I can enter the same keyword argument twice, and the second over-rules the first

[colintbowers]

A short example:

f(x::Int ; kw::Int=0) = x * kw
f(2)
f(2, kw=3) #evaluates to 6
f(2, kw=3, kw=4) #evaluates to 8

I'm on v0.4, but I asked about it on the julia-users list here, and was told you can still do this in v0.5. The link also contains an argument for why this might be considered a feature and not a bug. Personally, it feels more like a bug to me - but I am a very small voice.

I stumbled onto this when typing out several keyword arguments, and my mind was still on the previous keyword as I typed the next one, and so I accidentally typed it twice. Easy to do, and it took me quite a while to work out where I'd gone wrong.

[JeffBezanson]

It seems pretty reasonable to make this a syntax error.

[martinholters]

As mentioned on the list, the current behavior might be useful with splatting for something like foo(;defaults..., overrides..) to work, but the same could be achieved with an explicit merge, of course.
Well, not exactly the same: If foo is defined as foo(;kwargs...), it can actually tell whether a value is given multiple times. This might be potentially useful if one actually wants to allow multiple values to one kwarg, but wants to avoid a collection for some reason E.g. process_file(input="foo.txt", input="bar.txt") could process both files. But this feels so fragile compared to "normal" kwarg behavior that I'm also in favor of making it a syntax error.

[ivarne]

Somewhat relevant discussion in #4916

[toivoh]

+1 for making it a syntax error.