Is it invalid to call Attribute with the same key on the same node/way/relation?

[cldellow]

For example, is this valid, and is the expected result that last-one-in wins?

function node_function(node)
  node:Layer("somelayer")
  node:Attribute('name', 'Unknown')

  if ...
    node:Attribute('name', 'an actual name')
  end
end

I think neither the OMT process.lua nor Shortbread's process.lua do this currently.

Poking about in AttributeStore, it doesn't contemplate that a user would do this, and stores both name pairs. At writing time, I think both get written out in the order of std::map's comparator, which results in the lexicographically biggest value winning -- in the above example, Unknown would be written.

I ask because I'm looking at making some changes to AttributeStore and am wondering if I need to preserve this behaviour, or if anything's fair game so long as it doesn't crash.

[systemed]

I would expect the last one to win - it's essentially what you'd expect if you were to write (in pseudocode)

let a = 5
let a = 3
print(a)

which suggests we're not doing it right at present!

AttributeStore is quite extensively rewritten in #499 so that's probably the code to target.

[cldellow]

Great, thanks!