Discourage multi-argument infix syntax: lint applications (x op (a, b)), also lint operator-name definitions

[som-snytt]

Lint multiarg infix applications, for example x op (a, b)

Also lint method definitions with multiple parameters where the method name begins with an operator char, for example def +(a: Int, b: Int), on the assumption that operator-style naming encourages infix at call site.

Message in both cases is "multiarg infix syntax looks like a tuple and will be deprecated"

This aligns Scala 2 with Dotty (and likely Scala 3) changes, as per scala/scala3#4311 and http://dotty.epfl.ch/docs/reference/changed-features/operators.html

[som-snytt]

Drafting to handle (i, j) == (42, 17).

[som-snytt]

Also don't warn on deprecated definitions.

Or context bounds: def ?[T: ru.TypeTag : ClassTag] = InternalInfo[T]

[som-snytt]

@lrytz it adds -Wconf!

[lrytz]

Nice work, thank you!

[som-snytt]

The inner parens are extraneous.

[SethTisue]

where the method name is [...] a plain identifier ending in an operator character

@som-snytt I ran into this case today over in #8998 and it surprised me:

[error] /Users/tisue/scala.213/src/library/scala/sys/process/ProcessBuilderImpl.scala:139:9: multiarg infix syntax looks like a tuple and will be deprecated
[error]     def lineStream_!(log: ProcessLogger, capacity: Integer): Stream[String] = lineStream(withInput = false, nonZeroException = false, Some(log), capacity)
[error]         ^

at first I assumed this was just a bug, but now I see it was intentional. but I don't understand the design thinking here? lineStream_! doesn't seem like a good example of what you meant to discourage? what is a good example of what you mean to discourage?

[som-snytt]

@SethTisue I don't know, Seth, but I certainly meant to discourage lineStream_!.

It did occur to me to go looking in the ecosystem for usages of names with trailing ops. I was too lazy or something.

sys.process has a background in funky op names that history has not been kind to, but maybe there are other important use cases besides setter_=.

Edit: Seth is persuasive and convinced me that it shouldn't warn.

[jeffrey-aguilera]

This really does not like Slick. It appears that scalac -X does not have an entry for -Xlint:multiarg-infix

[som-snytt]

Multiarg infix applications are not behind the lint flag, only definitions of multiarg operators (which suggest they would be used infix).

It would be friendlier if applications were also behind the lint flag.

Edit: scala/bug#12058