Huiming Wang, Zhaodonghui Li, Liying Cheng, De Wen Soh, Lidong Bing. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Query rewrite, which aims to improve query efficiency by altering an SQL query's structure without changing its result, has been an important research problem. In order to maintain equivalence between the rewritten query and the original one during rewriting, traditional query rewrite methods always rewrite the queries following certain rewrite rules. However, some problems still remain. First, existing methods of finding the optimal choice or sequence of rewrite rules are still limited and the process always costs a lot of resources. Methods involving discovering new rewrite rules typically require complicated proofs of structural logic or extensive user interactions. Second, current query rewrite methods usually rely highly on DBMS cost estimators which are often not accurate. In this paper, we address these problems by proposing a novel query rewrite method named LLM-R 2 , which leverages a large language model (LLM) to recommend rewrite rules for a database rewrite system. To further enhance the inference ability of the LLM in recommending rewrite rules, we train a contrastive model using a curriculum-based approach to learn query representations and select effective query demonstrations for the LLM. Experimental results show that our method significantly improves the query execution efficiency and outperforms the baseline methods. In addition, our method exhibits high robustness across different datasets.

Query plan is widely used as input in machine learning for databases (ML4DB) research, with query plan representation as a critical step. However, existing studies typically focus on one task, and propose a novel design to represent query plans along with a ML4DB framework, without comparing with other representation methods designed for a different task. This raises a critical question: How do we select a query plan representation method in a ML4DB system? To address this question, we perform a comparative study on ten representation methods on three distinct ML4DB tasks: cost estimation, index selection and query optimization. Our extensive experiments not only verify the interchangeability of representation methods across different tasks, but also identify consistently high-performing models. Further, we dissect the query plan representation into two core components: feature encoding and tree model, and evaluate the impact of design choices for each in different scenarios. Our results show that the findings for tasks optimizing absolute errors are different from findings for tasks optimizing relative errors. Some findings challenge widely-held assumptions, i.e., one finding shows that tree models do not significantly impact cost estimation results, but only play a significant role to optimize relative performance. Practical guidelines and future directions are provided based on the findings of the study.