Well, the standard way to implement Distinct() on an IEnumerable<> would be to use a HashSet, with two variations: 1) use it strictly as a test (if (hs.Add(item)) yield return item), or 2) read the entire enumerable in to the HashSet, and then iterator thru the set. The first method would be stable; the second, not necessarily.

You're correct as far as the documentation saying that it's "unordered", and other providers are technically free to implement it in another way, but when dealing with a stable data structure such as a string, in a single-threaded context, in a lazily-evaluated LINQ chain, there is simply no other efficient way to yield the items from the enumerable than the original sequence. The only other way is to consume the entire thing and sort it and eliminate duplicates that way, but that both takes more immediate memory and delays the pipeline, so it's unlikely this implementation would change. There might be a theoretical potential for very large data sets to implement some sort of parallel distributed Distinct which then merges the sets at the end, and in that case order might not be preserved, but I'm not sure if that's performant, and in any case we don't use .AsParallel() here, so that's not an option either.

tl;dr: It may be undocumented behavior, but it's very predictable undocumented behavior in this particular case.

ty mate. That's exactly what I thought too.

strings are an IEnumerable. The ToCharArray() is unnecessary.

What makes you think the Replace() would leave spaces on the start or end?

It's not redundant, just bad form. input[i] is a char, " " is a string. The line should be "if (input[i] == ' ')" or better "if (Char.IsWhiteSpace(input[i]))"