> This is the main reason for so many people being busy with re-inventing the wheel.
I don't know about other people, but I'm reinventing the wheel to learn better.
An algorithm you reinvented only to discover a better version later will stick to your head. And it's also fun doing it.
Also, a library or a tool you "reinvent" is many times better for your needs considering that assembling software pieces is not actually like putting together Lego blocks (although corporates are still trying).
Quicksort reinvention is also useful, because that standard qsort you've got laying around there might not behave as you expect (like glib's qsort that uses merge-sort instead) or it might not be optimized for your datasets (because you know, depending on the partitioning function you might end up in the worst-case scenario quite a lot).
> I don't know about other people, but I'm reinventing the wheel to learn better.
That's an excellent reason for doing so. But the GGPs point was I think people reinventing the wheel simply because they are not aware of what was done in the past.
> An algorithm you reinvented only to discover a better version later will stick to your head. And it's also fun doing it.
That's very true. I played around with clustering for a while, then decided to get serious about it which means downloading and studying all the papers that I can find in open access journals or on websites of the authors. That way you get the best of both worlds, first you gain the experience of doing it yourself and appreciating the problems of that particular domain, then you get to see how others dealt with those problems.
I don't know about other people, but I'm reinventing the wheel to learn better.
An algorithm you reinvented only to discover a better version later will stick to your head. And it's also fun doing it.
Also, a library or a tool you "reinvent" is many times better for your needs considering that assembling software pieces is not actually like putting together Lego blocks (although corporates are still trying).
Quicksort reinvention is also useful, because that standard qsort you've got laying around there might not behave as you expect (like glib's qsort that uses merge-sort instead) or it might not be optimized for your datasets (because you know, depending on the partitioning function you might end up in the worst-case scenario quite a lot).