Extend safe navigation docs about long &. chains

[fatkodima]

Closes #946.

[pirj]

I love how you put it.

Considering the law of Demeter, long chains are inherently bad. But this principle has downsides, and they are due to forcing to write wrappers (like presenters), or propagate increasing the interface of the top-level receiver class.

It really depends. Can it be that an address doesn’t have a zip? It’s business logic, can vary.

This change seems for the best. Can you try finding confirmation that the suggested style is widespread?
Is there a cop already to facilitate finding this ?

[fatkodima]

Can you try finding confirmation that the suggested style is widespread?

At least in my company we are striving to adhere to this. I will try to investigate other codebases.

Is there a cop already to facilitate finding this ?

I am currently implementing this cop - rubocop/rubocop#13171

[fatkodima]

Found only 1 offense in rails - https://github.com/rails/rails/blob/53ce1964f0bd6491fdc08aebc45f61275100f818/actionpack/lib/action_controller/metal/exceptions.rb#L43, so people definitely stick to it

[koic]

I'm unsure if this example is appropriate. According to rubocop/rubocop#13228, up to two consecutive &. are allowed by default. It would be better to explicitly clarify, with an explanation and example code, how many consecutive &. are considered "Long safe navigation chain".

[pirj]

Rails code might not be representative, I believe they are strict in code quality.

How about other repos from real-world-rails-apps and real-world-ruby in general?

[bbatsov]

We already have above the following rule:

Avoid chaining of &.. Replace with . and an explicit check. Perhaps it just needs to be expanded and clarified a bit instead of adding a new guideline?

[fatkodima]

Perhaps it just needs to be expanded and clarified a bit instead of adding a new guideline?

Updated existing section for safe navigation. I am bad at writing docs - concrete suggestions welcome.

[bbatsov]

Probably we shouldn't mention a specific number here and just expand a bit the rationale - namely that the longer the chain is, the harder it becomes to track what on it could be returning a nil. That's always been my fundamental problem with any safe navigation chains.

[fatkodima]

Updated with a suggested rationale.