- GiNaC-list - ginac.de

Compilation under macos + question on matrices
by Laurent Bartholdi 16 Feb '13

16 Feb '13

Dear all, I'm new to ginac, and very eager to learn more about it. I installed the latest (1.6.2) version on a recent (mountain lion 12.2.1) system, using the automatic packaging system 'homebrew'. 1. There must be a bug somewhere along the line, because the most naive, simple program from the manual (and all other examples I tried) cause a malloc error: % cat x.C #include <iostream> #include <ginac/ginac.h> int main() { GiNaC::symbol v("v"); std::cout << v << std::endl; return 0; } % g++ x.C -lginac && ./a.out v a.out(31763) malloc: *** error for object 0x7fff7b419570: pointer being freed was not allocated *** set a breakpoint in malloc_error_break to debug Abort trap: 6 2. Setting matrix entries causes crash: ... GiNaC::matrix m(10,10); m[1][2] = 1; ... causes "libc++abi.dylib: terminate called throwing an exception" 3. I can't find the tensor product of matrices (aka Kronecker product) in the documentation. If M in an axb matrix and N is a cxd matrix, this would be the (ac)x(bd) matrix with entries the products of entries of M and N. While it's easy to implement myself, I'm surprised not to find it in the library. Did I miss something? I'm not sure I know well how this mailing list works, so I'd appreciate a reply directly to my email address. Thanks in advance, Laurent -- Prof. Dr. Laurent Bartholdi \ laurent.bartholdi<at>gmail<dot>com G.-A. Universität zu Göttingen \ Phone: +49 551 39 7826 Bunsenstraße 3-5 \ Secr: +49 551 39 7752 D-37073 Göttingen, Germany \ Fax: +49 551 39 22674

1 0

Re: [GiNaC-list] set_TeX_name() ?
by Dale Lukas Peterson 23 Jan '13

23 Jan '13

> Just as a suggestion: instead of an introduction of a separate > method set_TeX_name(), may be simpler to modify set_name() to have an > optional second argument for the Tex name (similar to the symbol() > method). Here are the three options I can see. First, a default parameter as I think you are suggesting: void set_name(const std::string & n, const std::string & texname = "") { name = n; TeX_name = texname; } Second, overloading set_name: void set_name(const std::string & n) { name = n; } void set_name(const std::string & n, const std::string & texname) { name = n; TeX_name = texname; } Third, the option I originally proposed: void set_TeX_name(const std::string & n) { TeX_name = n; } I prefer my original proposal since it seems the most orthogonal to me. But any of them is really fine with me so that there is a way to change the TeX_name after you've created an instance of symbol. If there is interest in keeping fewer methods, then I guess one of the first two options would be preferable. Thanks, Dale

3 11

Automatic simplification of C# expressions
by Eduardo Cavazos 20 Jan '13

20 Jan '13

Hello, I've started building a small library which implements automatic simplification of C# expressions: https://github.com/dharmatech/Symbolism I guess this is mostly off-topic, so sorry about that. I wanted to mention the library here because GiNaC has been hugely influential (as has SymbolicC++). It's a very limited library, but I think it might be useful for someone wanting to experiment with building GiNaC type functionality in C# without starting from scratch. Ed

1 0

How can I build an expression one character at a time?
by Roger Gordon 08 Jan '13

08 Jan '13

Hi guys, I recently started using ginac, and one thing I couldnt figure out from the tutorial is how can I make an expression one piece at a time. So if I wanted to build an expresison out of a list like x,=,5,*,4 I imagine doing something like ex expr; expr.append(x); expr.append(==); expr.append(5); expr.append(4); I understand the numbers are numerics, and x would be a symbol, but I dont know what the operators would be ( * / + - ). Does anyone how I can do this? Does this even make any sense to you? Thanks for any replies! R. Gordon

2 1

Simple question about how to solve for a simple non-linear system of equations.
by Zach 06 Jan '13

06 Jan '13

Hi guys I know this is a really basic question but after looking at the tutorial and reference I couldn't quite figure out how to solve a nonlinear system of equations. (If this is even possible, I dont know). I have this: lst eqns, vars; > > eqns = D==pow(T,2); > > vars =T; > > ex L = lsolve(eqns,vars); > But lsolve (I guess the l for linear makes it obvious) cannot solve this nonlinear function. Is there another way I can accomplish this? Then I could do something like: ex ans = L.subs(lst(D==3)); like I would for a linear? Is this even possible? Thanks for any answer! Ginac is a great tool for me so far!

4 3

Variable Isolation
by David Narvaez 17 Dec '12

17 Dec '12

Hi all, Is there a ready-to-use method to isolate a variable in an expression? At least for first and second degree polynomials in at most two variables. David E. Narvaez

2 4

Compiled GiNAC lib for MSVC failed in debug mode
by Christiand 13 Dec '12

13 Dec '12

Hi All, I have successfully compiled GiNAC for MSVC. I compiled GiNAC as static library (*.lib). I works fine under release mode, but fails in debug mode. Does anyone experience same thing ? Thank before for your attention. ~ Christiand ~

1 0

Ginac compile with MSVC
by Rheinländer 13 Dec '12

13 Dec '12

Hi, the current version of GiNaC requires a new hack to make it work with MSVC. The problem is lst::info(). It is never called when GiNaC is compiled with MSVC, instead, basic::info() gets called which leads to exceptions being thrown by code like: if (!eqns.info(info_flags::list)) { throw(std::invalid_argument("lsolve(): 1st argument must be a list or an equation")); The only work-around I could find was to move the definition of lst::info() from the .cpp file to the .h file (please don't ask me why this should make a difference). I know it's a stupid hack, but in any case, I attach the modified lst.h and lst.cpp in case you want to include it in a future release of GiNaC. Jan

1 0

set_TeX_name() ?
by Dale Lukas Peterson 10 Dec '12

10 Dec '12

I see that there is a set_name method in the symbol class and was wondering if there is a reason there is no set_TeX_name() method? I have a situation where I would like to declare a number of symbols, all using the default constructor, but then would like to set the TeX string of these symbols later. It seems like this implementation would work fine: void set_TeX_name(const std::string & n) { TeX_name = n; } Attached is a patch, if this implementation seems reasonable. I'm not sure whether TeX_name should be mutable or not, it seems like perhaps it also should be? Thanks, Luke -- “People call me a perfectionist, but I'm not. I'm a rightist. I do something until it's right, and then I move on to the next thing.” ― James Cameron

2 2

Empty brackets in compiled C code.
by Zach 07 Dec '12

07 Dec '12

Hi guys. Finally, with the help of those on this mailing list I was able to compile the lenet5 demo into c. But at first I was getting a few of these errors from the compiled C code: /home/refinedcode/Development/lush/packages/gblearn2/demos/C/lenet5_fprop.c:327:278: > error: expected expression before ‘)’ token > So i opened up the C code and found that all the errors pointed towards this one line: (L3_19__x_)->Vtbl->M_net_cscscf(L3_19__x_, L2_1_image_height, > L2_2_image_width, L2_3_ki0, L2_4_kj0, L2_13_table0, L2_5_si0, L2_6_sj0, > L2_7_ki1, L2_8_kj1, L2_14_table1, L2_9_si1, L2_10_sj1, > ((()((((double)(L2_1_image_height-((L2_3_ki0)-1)))/((double)L2_5_si0))-((L2_7_ki1)-1)))/(()L2_9_si1)), > ((()((((double)(L2_2_image_width-((L2_4_kj0)-1)))/((double)L2_6_sj0))-((L2_8_kj1)-1)))/(()L2_10_sj1)), > L_Tmp20, L2_12_output_size, L1_net_param); > > As you can see there are some empty brackets that have caused these errors. It wasn't a big deal to me as removing the empty brackets fixed the problem and it ran fine but is there somewhere I should post this? I don't assume anyone else has seen this before? Maybe im doing something wrong that caused it. If anyone wants I can post more details about it but I doubt it would be anyone's priority. Thanks for helping me with my questions by the way! Lush works very well!

1 0