Assist to convert nested function to closure.

[mdx97]

I'm a bit worried about the has_semicolon function. The syntax node returned for AssistsContext::find_node_at_offset doesn't seem to capture the semicolon of the nested function if it has one. The current implementation was the only workaround that I could think of. If there is a better way let me know! Maybe we have to fix something upstream in Rowan?

closes #13230

[bors]

☔ The latest upstream changes (presumably #13516) made this pull request unmergeable. Please resolve the merge conflicts.

[Veykril]

We should also check that the function has no generic parameters, as we can't transform that into a closure (+ a test that verifies this)

[Veykril]

This will trigger incorrectly if we have an associated function nested in a function (very rare that someone does this but it could happen), so it might be nicer to do the following

function.ancestors().skip(1).find_map(ast::Item::cast).unwrap_or(false, |it| matches!(it, ast::item::Fn(_))

[mdx97]

I'm not sure what that snippet is supposed to do, namely because there are two arguments that are passed to unwrap_or. Would it be easier to just check if the function is an associated function?

[Veykril]

oh my bad, this was supposed to be a map_or_else. But yes it would be easier to just check if it is an assocaited function actually, that is go from the ast::Fn to the hir version and check if that is an assoc item.

[Veykril]

Sorry for the late review, was a bit busy.
Could you remove the merge commit? We try to keep our commit history merge commit free if possible (and do rebases instead).

[jplatte]

Ping @mdx97, are you going to come back to this? I can take over and rebase if not.

[mdx97]

@jplatte You can take this over if you'd like. I don't really have a timeline on when I could get back to this.

[Veykril]

Closing in favor of #14455