Hello, On Sat, Oct 04, 2008 at 09:58:25PM -0400, Krishna Mohan Gundu wrote:

...

Is there a better way to test if a symbol is a vector?

1. I don't think subclassing symbol is a good idea. 2. `vector' is a bad name (because there is std::vector) 3. What's wrong with schoolbook method: typeid(blah) == typeid(YourVector)

class symbol_base : class basic { ... symbol_base(const std::string & initname, type_info* ti); ... type_info const& get_domain_type() const { return *pti; } ... private: type_info const* pti;

symbol is a bit performance (and memory) critical. Adding yet another 8 bytes is not appreciated at all.

Now things like is_domain_a<real_domain>(e) is_domain_a<vector_domain>(e) makes sense.

What's wrong with typeid(e) == typeid(vector_symbol)

as friend cannot be a template.

Sure it can. See the example code in the paragraph 14.5.3.1 of the standard, and this silly program: $ cat test.cc template<typename T> class A; template<typename T> class B; template<typename T> class A { T i; template<typename T1> friend class B; public: A() : i(0) { } A(const T& i_) : i(i_) { } }; template<typename T> class B { T j; public: B() : j(0) { } B(const T& j_) : j(j_) { } template<typename T1> B& operator=(const A<T1>& a) { j = T(a.i); return *this; } T get_j() const { return j; } }; int main() { A<int> a(0); B<long> b(-1); b = a; return b.get_j(); } $ g++ -Wall test.cc $ ./a.out $ echo $? 0

Let me know if you think of any simpler alternatives.

Use already existing code (i.e. clifford class), add whatever functionality you need. Don't hesitate to ask questions if something goes wrong and/or you don't understand something, etc. Best regards, Alexei -- All science is either physics or stamp collecting.

nomad wrote:

Alexei Sheplyakov-2 wrote:

...

...

I am planning to do it as follows 1) extend a symbol to a vector

I don't think it's a good idea. Symbol is something simple and in a sense "atomic". Also it's quite performance and memory critical. You'd better use the clifford class.

I need to think about this.

It is now somewhat clear to me how to proceed. I will represent a vector by a vector symbol, a basis symbol and possibly fixed size. I will represent scalar and vector products as separate products different from '*'. When substitutions are made for the vector symbol and basis symbol, I will evaluate the scalar and vector products. One thing that is not clear to me however is indexing into vectors. If we have vector A and B, what should something like A.0 * B.1 signify? Should indexing be supported at all? -- View this message in context: http://www.nabble.com/extending-the-symbol-tp19818944p19853690.html Sent from the Ginac - General mailing list archive at Nabble.com.

Hello All, I am attaching the code for performing vector algebra using GiNaC. It supports inner products, cross products and components of vectors. Currently cross product between two 2D vectors returns a scalar and between two 3D vectors returns a vector. I hope that someone will find it useful. Here are a few examples // define vectors vector v1("v1"), v2("v2"), v3("v3"); // vector components cout << v1[2] << endl; // v1[2] // various products cout << inner_prod(v1, v2) << endl; // (v1 . v2) cout << cross_prod(v1, v2) << endl; // (v1 x v2) cout << inner_prod(v1, cross_prod(v2, v3)) << endl; // (v1 . (v2 x v3)) cout << cross_prod(v1, cross_prod(v2, v3)) << endl; // (v1 x (v2 x v3)) // identities (nothing fancy yet) cout << cross_prod(v1, v1) << endl; // 0 // expand ex e1 = inner_prod(v1, v2 + v3); cout << e1 << endl; // (v1 . (v2 + v3)) cout << expand(e1) << endl; // (v1 . v2) + (v1 . v3); // substitution (expand must be used before substitution) lst l1(1,2,3), l2(1,0,1), l3(1,1,0); cout << v1[2].subs(v1 == l1) << endl; // 3 cout << expand(e1).subs(lst(v1 == l1, v2 == l2, v3 == l3)) << endl; // 7 cheers, Krishna.

Hello,

Can you please attach a copy of your replies to my email address? I disable email delivery whenever I subscribe to forums as most of the times I post infrequently.

Sure. Also you can set proper Reply-To: header so my email client software will do this automatically.

If vector and scalar products of vectors are not given their own identities but are implemented using the clifford algebra as in the post http://www.ginac.de/pipermail/ginac-list/2005-June/000671.html then the output no longer bears any resemblance to the original definition. For example the consider the following output

"-C.0*B.2*A.1-C.1*A.2*B.0+C.2*B.0*A.1+C.0*A.2*B.1+C.1*B.2*A.0-C.2*B.1*A.0" There is no resemblance to the original definition

In general, the canonical form of GiNaC expressions is unpredictable, i.e. it changes from run to run (no matter what is the type of the expression). The following script demonstrates this: for i in `seq 1 100`; do echo "a = x; b = y; x - y;" | ginsh done | sort -n | uniq x x-y y -y+x Depending on luck, canonical form of the same expression (x-y) might be either 'x-y' or '-y+x' (of course it's the same during one run). This is kind of design decision: we trade deterministic output for speed. (Still nothing prevents interested parties from writing a custom printing method which outputs terms in a more or less predictable manner).

which could be printed nicely "A . (B x C)"

To emphasize my point, I dont really care how the triple product is represented inside. But when things go wrong (due to bugs) I would like to easily debug the code. For example, if I was expecting A . (C x B), the second output will immediately point out that to me.

This problem can be solved with a custom printing method. But the real question is: do you really need/want to represent tensor expressions in a component form?

On the other hand if scalar and vector products of vectors are given their own identities, then there is no need for clifford algebra any more to evaluate them.

You can write a new class from scratch if modifying clifford (or indexed) is not an option. Best regards, Alexei -- All science is either physics or stamp collecting.