Yiding Cui

Results 65 comments of Yiding Cui

![img_v3_02ar_8959d11b-7399-42f9-b5a7-42b6c76fbc6g](https://github.com/pingcap/tidb/assets/7846227/92ab3a30-3907-4af1-99ce-d74d3addc2f9) ![img_v3_02ar_caf1d1de-3c40-4f75-aa83-404dd4f8e0fg](https://github.com/pingcap/tidb/assets/7846227/3f602a75-0003-4297-8bb9-ccf55dca57bb) ![image](https://github.com/pingcap/tidb/assets/7846227/678abb63-133b-44a5-bb80-5b96862bb23e) The output columns of the top plan are directly passed as the `parentUsedCols` and then the `parentUsedCols` is modified during the pruning. So the output columns changed...

To support the physical cost during join reorder. Introducing the memo struct of cascades is necessary. We need to bring memoization/dynamic programming into the logical optimization phase.

Table `t` has few rows, and it can apply index join with `mk`(The one has 7.480087e+06 rows). So the cost of `t join mk` is a small value. But we...

The result should still be better than 7.1. So we merge it first.

Like itself is designed and used for the single pattern matching problem. We must be aware that single-pattern matching and multiple-pattern matching are different problems. They need different data structures...

It's about the filter's null-rejective testing. We get a false-positive result for the `(((ref_0.c6 is null) and false) and null) xor (true or ('111' >= the subquery col)))`

> @winoros we don't get false positive result for `(((ref_0.c6 is null) and false) and null) xor (true or ('111' >= the subquery col)))`. First of all, we have a...