I hacked together a script that parses METADATA.jl and the Julia general package registry in order to figure out if a package works on Julia v1.0. It is still very much wip and may contain some errors, but I’m temporarily hosting the list here:
The list is sorted by last push to the github repository, so most of the packages that do not work are near the bottom.
If there is some general interest I can continue publishing that list. Maybe we should also start a wishlist with packages that should be upgraded and another list with packages that are clearly deprecated.
Thanks you, it is a very useful list, I understand the dependences are useful to know why a package is not 1.0 ready (sometimes it is by the packages it uses and not by itself). The wishful list is a great idea also. I am starting with Julia and I could help porting packages. Great work!
Many thanks - this is a very useful list.
If hosting is a problem for you I have put a copy up on our London Julia site and would be happy to keep it there and also run the script on a regular basis, when you are happy with it.
Note that, for example, my
Fatou package has tests which fail on CI but the package does actually work on Julia version 1.0.1 and above, it just doesn’t work on automated CI testing right now.
This is a very nicely formatted and useful resource, but not necessarily accurate in every single case.
@sherrinm I will have to clean up my script first, but then I can send it to you.
You are right with Fatou.jl. The problem is that I am testing for Julia 1.0.0 compatibility. I think I still have not figured out the exact logic on how I should parse Compat.toml. That being said, Fatou.jl actually segfaults for me when I write
using Fatou. This is with Julia 1.0.2.
You need to check the readme on github,
using Fatou doesn’t work but
using Reduce,Fatou does, so an automatic testing is not going to catch that. Maybe someday it will be fixed.
Latest MAT.jl is v0.4.0, released Sep 15, 2017, and it does not work with Julia v1.0 contrary to your list. MAT#master seems to work. I think your list should show whether there is a working released version.
Fatou is now correctly classified. Thanks for the tip.
I agree with you that MAT.jl should not be classified as working. The problem is that a while ago all packages that do not work with Julia 1.0 where capped to a version <= 0.7 in the general registry. For reasons unbeknownst to me, that did not happen for MAT.jl. As a consequence, it is indeed possible to install version 0.4 of MAT.jl.
I could fix that by hardcoding this special case (as I did with the deprecated packages) but I don’t like this solution very much.
By the way, I believe the issue with the segfault should be completely fixed now with Reduce v1.0.3, so it is only necessary to use the regular command
using Fatou once again, without experiencing issues.