But people want what they want, except better. Where I work, numpy is
the big reason I can't recommend Lua (which is fine; I'm not an
FWIW, I've been a self-described evangelist and have now calmed down about that.
For our project, Lua is truly the best choice. We're integrating broadcast video APIs and are using Lua to expose behavior that would otherwise need to be written in C. For example: we're writing all of the iterators in Lua so that file position, styling and composite keyframing can be re-made "in the field."
We're on machines that have a hard cost of about 5k (plus the broadcast hardware). They have gobs of memory and processor cores. Yet, Lua's size and installation ease are extremely important. It's a dependency that is very easy to handle in our project and its size makes us trust it. Any time we have had library that causes issues and our answer has been to "try the update," it's proven to be a burden to us. Lua is the perfect example of the opposite of that kind of experience.
But as an application language, I can think of few modern choices that are worse than Lua. It's verbose, there aren't many libraries, etc.
All things in their place. What is so, if I may, *precious* about Lua is that it is so focused. As I would lament the lack of XYZ, I was also enjoying the benefits of a focus that was keeping XYZ away.
My point is that this is a truly special project. By special, I mean "rare" and I also mean "specialized." Perhaps when Lua becomes more popular, it will also be the day that this is no longer true.