Replace optimade-client with ipyoptimade

[unkcpz]

Able to install with aiida-core v2.5.0 but turns out there are quite a lot things we need to fix.

As discussed (see the thread in this PR), will only replace the optimade-client with ipyoptimade~=0.1 which is the version right after the migration without the support to the pydantic v2 (which comes in version v0.2).

[danielhollas]

Thanks for working on this!
I do think we have to workaround the pydantic versions somehow to support earlier aiida-core versions

[danielhollas]

We should be very careful about bumping the minimum aiida-core version since it might cause issues for downstream apps. If at all possible, please make the ipyoptimade compatible with earlier aiida-core versions. Ideally, we should allow pip to pick the correct version based on which aiida-core is already installed in the image.

[unkcpz]

Good point, I think very hard and here is some thought.

ipyoptimade has nothing to do with aiida-core but with pydantic.

• ipyoptimade<0.2.0 which right after I made the migration, it only support pydantic==v1.0
• ipyoptimade==0.2.0 is the version that support pydantic v2.

So if we need something that need aiida-core==2.5.0 then need pydantic v2 is the environment and thus need to use ipyoptimade==0.2.0.
The problem now comes with that the use of optimade widget comes from aiidalab-widgets-base and it is another base dependencies in an AiiDAlab environment.

• One workaround is not use optimade widget from aiidalab-widgets-base but from ipyoptimade directly.Then nothing need to change here.
• Maybe another workaround worth to try is what you mentioned, that the ipyoptimade is an extra dependency.

In summary, I think we can not guarantee that one app not break another app. But we need to keep the bottom line that we need to try our best the new version of aiidalab docker-stack will not break the app.

[unkcpz]

It would be very helpful if you have a clear "break path" example such as: update of app-A or component-A would cause the break of some part of app-B.

[danielhollas]

I think our general goal with AWB is to be as independent of various other core dependencies as possible, to avoid coupling between versions of different dependencies.

ipyoptimade<0.2.0 which right after I made the migration, it only support pydantic==v1.0
ipyoptimade==0.2.0 is the version that support pydantic v2.

This is great that there are diifferent versions that support different pydantic versions! In the best case scenario, we should allow both versions in AWB and let pip to pick the correct version. Whether that will work in practice is unclear to me, since pip can ignore the existing aiida-core version constraints. This is another reason why we really should publish AWB as conda package and install it together with aiida-core in the base image.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 87.07%. Comparing base (bd29d28) to head (f4bd947).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #553   +/-   ##
=======================================
  Coverage   87.07%   87.07%           
=======================================
  Files          27       27           
  Lines        4649     4649           
=======================================
  Hits         4048     4048           
  Misses        601      601           

Flag	Coverage Δ
python-3.10	87.07% <100.00%> (ø)
python-3.9	87.10% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[danielhollas]

@unkcpz could you test what happens when you bump the aiida-core version in GHA to 2.5.0?

Otherwise I think this is good to go. Since aiida-core <2.5 versions did not depend on pydantic at all, I think this should be safe. I also double checked that no other packages besides pydantic depend on pydantic.

[unkcpz]

@unkcpz could you test what happens when you bump the aiida-core version in GHA to 2.5.0?

Ah, that's true, there are some deprecation warnings needs to be addressed, but better to do with another PR I think, just to bump the aiida-core version.

[danielhollas]

Okay, cool. Good to see that the installation with 2.5 succeeded at least.

[danielhollas]

We don't have that image yet so this was bound to fail.

[unkcpz]

Emmm. Then it is a bit tricky, that one has a test depends on AWB, so we need release docker image first or this?

[danielhollas]

Let's revert back to v2.4.3 and merge this first. I think then we might proceed with releasing the image.

Once this is merged, maybe you could make a alpha pre-release of AWB (2.2.0a0) which we can then use for both testing the image and QeApp (which needs to be updated to specify optimade extra dependency.

[unkcpz]

Releasing the image of bumping aiida-core to 2.5.0 you mean? But the test will fail since awb is not yet support 2.5.0.

This is a chicken egg problem. I think we should make a release of awb without changing the aiida-core version in notebook CI test. Make the release and then update the image.

[danielhollas]

Sorry for not being clear. I agree we can't release the image just yet, and we therefore need to keep the 2.4.3 version the the notebook tests here. Here are the steps that are needed afaik:

1. Merge this PR
2. Release a pre-release version of AWB
3. Make QeApp work with the pre-released version. Verify that it works in the axisting aiidalab images.
4. Release new version of AWB, v2.2.0
5. Release new version of QeApp
6. Release the new image with aiida v2.5

[unkcpz]

Thanks for clarification. But I think it will not working for the step 6.

Release the new image with aiida v2.5

This is impossible if AWB not yet have pydantic v2 supported.

aiidalab docker stack full stack test:
aiida-core 2.5.0 -> pydantic v2 -> aiidalab-qe (pydantic v2 support) -> AWB (pydantic v2 support)

So if not yet AWB pydantic v2 supported

For AWB test, it depend on the image with pydantic v2 supported, but that image depend on the this as shown above.

[unkcpz]

But anyway, before step 6, I think it is all doable. Let's get this merged and when proceed with can see the potential problem more clear and able to find the solution.