Elixir already has Map.merge/2 and Map.merge/3. The latter allows to pass a function to resolve conflicts between the two maps.

Why do you believe that an infix operator would make maps more usable? Also, how would this be used with structs?

In my opinion, I do not find the new operator to enhance readability. I would prefer something like:

defmodule MapMerge do
  def a ||| b do
    Map.merge(a, b)
  end
end

import MapMerge
%{foo: 1, bar: 2} ||| %{bar: "overridden", baz: 4}

which is already possible without changing the core language.

WOW! I did not know you could just def infix operators like that. I expected more macro-ish overhead. That’s great!

Yes! This completely works. The only things you need to be aware of:

1. Only a small subset of operators you can conceive are recognized as such by the compiler.
2. Of these, most are already taken by Kernel. Only \\, <-, |, ~>>, <<~, ~>, <~, <~>, <|>, <<<, >>>, |||, &&&, and ^^^ are not used by default (some of these are used by the Bitwise module) and are user-overridable. (you can use the others, but then when someone uses your library they need to use e.g. import Kernel, except: [ +: 2])
3. You cannot alter the operator precedence of an operator, this is set by the compiler. See the Operator Table to check what they are.
4. Overloaded operators usually do not work in guard clauses, unless you build it as a macro using only Erlang BIFS. For most useful operations, this is impossible.
5. … and finally, most syntax highlighters don’t highlight them properly yet

Oh, and an aside: all recognized operators are valid ways to start a symbol. So :<-> is fine, as is [&&&: 3].