@gknauth i agree that it had a great message and information. i’ve been passing it along to the astro faculty i know who are heavily involved in teaching computation. i think there’s good ideas for us in the sciences to learn

Another great thing I got from Kathi’s talk was her mention of Advanced R, the book by Hadley Wickham. I bought the book before I even realized it’s online https://adv-r.hadley.nz/ but I’m glad I bought it. I do like to support great authors, and this book is a gem for understanding lots of things about R. I’ve done some R, not tons, partly because it seemed “weird,” but looking at Wickham’s book really makes me want to dive in, not just for R reasons, but because I think it could help me in other languages too and in building software people find useful and want to use. (Most of the ML/DS stuff at work is in Python, because “batteries included,” I think, but if you really want to push the envelope, I think it’s going to happen in a functional language.

The last RacketCon made me take a second look at Rust, primarily to port something I liked that was written in Perl, into something portable that would be bulletproof and run very fast.

And one of the reasons I got into Scala way back when (~10 years or more ago) was I was reading Odersky’s book, I saw the creators of Racket saying nice things about Odersky’s work, and that was enough for me to give it a try.

I realize Python & R have a strong foothold for data sciency stuff, but I think Julia has a lot of promise, and is a better designed language. My plan for that domain is to do as much as I can in Racket (general purpose coding, I/O, data cleaning, etc.), and delegate the few things where Julia shines to it.

I’m curious about Julia too. But in astronomy, there’s a ton of intertia behind python and astro-specific packages (plus of course numpy, scipy, pandas). I’ve seen there’s a way for Julia to call python packages. But I wonder if python is 90% good enough, and Julia will never gain a foothold.

It’s hard to say - language success is such a mysterious thing, but I do see lots of momentum with Julia, and the foundations seem pretty solid.

there does seem to be buzz about it. but i’ve yet to see an astronomy paper that uses it

time will tell though

it’s early still

Here’s something, but no idea how useful it might be: https://bids.berkeley.edu/news/julia-astro-enabling-next-generation-analysis-astronomy

yeah, i’ve seen that. it seems in a way to be missing the mark… julia astrolib looks to be replicating the functionality of the IDL astrolib (IDL was dominant before python came along), but the “state of the art” for python is astropy (https://www.astropy.org/) which seems at least a factor of 10 (perhaps 100) more featured, has almost universal recognition, and has a sizable, distributed developer community. so the julia astrolib (and some of the other packages) seems to be aiming way too low in terms of functionality

and looking at the history of packages like astropy (and their large collection of affiliate packages), the development has been going on for 6–10 years. so i think for julia to “catch up” in astronomy, it’ll need to attract those python developers. maybe it’s happening, but i haven’t really heard anyone talk about it. I do want Julia to succeed, but I’m afraid python is “good enough” for astronomy at this point. I also don’t have time (or the experience really) to push for Julia in a meaningful way. But that might change if one of the grants i’m thinking of submitting ends up working out. :man-shrugging:

I posted this at work: https://arstechnica.com/science/2020/10/the-unreasonable-effectiveness-of-the-julia-programming-language/ and our data science guy responded: “Yes I LOVE Julia, but it had to be weird and start array indices at 1 :sadpanda: I’ve been waiting for it to get an easy to use compiler.”

If someone ports Emacs to Julia, then Julia will surely take off.

@badkins just noticed the date on the juliaastro link! it’s been around quite a bit longer than i’d thought. given that, my guess is it was close to feature parity (certainly in terms of the planning) with astropy at that time, but python/astropy attracted more developers