release of version 2.3

[quaquel]

[coveralls]

Coverage remained the same at 81.139% when pulling 98e9d8cfa372fecda1e1fed98e6bfa0f7e55850f on 2.3-release into f1ca43423893352907ce5751f710045480d1958a on master.

[EwoutH]

@quaquel Awesome! I rewrote the changelog a bit, let me know what you think. The rest looks good!

I also added that this is the lastrelease series with Python 3.8 support. I do like to start implementing some type hinting, and having Python 3.9 as minimum version really helps for that. Is that fine by you?

One other thing, we should adres a few of the deprecations (in https://github.com/quaquel/EMAworkbench/issues/201) before release, because the 2.3.0 release series will likely live for quite some time (maybe even half a year). Especially the Sobol deprecation in SALib is coming quite soon.

[EwoutH]

Furthermore, do you want to do something with #82 and #102 before release?

[quaquel]

We are constrained by what is available on e.g., DelftBlue. In general, these systems are slow to adopt new versions. DelftBlue is still on 3.8 and as long as that is not changed, we cannot move forward either.

I'll see what I can do on #82 and #102 over the weekend. Neither one is complicated so it should be nice if those are closed. #201 requires some testing against the 1.4.6 of SALib.

The fact that 2.4 is likely to come only in spring does not mean that we can do intermediate updates to deal with e.g. depreciation issues.

[EwoutH]

They should really go straight for 3.11, that's 20%+ compute time saved right there.

Note that current versions can still keep running on 3.8, the version bump will only apply to the next release.

The fact that 2.4 is likely to come only in spring does not mean that we can do intermediate updates to deal with e.g. depreciation issue

Certainly true, but let's try to ship with the minimum of known (future) bugs.

[quaquel]

Any idea when 1.5 of SALib is expected? I am checking 1.4.6 right now but there sobol is not yet available. That is, only with 1.5 will sobol.py become available, but also saltelli.py will then immediately be removed. This means that there is nothing we can do at this point.

[quaquel]

They should really go straight for 3.11, that's 20%+ compute time saved right there.

In my experience, HPC software moves very slowly. There are so many complex interdependencies that have to be managed. This also means that I would not put into the release that 2.3 will be the latest with 3.8 support.

What does 3.9 offer for type hinting?

[quaquel]

Apart from the 3.8 point, the rest of the rewrite looks good.

[EwoutH]

I am checking 1.4.6 right now but there sobol is not yet available.

It should be in it: https://github.com/SALib/SALib/releases/tag/v1.4.6

But we can fix it in a 2.3.1 if preferred.

Let me come back on the Python 3.8 point.

[quaquel]

It should be in it: https://github.com/SALib/SALib/releases/tag/v1.4.6

my bad, forgot to update my virtual environment. Basically, by making SAlib 1.4.6 the minimum required version and shifting from saltelli to Sobol the problem can be avoided.

[EwoutH]

Perfect, could you cherry-pick a12c819 to a new PR?

[quaquel]

I guess we are ready to release this?

We moved the plotting stuff to 2.4, and MESA is not critical. I doubt I will get to it in the coming days so I suggest we move that to 2.4 as well.

[EwoutH]

Yes, let's release!

I already like that we have documented this. Do you want to do it?