Trying to use JLD.jl with Julia v1.7.0-beta3

I need to save variables of different types in Julia, and I’ve tried JLD.jl in Julia v1.6.2 without any problem. However, since I’m using the new Mac M1, I moved to the new upcoming release v1.7.0-beta3 (and it’s really much faster!), but I get an error after trying to use the package:

julia> using JLD
[ Info: Precompiling JLD [4138dd39-2aa7-5051-a626-17a0bb65d9c8]
ERROR: LoadError: UndefVarError: libhdf5 not defined

I also have problems (cannot use or compile) with other related packages: JLD2.jl, BSON.jl and HDF5.jl. Is this a normal behaviour because it’s a beta version? Will I be able to use these packages after the stable release of v1.7?

1 Like

jld does not work with 1.7, see error on Julia 1.7.0-beta2 · Issue #300 · JuliaIO/JLD.jl · GitHub

On my (Windows 10) machine, JLD2 and HDF5 work well with 1.7.0-beta3

1 Like

You are right: JLD2 works fine in v.1.7.0-beta3. I don’t know why I couldn’t get it to work before. Thank you very much.

Native aarch64-apple-darwin build of Julia is still missing many binary libraries, it’ll take a while to catch up. The x86_64-apple-darwin build running with Rosetta should work fine instead

1 Like

But the execution time ratio is about 1/1.6, where the 1 is for the v1.7.

In case you missed it, the aarch64-apple-darwin platform is classified as Tier 3 in terms of support, which means a lot of things won’t work. There are some outstanding bugs that still have to be fixed. Unless you can provide help to fix them, using the x86_64-apple-darwin build is a more stable, albeit slower, option.

On top of that, as I already said, some binaries will be missing for some time, and you can’t do much more than being patient.


JLD.jl now works on 1.7.

I find it hard to understand - should I use JLD or JLD2?

I find it hard to understand - should I use JLD or JLD2?

That’s clearly your decision. Both seem OK with recent Julia versions. (For my own work, I moved to JLD2 a year ago.) /Paul S

1 Like