Translating Relational Queries into Spreadsheets

TL;DR

This paper shows that spreadsheets can implement all SQL data transformations using formulas alone, enabling database-like queries without additional tools, and introduces efficient sorting algorithms within spreadsheets.

Contribution

It presents a query compiler translating SQL into spreadsheet formulas, allowing complex data operations without macros or external database systems.

Findings

Spreadsheet can implement all SQL data transformations.

A new $O(n ext{log}^2 n)$ sorting algorithm in spreadsheets.

Spreadsheets can perform graph algorithms like DFS and BFS.

Abstract

Spreadsheets are among the most commonly used applications for data management and analysis. Perhaps they are even among the most widely used computer applications of all kinds. They combine in a natural and intuitive way data processing with very diverse supplementary features: statistical functions, visualization tools, pivot tables, pivot charts, linear programming solvers, Web queries periodically downloading data from external sources, etc. However, the spreadsheet paradigm of computation still lacks sufficient analysis. In this article we demonstrate that a spreadsheet can implement all data transformations definable in SQL, without any use of macros or built-in programming languages, merely by utilizing spreadsheet formulas. We provide a query compiler, which translates any given SQL query into a worksheet of the same semantics, including NULL values. Thereby database operations become available to the users who do not want to migrate to a database. They can define their queries using a high-level language and then get their execution plans in a plain vanilla spreadsheet. No sophisticated database system, no spreadsheet plugins or macros are needed. The functions available in spreadsheets impose severe limitations on the algorithms one can implement. In this paper we offer $O(n\log^2n)$ sorting spreadsheet, but using a non-constant number of rows, improving on the previously known $O(n^2)$ ones. It is therefore surprising, that a spreadsheet can implement, as we demonstrate, Depth-First-Search and Breadth-First-Search on graphs, thereby reaching beyond queries definable in SQL-92.