So, my first question and answer to the site, I believe, was: Recommendations for a usable, fast C++ matrix library?

It's gotten a lot of views and votes, but now that I'm a mod and I've been on the site for over 2 months, I look back at it, and I feel like it's gotten out of hand, and isn't a good question as posed because it's starting to attract answers that aren't so great. Since I wrote the question and answered it, I feel like it's my responsibility to fix it.

The question was not as focused as it could have been. I wrote the question because I wanted a library to use for a project I'm working on, and I spent a couple of days looking for an answer to that question. I knew that I'd be using dense linear algebra as I started out. I also knew that I might want to use sparse linear algebra later. I probably should have specified that I wanted to know about established libraries (i.e., ones with some sort of institutional backing).

My answer was also not as focused as it could have been. I listed 11 libraries in one answer, which is not good if one wants a list I tried to cover dense libraries, sparse libraries, and hybrid dense-sparse libraries, which reflects the origin of the question; I didn't know at the time if it would be good to stick with a dense library, or plug a sparse library into the project instead. (I still don't know.)

Given that we're starting to refine and sharpen the idea of what we want out of the site, how can I fix this mess? (I have my own ideas, but I'll post them after I've gotten some sleep...)


To be honest, I can't really think of a way to make the question really ideal for the Stack Exchange model without completely rewriting it, and potentially invalidating some of the existing answers in the process. A good-for-SE question probably would have gone something like "I'm currently using matrix library X but it doesn't work because Y. What can I replace it that would do Y correctly?" But you probably can't retroactively insert an existing setup (X) and problem (Y); presumably you asked the question the way you did because you didn't have X and Y in the first place, and were just trying to build a resource that would be useful in the future. And even if you do have some X and Y, rewriting the question to incorporate them would make it quite different.

If the problem is that it's starting to attract bad answers, the best solution might just be to close it. The question can still remain on the site as a resource to later visitors, or your list could be incorporated into the tag wiki for linear algebra.


I don't have any good ideas. It was a fairly specific question, so that part of it was fine. I think the problem was that your answer was so well-composed, researched and instantaneous, that nobody else was motivated to respond. I also think that the answers should have been atomic, which would have helped substantially.


My thought is to triage it for now, so that there's at least a defensible reason for downvoting or deleting bad answers in the interim.

Breaking up my big answer into atomic pieces is another option, followed by wikifying the answers (and maybe even the question).

  • If you're okay with breaking it up into atomic pieces and wikifying, I think it would be helpful. – Aron Ahmadia Feb 16 '12 at 18:58
  • I think it's a good idea. I'm still figuring out the SE format. To paraphrase a conversation I had with other mods, I was told that it's better when answers that are lists are in one big post rather than lots of small posts. At the same time, I also agree with you about the whole "wheat with the chaff" argument that throwing together good and bad recommendations may not be helpful. I'll see what expert opinions I can dig up and report back. – Geoff Oxberry Feb 16 '12 at 19:45
  • Geoff, the biggest problem with that answer is that it doesn't provide strong guidance as to which library to pick up. If your answer had been more discerning, it might have remained stronger as a list. This isn't a criticism of you, I'm just trying to explain my reasoning, we will inevitably be making more of these decisions. – Aron Ahmadia Feb 16 '12 at 20:00
  • I don't take it personally; it's constructive criticism, and I appreciate the feedback. I'll first try to rework the single answer into one that makes stronger recommendations about scenarios in which each library might be good, and then look at breaking it up. I definitely welcome the chance to improve my answer (and more generally, my writing). However, given the number of things that are coming up in modding and research these days, it may be a little while... – Geoff Oxberry Feb 16 '12 at 21:05

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .