Non idempotent formatting

[olafurpg]

See

 -  def log(formatToken: FormatToken): String = s"""${log(formatToken.left)}
 -                                                 |${log(formatToken.between:_*)}
 -                                                 |${log(formatToken.right)}""".stripMargin
 +  def log(formatToken: FormatToken): String =
 +    s"""${log(formatToken.left)}
 +       |${log(
 +           formatToken.between: _*)}
 +       |${log(formatToken.right)}""".stripMargin

Reformatting again produces:

  def log(formatToken: FormatToken): String =
    s"""${log(formatToken.left)}
       |${log(formatToken.between: _*)}
       |${log(formatToken.right)}""".stripMargin

in commit olafurpg@779c1ee?diff=unified#diff-b276550f97eeb5bceb770dc3d64811ceL19

[paulp]

Here's a non-idempotent situation not involving an interpolated string.

class A {
  def traced(in: A => Unit, out: B => Unit): Fun[A, B] = ( f
    .   mapIn[A] { x => in(x) ; x }
    .  mapOut[B] { x => out(x) ; x }
  )
}

becomes

class A {
  def traced(in: A => Unit, out: B => Unit): Fun[A, B] = (f
    .mapIn[A] { x =>
      in(x); x
    }
    .mapOut[B] { x =>
      out(x); x
    })
}

which becomes

class A {
  def traced(in: A => Unit, out: B => Unit): Fun[A, B] =
    (f.mapIn[A] { x =>
        in(x); x
      }
      .mapOut[B] { x =>
        out(x); x
      })
}

Also, it was a lot better before anything happened to it.

And in the "watch out for the future" department, those leading dots are only necessary through 2.10. In 2.11 and beyond that can be written

  def traced(in: A => Unit, out: B => Unit): Fun[A, B] = ( f
      mapIn[A] { x => in(x) ; x }
     mapOut[B] { x => out(x) ; x }
  )

I think dotless syntax is far superior for readability but I doubt most people can handle it unless their formatter does all the handling of it, hint hint.

[olafurpg]

Thanks for the example!

I think dotless syntax is far superior for readability but I doubt most people can handle it unless their formatter does all the handling of it, hint hint.

Some auto migration tool would be nice, indeed. I'm not sure about baking it with scalafmt, I think it better belongs to another tool like scala-tidy, similar to clang-tidy.