Import start.py using absolute path.

[yakutovicha]

Previously, the apps folder was considered a python module. Since
this is not true anymore [1,2], the loading of the apps' start.py
needs to be adapted.

[1] aiidateam/aiida-prerequisites#37
[2] aiidalab/aiidalab-docker-stack#213

[csadorf]

Looks good. Could you briefly comment about the motivation to use this particular approach as opposed to importlib.machinery.SourceFileLoader? I did some cursory search and found the following article: https://www.geeksforgeeks.org/how-to-import-a-python-module-given-the-full-path/

[yakutovicha]

Looks good. Could you briefly comment about the motivation to use this particular approach as opposed to importlib.machinery.SourceFileLoader? I did some cursory search and found the following article: https://www.geeksforgeeks.org/how-to-import-a-python-module-given-the-full-path/

I didn't find the importlib.machinery.SourceFileLoader approach, before. But since it is shorter, I will use it.

[yakutovicha]

I didn't find the importlib.machinery.SourceFileLoader approach, before. But since it is shorter, I will use it.

The amount of code is the only difference I see between those two approaces. @csadorf are you aware of the other differences?

[csadorf]

I didn't find the importlib.machinery.SourceFileLoader approach, before. But since it is shorter, I will use it.

The amount of code is the only difference I see between those two approaces. @csadorf are you aware of the other differences?

No, I am not aware of any differences besides the length and that the currently implemented approach allows one to control the execution of the module. It looks like both approaches are part of the public API and would be equally well-supported.

[yakutovicha]

I didn't find the importlib.machinery.SourceFileLoader approach, before. But since it is shorter, I will use it.

The amount of code is the only difference I see between those two approaces. @csadorf are you aware of the other differences?

No, I am not aware of any differences besides the length and that the currently implemented approach allows one to control the execution of the module. It looks like both approaches are part of the public API and would be equally well-supported.

ok, then the shorter one wins, I think.

[yakutovicha]

changed, tested - works.

[csadorf]

Assuming it works, approved! 😃