• Streams from a java.io.Reader, so can process very large files
• Unlike java.io.BufferedReader.readLine(), retains the line-break at the end of the line, whether it's UNIX, Windows, etc
• Implements the scala.collection.Iterator trait. Lines are not retained in memory after being returned by next().
• Limits memory consumption to a single ring-buffer the lives the lifetime of the iterator, and, for lines that loop round the ring-buffer, temporary StringBuilder instances that are garbage-collected after the line is returned.
• Reading a line using next() will return as soon as the java.io.Reader has returned enough characters - will at most only attempt to fill the ring-buffer.

For my purposes, the first two requirements were the most important. If I hadn't wanted to retain the line-breaks, I'd probably have used java.io.BufferedReader.readLine().

libraryDependencies += "com.madgag" %% "line-break-preserving-line-splitting" % "[version]"

import com.madgag.linesplitting._

val splitLines = {
  val reader: java.io.Reader = new java.io.StringReader("Foo\nBar\rBoo")

  val iter = new LineBreakPreservingIterator(reader)
  
  iter.toSeq
}.toSeq
// splitLines: Seq[String] = List(
//   """Foo
// """,
//   """Bar
""",
//   "Boo"
// )

splitLines.map(_.replace("\n", "[LF]").replace("\r", "[CR]"))
// res0: Seq[String] = List("Foo[LF]", "Bar[CR]", "Boo")