptor\replaced.hpp(116) : error C4430: missing type specifier - int assumed. Note: C++ does not support default-int 2. I see indirected adaptor, but I fail to see counted adaptor. With Boost.Iterator, we have both counting and indirecting iterator, these two do opposite things. For example, I want to rewrite this with Boost.RangeEx: #include <numeric> #include <algorithm> #include <iostream> #include <functional> #include int main() { std::cout << std::accumulate( boost::make_counting_iterator(2), boost::make_counting_iterator(7), 1, std::multiplies<int>()) << std::endl; return 0; } 3. push_back/push_front It's not evident from their names, neither from their documentation that they would use "c.insert(where, begin, end)" not a loop of push_back. Suggest clarification in documentation and renaming them to: insert_back / insert_front. 4. Documentation - what about performance, exception safety, thread safety ? In my opinion the documentation should be explicit in what they are depend on the implementation of STL being used and adds no overhead on its own. Each algorithm should contain explicit notice which STL algorithm it wraps, and it would be convenient to link to SGI STL site. 5. I heard that some STL implementation fail to take an advantage of a range of random iterators to reserve() before range insertion. Is it true ? Is it possible to fix with Boost.RangeEx ? ===== Then, my opinion

What is your evaluation of the design? Good design.

What is your evaluation of the documentation? It has some issues mentioned above.

What is your evaluation of the potential usefulness of the library? Very useful. Will reduce a lot of boilerplate code. In particular, it provides the most elegant solution of "for-each-value-in-a-map" problem. I have not used old Range library, because I have not found it that useful.

Did you try to use the library? With what compiler? Did you have any problems? MSVC8, the problem described above. Without Boost.Thread included before Boost.Range it seems to work. Maybe tomorrow (March 2nd) will try on production code.

How much effort did you put into your evaluation? A glance? A quick - reading? In-depth study? A quick reading.

Are you knowledgeable about the problem domain? Been writing production code in C++ for 2 years, using STL and Boost a lot. Work in a project where all STL containers and most of STL algorithms are used, also used Boost.Iterators and Boost.Foreach. I have not used Boost.Range before, beause have not found it much useful.

Do you think the library should be accepted as a Boost library? Yes.