ability to evaluate code via elm-repl-push{-decl} without switching to the REPL buffer

[berkan]

This change allows elm-repl-push and elm-repl-push-decl to work without switching to the REPL buffer.

The use-case I'm addressing is to evaluate code directly from a source code buffer, with minimum disruptions.

The details of the change are:

• elm-repl-push & elm-repl-push-decl accept a universal argument which controls whether or not to switch to the REPL buffer after the push.

  The default behaviour is:

  • if the universal argument is not given or nil, switch to the REPL buffer (the same as the current behaviour)
  • if the universal argument is non-nil, stay in the current buffer after the push

• Setting the variable elm-interactive-flip-arg-switch-behaviour to non-nil reverses the default behaviour above.
  So, anybody who didn't want to switch to the REPL buffer after a push and don't want to use the universal argument all the time can set this to t.

• After a push, functions in elm-interactive-after-push-hook are run with the result of the last push

  • Currently this hook contains elm--print-result which prints the first line of a possibly multi-line string in the echo area (this is so that the echo area doesn't grow and create an unnecessary visual distraction)

[purcell]

Thanks. This solution is greatly overthinking things IMO, and adds unnecessary complexity: it would be conceptually much simpler to just add a separate command that evaluates the selected region and prints the result in the minibuffer.