ccgsl / News: Recent posts

Some minor updates. Two bugs fixed in vector_char and minor changes to documentation.

Version 2.7.1 released

This version works with GSL-2.7 and should compile with very few warnings with C++20.

This version works with GSL-2.7.

This version fixes a bug that caused small memory leaks. Specifically the wrap_*_without_ownership() functions in the vector and matrix classes would create an object that leaked four bytes when deleted. These functions are unusual, but are used, for example, with multimin. So four bytes would be lost each time one created an mutimin optimiser. There are a handful of other places where these can be called. The updates stop this leak.... read more

This version has more documentation updates than anything else. But it also has a few minor changes and substantially reduces the amount of code needed for function_scl, for example. The code now covers everything documented in GSL 2.6 except a few features that are better supported natively in C++.

This version has a few minor changes and lots of additional tests. It should now be possible to use all of gsl 2.6 except some functions like sorting built into the C++ standard template library through ccgsl. Nearly everything that used pointers to arrays can now use std::vector<t> or gsl_vector instead. Nearly all gsl functions that take function pointer and void* param arguments can use C++ functions or function objects instead.... read more</t>

This one has a combination of documentation changes and additional functions to allow us to use, for example, std::vector or std::array in place of double[] in functions that did not already have this feature.

This version adds some new features so that you should always be able to use function objects in place of function pointers.

This covers nearly all of GSL 2.6. There will be some bugs and inconsistencies. But, since it mainly wraps the Gnu Scientific Library these should be less problematic.

This covers a lot of version 2.3 of the Gnu scientific library. It still wants multilarge, rstat, spblas, splinalg and spmatrix.

This is a minor update on version 1,16, which corresponds to gsl-1.16. There are minor changes in documentation and additions of iterators for the permutation and combination classes. The combination class also now has an option to initialise the combiantion to the first valid combination on construction.

JDL

This is an update. It corresponds to GSL-1.16.

This is another minor update. Only the configuration files are changed; so if this version doesn’t work try 0.99-13

This is another release with very minor changes. These make it possible to select various vectors, subvectors and submatrices with 0 entries, rows or columns, even from a vector or matrix with 0 entries, rows or columns.

This version fixes a few obscure bugs and creates overloaded functions (sometimes through templates or default arguments) so that you no longer have to supply stride arguments to functions that require an array.

In a few cases it also allows std::vector<T> or gsl::vector* where previously a C-style array was expected.

The permute namespace is tidied up so that permute functions include the standard (forward) permutations and take a gsl::permutation argument rather than the size_t const array that is usual in GSL.... read more

More minor changes. These are to the matrix classes and allow C++11 expressions of the following form:

for( auto row : matrix ){ … }

Here matrix is an object from a matrix class and wor is an element of the following vector class. Reverse iteration is possible, though clunkier. It should be slightly more efficient to use the matrix.set() and matrix.get() operations together with row and column indices.

Added real() and imag() views to the vector_complex classes.

Also fixed a minor bug in a some vector_complex and vector_matrix classes that meant that if, for example, a vector_matrix was declared but not given an initial size, its destructor would attempt to decrement nullptr.

No real new material. But I’ve added some overloaded functions to the vector and matrix classes. An example might illustrate why. In C we have to use subvector( v ) and const_subvector( v ) to construct non-const and const subvectors. But in C++ we can overload v.subvector() to return either a non-const or const vector depending on whether v is const or not.

The const_ versions of functions are now effectively obsolete, but they are not deprecated because it may be semantically useful to use them in an expression like:... read more

Only one new function (in bspline), which is hardly needed and no bugs found to be fixed. But added tests of bspline and stats in source code. The tests don’t get installed.

This covers everything in GSL that can sensibly be accessed through C++ wrappers.

This version includes classes to make any reasonable function object into gsl_function object. It also adds direct support for linear regression and numerical differentiation.

ccgsl 0.1 is now available.