Refactor/attach comments on node

[clementdessoude]

Fix #262 and fix #276.

Will help to fix issue #270

[clementdessoude]

There is still an issue with handling some trailing comments. This behavior is also found in master

Input:

class T {
  void t() {
    if (e) {
      // empty
    } // comment
   else {
      
    }
  }
}

Output:

class T {

  void t() {
    if (e) {
      // empty
    } else {} // comment
  }
}

[clementdessoude]

It is ready for review. As I modified the behavior in the parser, I requested a review from you @bd82. Please let mme know if you have time to review or not 😄

I think there will be some edge cases that are not correctly handled, like those I corrected in labeledStatements and ifStatements. I think it could be good to fix these cases in separate PRs, as it will be difficult to review otherwise.

[bd82]

Hi @clement26695 I do not actually remember what was the previous behavior of comments handling and I do not have the time to go in depth in this large PR.

Can you describe what was done? what was the change and why?

[clementdessoude]

Basically, I attached the comments on nodes when possible, instead of tokens. Indeed, it is much easier to process it then in the printer, due to the way the group command work in prettier.

So, if you have time to look at how we attach the comments on nodes, it would be great, but otherwise, @Shaolans could have a look, and we could try to optimise it (if necessary) in other PRs

[bd82]

Attaching the comments on nodes instead of Tokens seems to make sense.
I looked at the code at a bit but only partially understood it, perhaps some more comments (pun intended :smile) could be useful to explain the logic e.g:

• What is the condition for a comment to be a trailing/leading comment to be associated with a node?
• What happens with a sequence of trailing/leading comments?
• What are the helper data structures you are building.

[clementdessoude]

You're right, it is quite unclear. Will try to clarify/document the process

[clementdessoude]

Done ! I hope it is more understandable

[bd82]

Done ! I hope it is more understandable

Yes makes much more sense now.

I think you could have also added info to the CSTNodes during parsing:

• First Token in CSTNode
• Last Token in CSTNode

Instead of creating some of these helper data structures and than perform the logic
directly on the CSTNodes and Comments array. However I do not have an opinion which approach is better.

[Shaolans]

Why a new line is required ?

[clementdessoude]

For me, it is just good practice to end a file by an empty line. But I don't know exactly why 😅

[Shaolans]

IMO, we should not enforce it since it JHipster app do not have that additional line or at least discuss about it.
Anyway great work !