1: \begin{abstract}

2:   A famous result by Jeavons, Cohen, and Gyssens shows that every constraint satisfaction problem (CSP) 

3:   where the constraints are preserved by

4:   a \emph{semi-lattice operation} can be solved in polynomial time. This is one of the basic facts for the so-called

5:   \emph{universal-algebraic approach} to a systematic theory

6:   of tractability and hardness in finite domain constraint satisfaction.

7:   

8:   Not surprisingly, the theorem of Jeavons et al.\ fails for arbitrary infinite domain CSPs. 

9:   Many CSPs of practical interest, though, and in particular

10:   those CSPs that are motivated by \emph{qualitative reasoning calculi} from Artificial Intelligence, 

11:   can be formulated with constraint languages that are rather well-behaved from a model-theoretic point of view. 

12:   In particular, the automorphism group of these constraint languages tends to be \emph{large} in the sense that 

13:   the number of orbits of $n$-subsets of the automorphism group is bounded by some function in $n$.

14:   %the automorphism group has \emph{few} orbits of $n$-subsets. 

15:   

16:   In this paper we present a generalization of the theorem by Jeavons et al.\ to infinite

17:   domain CSPs where the number of orbits of $n$-subsets grows sub-exponentially in $n$, and prove that

18:   preservation under a semi-lattice operation for such CSPs implies polynomial-time tractability. 

19:   Unlike the result of Jeavons et al., this includes many CSPs that cannot be solved by Datalog.

20: \end{abstract}

21: