I agree that this is problematic. Our Dicts and hashing in general have some rather large issues at the moment, which is why I'm working on the state of numeric hashing right now. This is an area that we'll get straightened out in 0.4.
On Thu, Mar 6, 2014 at 10:25 AM, David P. Sanders <[email protected]>wrote: > Hi, > > I was recently bitten by this: > > julia> s = Set(3, 4, 5) > Set{Int64}({5, 4,3}) > > julia> (1, 1) in s > false > > I.e., I had changed the type stored in a Set from a pair to an integer, > and had a strange bug for several days, which I finally traced to this. > I would much prefer that `in` give an error if the wrong type is used, > rather than 'false'! > Is there a good reason for wanting this behaviour? > > Thanks, > David. >
