Simplify argument type checks and exception raise about fedbiomed HOT 6 OPEN

I fully agree 100%!

I have a couple more opinions:

• for exceptions, I think there used to be an interest in converting everything into a Fed-BioMed exception because these could be treated differently by the runtime. For example, we may decide that the node CLI should crash if a non-Fedbiomed exception is raised (because we don't know what to expect), but if in fedbiomed exception was raised we might assume that at least it is a known issue/possibility and therefore no need to crash the node. However, I think that it would make more sense to raise a Fedbiomed AttributeError in your example, rather than what we currently do.
• I would be in favour of strongly reducing the types of check we do on TrainingArguments (or even more extrem, I would be in favour of removing them altogether)

from fedbiomed.

Hi @srcansiz,

For me, this kind of things about agreeing on coding practices (or changing the existing ones) should be discussed and decided together, maybe in a retrospective meeting (or dedicated meeting ?). And then written and followed :-)

At first sight, I would tend to agree with 1. but
disagree with 2. and 3. that we purposely implemented for robustness.

from fedbiomed.

Hi @sharkovsky, @mvesin,

Thanks for your comments. Yes I agree @mvesin It can be something that we can discuss during the retrospective or we can have an internal meeting during this sprint. The main idea is to reduce the complexity of the code, avoid writing unnecessary code, and thanks to that reducing the time spent while coding, unit-testing etc.

from fedbiomed.

If matters of coding rules (these ones, maybe others) are of interest to the team, maybe we can dedicate next sprint retrospective to discussing & deciding on some of these aspects ?

from fedbiomed.

Yes, I agree it matters of coding rules even though we don't have specific written rule for that. My point is that it seems we won't have enough time discuss and decide it during retrospective. However, I think we will need maximum 1 hour to discuss and decide, we can spare 1 hour when everyone is in SED. But if you wanna delay it for the next sprint yes we can.

from fedbiomed.

I agree we have few coding rules, but exceptions is one of the few really specified and documented aspects: https://github.com/fedbiomed/fedbiomed/blob/develop/CODING_RULES.md

But it doesn't seem we apply these rules (eg I am the first one to use except Exceptions:) ...
It probably means we need to agree on evolved rules.

from fedbiomed.