Query Reverse Engineering of Pre-deleted Uncorrelated Operators

Recent years have seen an increasing reliance on data processing to accomplish work tasks. However, many users do not have the programming background to write complex programs, especially query statements. Query Reverse Engineering solves the problem of deriving query statements from the database and the desired output table in reverse. SQUARES, which is based on Domain-Specific Languages (DSL), is one of the most advanced models in this field. However, the existence of uncorrelated DSL operators constrains the synthesis efficiency. This paper proposes PdQRE based on SQUARES, which improves efficiency by predicting whether DSL operators are correlated with the query statement and pre-deleting uncorrelated operators. On the test-55 dataset, the synthesis rate of PdQRE improved from 80.0% to 89.1%, and the average synthesis time was reduced from 251 s to 127 s compared to SQUARES. Comparison with Scythe et al. in the Recent posts dataset shows that PdQRE outperforms other models in Query Synthesis.