E-Graph Union and Intersection
1  E-Graphs
2  E-Graphs and Proofs
3  E-Graph Union
4  E-Graph Intersection
4.1  Sidenote:   Existing E-Graph Intersection Paper
5  Runtime Analysis
8.5

E-Graph Union and Intersection

1  E-Graphs

E-Graphs are a data structure for representing equalities between terms efficiently. This post assumes you know what an e-graph is, so here’s a great post about them.

2  E-Graphs and Proofs

E-Graph implementations (including egg) typically provide a way to generate a proof certificate for why two terms are equal. For example, the e-graph may know a + b is equal to b + a using an identity x + y = y + x. The proof would say it used the identity to infer the specific terms a + b and b + a are equal. Each particular use of the rule results in an equality like a + b = b + a. You can think of a proof certificate as a set of equalities on specific terms E = \{ T_1 = T_2, T_3 = T_4, ... \}.

E-Graphs compute the congruence closure over terms:

Thm1: If you insert a set of equalities E into the e-graph, then it will say two terms T_i and T_j are equal iff there is a valid proof between them (potentially using congruence).

We can also do the reverse: given an e-graph, we can extract back out the set of equalities that generate the e-graph. There are multiple choices for this set of equalities, but the simplest choice is the terms originally inserted into the e-graph either directly or by axiom instantiations. This is the key idea behind this post: e-graphs are just a way to prove things are equal based on a set of equalities. Using this observation, we can predict what an e-graph can prove.

3  E-Graph Union

Let’s write the union of two e-graph E_1 \cup E_2 = E_3. Intuitively, E_3 should represent terms who are equal due to a combination of facts from E_1 and E_2. More formally, any two terms T_1 and T_2 should be equal in E_3 iff there is a valid proof that they are equal involving only facts from E_1 and E_2.

There’s a simple algorithm to union e-graphs, and it relies on the set of equations that originally generated the e-graph \{ T_1 = T_2, T_3 = T_4, ... \}. First, make a new e-graph E_3. Now assert every equality T_i = T_j from E_1 in E_3. Similarly, assert every equality T_k = T_m from E_2 in E_3. Do this by adding the two terms to E_3 and unioning them in the e-graph.

Here’s some rust code that does it for egg (though it’s not merged into main yet):

By our Thm1, this new e-graph will tell you when two terms are equal because of a combination of facts from each e-graph.

4  E-Graph Intersection

E-Graph intersection is only slightly more complex than the union. We would like the intersection of two e-graphs to capture equalities between terms that are true in both e-graphs.

To guarantee this property, we check which equalities are true in both E_1 and E_2. For every equality T_i = T_j in E_1, insert T_i and T_j into E_2 and check if they are equal. If they are, assert the equality in E_3.

Here’s some code that performs the intersection:

Note that in the code above, we need to maintain congruence closure in the second e-graph by calling the rebuild function. It’s crucial so that we can query it for which equalities are true.

The resulting e-graph captures the intersection perfectly. Any two terms T_1 and T_2 that are equal in the intersection has a proof in E_1. This proof contains only facts which are also true in E_2. We added these facts to E_3, so they must also have a proof in E_3.

4.1  Sidenote: Existing E-Graph Intersection Paper

E-Graph intersection is described in this paper, but the paper offloads some of the problem. Instead of guaranteeing anything about the resulting e-graph, it only guarantees that the resulting e-graph will give the correct result for some set of "important" terms I that you specify in advance. The paper notes that the resulting e-graph E_3 cannot represent all interesting terms since that could be infinite. The insight of this blog post is that choosing the right I above (by filtering to facts that are true in both e-graphs) results in an e-graph that gives the correct answer for any terms inserted in the future.

5  Runtime Analysis

If we assume that all the equalities in E_1 and E_2 are bounded in size by some constant, then the overall runtime of union and intersection is O(n \log{n}), the run time of maintaining congruence closure for n equalities (see this paper). We can, however, get a faster implementation by carefully tracking the new identities of sub-terms in our target e-graph. This saves us the additional cost of copying the same sub-terms over to the new e-graph each time.