Jacob Walls

> The feature proposed in this PR makes it possible, optionally, to return "/mnt/root/dir/{MISSING_VALUE}/logs", which is what we need to detect the "/mnt/" prefix, and also provide better feedback to...

> I guess l'm wondering why you would detect that by calling inferred(), wouldn't you just directly inspect the AST? Oh, I see, it's so that both cases (mnt injected...

I guess I'm not certain we even know it's a string. Don't these uninferable cases include failures to format?

Re the factory proposal, as far as I understand, we have the ability to do that already: ```py >>> import astroid >>> def my_transform(node): ... def _infer_with_values(self, context=None, **kwargs): ......

@robgaston also mentioned hashing an object (elastic dsl object, queryset, the args to create one) as a potential way to keep using GET

> I’m starting to feel that we might be adding more complexity than necessary, and it may be overall better to just stick with the settings declaration. I’d love to...

> The expectation is that I would then see bar's console log. However I see arches-core console log. Thanks for the repro steps -- I followed it and got the...

(Hm, if project files are supposed to overwrite arches applications files, perhaps the new behavior isn't really an improvement.)

Retested the scenario from above, looking clear now 😎

Ah I thought we had consensus on collapsing the app/project distinction, but no problem to cleave that off here and propose it separately to vet it throughly 👍 I'll just...