Example-guided synthesis of relational queries

Abstract

Program synthesis tasks are commonly specified via inputoutput examples. Existing enumerative techniques for such tasks are primarily guided by program syntax and only make indirect use of the examples. We identify a class of synthesis algorithms for programming-by-examples, which we call Example-Guided Synthesis (EGS), that exploits latent structure in the provided examples while generating candidate programs. We present an instance of EGS for the synthesis of relational queries and evaluate it on 86 tasks from three application domains: knowledge discovery, program analysis, and database querying. Our evaluation shows that EGS outperforms state-of-the-art synthesizers based on enumerative search, constraint solving, and hybrid techniques in terms of synthesis time, quality of synthesized programs, and ability to prove unrealizability. CCS Concepts:• Information systems → Relational database query languages; • Software and its engineering → Automatic programming; Programming by example; • Theory of computation → Constraint and logic programming.

Learning formulas in finite variable logics

“…Synthesis from examples, or programming by examples (PBE), has been active in recent years and has seen successes in practice (e.g., [Polozov and Gulwani 2015]). In PBE, the goal is to synthesize a program consistent with a set of input-output examples; several domains have been explored, e.g., synthesis of database queries from examples [Shen et al 2014;Thakkar et al 2021;Wang et al 2017a] and from analysis of database-backed application code [Cheung et al 2013], synthesis of data completion scripts [Wang et al 2017c], data structure transformations [Feser et al 2015], and typed functional programs [Osera and Zdancewic 2015;Polikarpova et al 2016]. A common approach to PBE involves version space algebra [Mitchell 1982], where the idea is to capture the set of all programs that work on each example in a compact representation and then intersect the sets for each example to represent programs consistent with all examples (e.g., see [Gulwani 2011]).…”

Synthesizing Analytical SQL Queries from Computation Demonstration

“…Relational Query Synthesis. Due to the importance of relational queries in the database industry and the programming challenges associated with them, many relational query synthesizers have been developed [29,31,36,37,39,47]. Among them, Scythe [39] supports standard SQL queries with subqueries and aggregations; EGS [36], ProSynth [29] and Alps [31] are datalog query synthesizers that synthesizes datalog queries using syntax and provenance guided search from input-output examples; Morpheus [12] and Viser [40] are table transformation synthesizers for data wrangling.…”

“…Sickle's focus is the computation-rich analytical SQL. Analytical SQL are not expressible in logic-based synthesizers [36] that requires relational queries to be "pure" (free of computations); algorithms for standard SQL [39,40] are insufficient due to lack of support for analytical operators (Section 5). As Sickle focuses on analytical computations, Sickle is not specially designed for data wrangling [12] or knowledge base querying tasks [29]; there is an opportunity to combine Sickle with Morpheus [13] or ProSynth [29] for tasks requiring all types of relational query features.…”