It could be a good idea to have a thread where people show their interest in new Julia packages with features non-existent yet, and other people can also get ideas about what packages to develop.
For example, I haven’t been able to find any package for…
- Multiple imputation.
It is very unlikely that people would develop and maintain packages for others (that they don’t need themselves).
The bottleneck is not the lack of ideas.
This is true.
And also, there are things that I would use, and would be willing to contribute development time too, but am unlikely to undertake on my own. A thread (or discourse topic maybe) for finding like-minded folks to build something together might be useful.
Perhaps not professionals but students looking for a ideas for their projects or thesis.
If that’s the case they should looking to numerical differential equations! If any students are interested in numerical differential equations, automatic differentiation (Zygote), and the associated software ecosystem (global sensitivity analysis, sparse linear algebra, symbolic transformations), the JuliaDiffEq can offer mentorship (and future recommendation letters with a fancy MIT logo). Additionally, grant funding is available for part time positions if one demonstrates strong productivity. Please contact me if you’re interested.
(This is both a real advertisement and a comment: if you’re going to ask for people to do work, you have to make a fair deal and put some skin in the game!)
… should, primarily, talk to their advisor
Seriously, there tends to be little overlap between the “please implement some fairly standard methods” and interesting research projects. But students interested in the latter will just write code without prompting.
In practice most packages start by a single person implementing some core functionality or at least a proof of concept, then grow from there.
Most of the desired but unimplemented features already have issues. There is no shortage of work to do.
Indeed. Small amounts of standard methods can be used as a training exercise to get started, but if the student is actually looking for research credits they would be surely disappointed (and stop working) if all they were doing was implementing existing methods. At some point it has to be research, or it’s brute software development that has to be paid work for anyone to have interest. Or you have to do it yourself / wait for some brave soul to do it on their own time.
A more useful thread might be if you have time or money to make half (or some fraction) of a particular package, but you need to find the other half to make it happen. You could look for help writing a Meta-analysis package. Otherwise packages are written when someone else needs them enough…