Program synthesis using abstraction refinement

Abstract

We present a new approach to example-guided program synthesis based on counterexample-guided abstraction refinement. Our method uses the abstract semantics of the underlying DSL to find a program P whose abstract behavior satisfies the examples. However, since program P may be spurious with respect to the concrete semantics, our approach iteratively refines the abstraction until we either find a program that satisfies the examples or prove that no such DSL program exists. Because many programs have the same input-output behavior in terms of their abstract semantics, this synthesis methodology significantly reduces the search space compared to existing techniques that use purely concrete semantics.While synthesis using abstraction refinement (SYNGAR) could be implemented in different settings, we propose a refinement-based synthesis algorithm that uses abstract finite tree automata (AFTA). Our technique uses a coarse initial program abstraction to construct an initial AFTA, which is iteratively refined by constructing a proof of incorrectness of any spurious program. In addition to ruling out the spurious program accepted by the previous AFTA, proofs of incorrectness are also useful for ruling out many other spurious programs.We implement these ideas in a framework called Blaze, which can be instantiated in different domains by providing a suitable DSL and its corresponding concrete and abstract semantics. We have used the Blaze framework to build synthesizers for string and matrix transformations, and we compare Blaze with existing techniques. Our results for the string domain show that Blaze compares favorably with FlashFill, a domainspecific synthesizer that is now deployed in Microsoft PowerShell. In the context of matrix manipulations, we compare Blaze against Prose, a state-of-the-art general-purpose VSA-based synthesizer, and show that Blaze results in a 90x speed-up over Prose. In both application domains, Blaze also consistently improves upon the performance of two other existing techniques by at least an order of magnitude. Abstract Synthesizer Checker Refiner candidate program Failure (no solution) Synthesized program counterexample & spurious program new abstraction End-users Domain expert Examples DSL w/ abstract semantics SYNGAR It can be shown that the language of the CFTA constructed from Fig. 3 is exactly the set of abstract syntax trees (ASTs) of DSL programs that are consistent with the input-output examples. 1 Hence, once we construct such a CFTA, the synthesis task boils down to finding an AST that is accepted by the automaton. However, since there are typically many ASTs accepted by the CFTA, one can use heuristics to identify the "best" program that satisfies the input-output examples. 1 The proof can be found in the extended version of this paper [Wang et al. 2017a].AFTA constr uction AST annotated with old abstract values AST annotated with concrete values Pr oof constr uction Rank Predicates Predicates 2 AST annotated with old abstract values AST annotated with concrete values Pr oof co...

Synthesizing database programs for schema refactoring

“…This paper is related to a long line of recent work on program synthesis [1, 2, 6, 7, 22, 26-28, 32, 35, 36, 41, 42, 45-49, 53, 55]. While the goal of program synthesis is always to produce a program that satisfies the given specification, different synthesizers use different forms of specifications, including input-output examples [6,22,26,42,53], logical constraints [47][48][49], refinement types [41], or a reference implementation [27,35,45]. Our technique belongs in the latter category in that it uses the original implementation as the specification.…”

Visualization by example

“…Program analysis for program synthesis. Given the large search space that must be explored by program synthesizers, a common trick is to perform lightweight program analysis to prune the search space [Feser et al 2015;Polikarpova et al 2016;Wang et al 2017bWang et al , 2018b. The particular flavor of program analysis varies between different synthesizers and ranges from domain-specific deduction [Feser et al 2015;Wang et al 2017b] to abstract interpretation [Wang et al 2018b] to SMT-based reasoning [Feng et al 2018;Polikarpova et al 2016].…”

“…The particular flavor of program analysis varies between different synthesizers and ranges from domain-specific deduction [Feser et al 2015;Wang et al 2017b] to abstract interpretation [Wang et al 2018b] to SMT-based reasoning [Feng et al 2018;Polikarpova et al 2016]. Furthermore, some of these techniques leverage program analysis to construct a compact version space [Polozov and Gulwani 2015;Wang et al 2018b] while others use it to prune partial programs in enumerative search [Feng et al 2018;Wang et al 2017b]. Similar to these efforts, we also use program analysis to prove infeasibility of partial programs but with two key differences: First, our analysis is tailored to table transformation programs and infers inclusion constraints between tables.…”

Syntax-Guided Termination Analysis

“…SyGuS [2] is applied to various tasks related to program synthesis, e.g., [13,17,25,33,35,41]. However, the formal grammar in those applications is typically given or constructed from user-provided examples.…”