Page History

This optimization tries to intersect multiple secondary indexes if the select conditions introduce them. Previously, as in the IntroduceSelectAccessMethodRule, we would pick the first index to contribute to the access path when there were multiple indexes available. Due to the lack of statistical information, the first one may not be the best choice. Moreover, even we chose the index of the lowest selectivity, it still may not be the best solution. Because we can further reduce the selectivity by intersecting it with the other secondary indexes. Having intersection into the plan will avoid the worst path. Furthermore, if we have statistical information later we can keep improving the decision by whether to introduce the intersection or not.

Optimization Rule

The logical changes are in the IntroduceSelectAccessMethodRule. After we analyzed the interesting functions and indexes, we pair them up as one to one mapping. (E.g. BTreeAccessMethod -> BTreeIndex On Salary.).

...

Because we only have one disk. First, the Tweet.create_at path has to wait for the User.create_at to finish a frame to operate the intersection. These two index search is battling the disk read. Second, although the intersection itself can be finished as long as one of the input is done, we can not stop the other index scan based on our push model. Hence, the primary search is also competing on the disk resource with the two index searches. 

...

Intersect Unclustered Secondary Indexes

As shown in the previous result, the index on Tweet.create_at is a clustered secondary index, which is a special case for the secondary index. To test a more general case, we create an RTree on the Tweet.place.boudingbox which is a rectangle area. We create a circle area around LA county. By increasing the radius, we can increase the selectivity of that RTree. The query is as below

...