Synthesizing Analytical SQL Queries from Computation Demonstration
summary
Proposes a new specification called computation demonstration where users provide cell‑level expressions and optional ellipses instead of only final values, and defines computational consistency (preserve cell provenance, allow partial matches, allow parameter ellipses). Implements a prototype synthesizer that uses abstract provenance analysis to prune programs during search. Evaluated on a benchmark of 80 analytic queries (60 web tasks + 20 TPC‑DS): significantly faster on hard tasks, visits far fewer candidates; a user study with 13 participants on 6 tasks shows faster creation for hard tasks and reduced effort using partial expressions.
AI-written summary · about
related talks
48mCompositional, Expressive, Performant Database Queries: Choose 3
14mAutomated Translation of Functional Big Data Queries to SQL
15mSpeeding up Symbolic Reasoning for Relational Queries
35mQuerying the queries: SQL Metaprogramming in Python
14mDetecting Logic Bugs in Database Engines via Equivalent Expression Transformation
41mOptimizing Relational Programs I
20mPolygon: Symbolic Reasoning for SQL using Conflict-Driven Under-Approximation Search
32mBecause SQL is everywhere but you don't want to use it