bug#16182: Acknowledgement (24.3.50; ruby-mode: Indentation style of multiline literals with hanging open paren inside other parens)

[Dmitry Gutov]

On 20.12.2013 17:46, Bozhidar Batsov wrote:
> On Friday, December 20, 2013 at 1:57 PM, Dmitry Gutov wrote:
> > On 20.12.2013 11:51, Bozhidar Batsov wrote:
> > > Just a small nitpick - everything that returns a value is actually an
> > > expression, not a statement.
> >
> > It can be both (see "expression statement"). This way it's not
> > ambiguous, because I'm really aligning to the statement: the containing
> > expression, which follows the bob or an [implicit] semicolon.
> >
> > In Rubocop, you've chosen to align to just the parent expression. Maybe
> > we should find a realistic example where one would be different from the
> > other.
> I don’t quite understand what you mean.

This example is indented just like Robocop master likes with (AlignWith: 
variable):

b = a = if 3 == 4
      1
    else
      2
    end

puts a
puts b


Someone correct me if I'm wrong, but I suspect that users who like less 
indentation would prefer to have the `if' body and closer to be aligned 
to the beginning of the statement, rather than to `a'.

That's what ruby-mode does now if `if' is in `ruby-align-to-stmt-keywords'.

Another reason to pick this behavior is that "align to parent" is harder 
to implement. SMIE has no AST: we can find the position of the parent 
token (=), but finding the position of `a' will require manual seeking.

And `a' could be more than a plan variable: maybe something like 
`b[a+1]' or `foo[:bar][:qux]`.

`=' is also not the only operator we can handle. Aside from its 
variations (||=, etc), we might want to support `||' and others.
And the left side of `||' can be an arbitrary expression.