SciRuby
is rb-gsl dead
Hello all — I was trying to install rb-gsl, and running into various issues. Among other things, it looks like it is no longer keeping up with gsl versions. gsl 2.8, for example, doesn't work; even after fixing warnings, we get errors like this
bspline.c:101:39: error: no member named 'k' in 'gsl_bspline_workspace'
B = gsl_vector_alloc(w->nbreak+w->k-2);
which suggests to me that the underlying gsl code has changed sufficiently that things are broken. It's potentially possible to keep things going, but it does worry me that errors might be sneaking in that we don't see.
It would be sad to lose ruby support for mathematics, but if anyone knows of a project that is working well, do please share.
@prasunanand @katafrakt @v0dro @pjotrp @mrkn @translunar Any chance on getting a new maintainer on this project?
@wynksaiddestroy from my very limited experience with this project, it requires very good knowledge of both GSL and Ruby. Every Ruby release is likely to break this, sometimes with changes added last minute. So it would probably be quite hard to find someone to spend a lot of time on a relatively not popular project, which is super unfortunate.
@wynksaiddestroy from my very limited experience with this project, it requires very good knowledge of both GSL and Ruby. Every Ruby release is likely to break this, sometimes with changes added last minute. So it would probably be quite hard to find someone to spend a lot of time on a relatively not popular project, which is super unfortunate.
@katafrakt I didn't want to sound ungrateful or accusatory. I don't know much about either C or GSL. But it's such a pity that there are no decent statistics libraries for Ruby. And I believe that the state of this gem can't get any worse. So why can't you become a maintainer, for example, and at least try to do better?
On Thu, Apr 03, 2025 at 01:04:16PM -0700, Fabian Winkler wrote:
@.*** I didn't want to sound ungrateful or accusatory. I don't know much about either C or GSL. But it's such a pity that there are no decent statistics libraries for Ruby. And I believe that the state of this gem can't get any worse. So why can't you become a maintainer, for example, and at least try to do better?
Also both rb-gsl and nmatrix have been downloaded 200K+ times. That means the packages are in use and not dead. Free software lives with its users and the users can take it on to work with them. That is how we started, even if there is no assigned developer/maintainer now.
On Thu, Apr 03, 2025 at 01:04:16PM -0700, Fabian Winkler wrote: @.*** I didn't want to sound ungrateful or accusatory. I don't know much about either C or GSL. But it's such a pity that there are no decent statistics libraries for Ruby. And I believe that the state of this gem can't get any worse. So why can't you become a maintainer, for example, and at least try to do better? Also both rb-gsl and nmatrix have been downloaded 200K+ times. That means the packages are in use and not dead. Free software lives with its users and the users can take it on to work with them. That is how we started, even if there is no assigned developer/maintainer now.
I'm not sure if the number of downloads is a good metric for the actual usage of the gem. It's certainly not usable in its current state. And there are other gems, such as distribution, that depend on rb-gsl and therefore don't work either. I'm sure I'm not the only one who feels this way, but before I start the tedious task of getting the gem to work myself, I use something else, e.g. numpy. But I don't really want to do that. I want rb-gsl and the other SciRuby gems to work.
We are, I fear, a dying breed! Probably the most powerful thing people could do for ruby is write a gem that enabled us to interface with the relevant Python packages. GSL is a wonderful library, but there are more powerful methods and tools in scikit-learn, and so on.
I was a big fan of Ruby and rb-gsl for many years, but I have moved on to Julia now. Julia has built-in support for multi-dimensional numerical arrays and is as fast as C since the code is automatically compiled to the CPU's (or GPU's!) native instruction set using LLVM at or before runtime. Julia also has a rich ecosystem of mathematical and plotting packages as well as fantastic GPU support. I would strongly encourage anyone thinking of switching away from Ruby and rb-gsl to consider Julia as an alternative.
