left and right euclidean;
An equivalence relation on a set is a binary relation on that is:
A setoid is a set equipped with an equivalence relation. (However, we should be cautious with this terminology, since the people who typically use setoids begin with an impoverished notion of set and then introduce setoids specifically to fix this, as described below.)
Equivalently, a setoid is a groupoid enriched over the cartesian monoidal category of truth values. Equivalently, a setoid is a groupoid that is 0-truncated. Then the equivalence relation on is a way of making into the set of objects of such a groupoid. Equivalently, a setoid is a (0,1)-category whose each morphism is iso, or a symmetric preordered set.
It may well be useful to consider several possible equivalence relations on a given set. When considering a single equivalence relation once and for all, however, it is common to take the quotient set and use that instead. As a groupoid, any setoid is equivalent to a set in this way (although in the absence of the axiom of choice, it is only a “weak” or ana-equivalence).
Setoids are still important in foundations of mathematics where quotient sets don't always exist and the above equivalence cannot be carried out. However, arguably this is a terminological mismatch, and such people should say ‘set’ where they say ‘setoid’ and something else (such as ‘preset’, ‘type’, or ‘completely presented set’) where they say ‘set’. (See Bishop set and page 9 of these lecture notes.)
A partial equivalence relation is a symmetric and transitive relation.
For the history of the notion of equivalence relation see this MO discussion.