Not much of a poster here, nor am I a professional developer (I jokingly refer to myself as an unprofessional developer, as I don’t have a formal background but code a decent amount).
I had the passing thought the other day that it seemed like Julia versions were slower in recent years, and was surprised to see no 1.13.0-rc2+ floating around (given that I recall 1.13.0-rc1 being available awhile back).
Here’s a crappy chart I made from pulling release dates from Github manually.I pulled from 1.6.0 to current, and my apologies if I typoed any of the data lol.
I was looking at rc1’s, 2’s, and 3-/4 (depending on version). The 3s and 4s weren’t actually needed though. Mainly, the chart shows the github release date for various versions and release candidates, the orange bars show the time between rc1 and rc2 for a given release (there was always an rc2), and the green bars show the time between an rc1 release and the 1.X.0 version release.
It does look like the rc1 duration (in progress) for 1.13.0-rc1 is off the charts, so I’m not sure what’s going on with that. 1.12.0-rc1 was the previous longest-lived rc1 at only 57 days, whereas 1.13.0-rc1 is now at 126.
Additionally, 1.11 and 1.12 were indeed a bit slower in terms of total rc time compared to prior releases.
Here’s a second crappy chart with the same dates/versions etc., but the bars are simply the time between the given 1.X.0 and 1.(X+1).0 releases in the data I copied down. 1.10 was quite fast, and indeed 1.11 and 1.12 have been noticeably slower. 1.12.0 was released 1 year to the day after 1.11.0’s release. “Version duration” is backwards-looking… like I said, it’s a crappy chart.
As I mentioned, I’m not sure what’s going on with 1.13, though it seems like it may match or exceed 1.12’s development time.
I thought the data may be interesting to others. My simple questions are along the lines of: did something change in funding/dev time available for core Julia? Is 1.13 just having some unforeseen challenges? Did 1.11 and 1.12 also have some challenges pushing the development time higher?
From a more basic dev standpoint, honestly the main things I’ve noticed since 1.10 are that things are seemingly running/pre-compiling slower, and I can redefine structs without detonating my REPL.
Side note: I don’t have Revise.jl called in my startup.jl, and upon review, it looks like struct definition is supposed to only happen in 1.12.X+ with Revise.jl active. I’m not loading Revise anywhere in VSCode or a REPL called from Powershell. I can perform the redefinition of “Inner” in the Revise tutorial with no issue: [can’t send link, it’s under “Struct Revision”], with “Revise” having no namespace “?” search result in my REPL. A VSCode REPL does show “Revise” in the namespace, so it’s doing something differently/implicitly loading it, which I didn’t realize. Confirmed both cases (powershell REPL and VSCode REPL) with
println(Main.varinfo(@__MODULE__; all=true, imported=true))
No idea why I can revise structs without Revise when search results say I’d have to be using Revise, but in any case, it’s convenient, especially for large structs that I have in some programs.
Perhaps my case is not very representative, but recent versions have had relatively few changes that change or improve my experience with the language. I believe others have better quantified slowdowns in recent versions than my anecdotal vibes, so I won’t speak with confidence that it has changed much (as I don’t know the broader quantitative results off the top of my head).
The latter comments here are meant to be more of a report without normative baggage. I’m sure people are working as hard as ever behind the scenes, and I know my perspective on inner workings of the language is pretty uninformed.