[Paper Review] Spreadsheets for Stream Partitions and Windows

Martin Hirzel, Rodric Rabbah+3 more|arXiv (Cornell University)|Mar 13, 2015

Advanced Database Systems and Queries2 references3 citations

TL;DR

The paper proposes extending spreadsheets with two new functions—WINDOW and PARTITION—to enable efficient programming of stream processing operators. These extensions allow reactive, stateless computation over variable-sized windows and key-based data partitioning, enabling domain experts to build scalable, real-time streaming applications directly in spreadsheets without low-level coding.

ABSTRACT

We discuss the suitability of spreadsheet processors as tools for programming streaming systems. We argue that, while spreadsheets can function as powerful models for stream operators, their fundamental boundedness limits their scope of application. We propose two extensions to the spreadsheet model and argue their utility in the context of programming streaming systems.

Motivation & Objective

• Address the limitations of traditional spreadsheets in handling unbounded, high-volume streaming data by extending their model to support stream-specific operations.
• Enable domain experts—familiar with spreadsheets but not necessarily with low-level stream programming—to develop and deploy streaming operators independently.
• Overcome the bounded nature of spreadsheets that restricts expressive power for windowing and partitioning in streaming workloads.
• Preserve the reactive, stateless, and composable nature of spreadsheets while enabling scalable, incremental computation over dynamic data streams.

Proposed method

• Introduce a new WINDOW function that allows a single cell to store a dynamic list of values (e.g., last 60 seconds of trades), which can be passed to aggregating functions like SUM, AVERAGE, or COUNT.
• Ensure reactivity by triggering recomputation only when new values enter the window, maintaining responsiveness and performance.
• Design the WINDOW function to enforce type safety by rejecting non-aggregate functions, preventing invalid usage and ensuring predictable semantics.
• Introduce a PARTITION function that behaves like SELECT but creates a separate, isolated computation instance for each distinct key (e.g., stock ticker), enabling per-key processing without code duplication.
• Enforce key consistency across PARTITION functions in a spreadsheet to prevent semantic inconsistencies and support efficient, scalable execution.
• Integrate the extended model into ActiveSheets, a Microsoft Excel-based platform that exports spreadsheet logic to SPL operators for IBM InfoSphere Streams.

Experimental results

Research questions

• RQ1 Can spreadsheets be effectively extended to model and execute real-time stream processing workloads with variable-sized windows?
• RQ2 How can key-based partitioning be modeled in a spreadsheet to support per-key computation without code duplication?
• RQ3 What extensions to the spreadsheet model are necessary to support reactive, incremental computation over unbounded data streams?
• RQ4 Can the reactive and stateless semantics of spreadsheets be preserved while enabling complex stream operations like windowing and partitioning?
• RQ5 To what extent can domain experts develop and deploy scalable streaming applications using only spreadsheet interfaces?

Key findings

• The WINDOW function enables natural and efficient modeling of variable-sized windows (e.g., last minute of data) by allowing cells to hold dynamic lists of values that trigger re-computation on new inputs.
• Aggregating functions like SUM and AVERAGE can be incrementally updated when new values enter the window, preserving performance and responsiveness.
• The PARTITION function allows users to define a single computation for one key (e.g., one stock symbol), which is then automatically replicated per distinct key in the data stream.
• By requiring consistent keys across PARTITION functions, the model ensures correctness and enables efficient, scalable execution without runtime conflicts.
• The extensions maintain the reactive, stateless, and composable nature of spreadsheets, allowing domain experts to build and export complex streaming operators directly in Excel.
• The approach enables rapid development of malleable, reusable streaming logic—such as VWAP-based bargain detection—without requiring low-level programming or external coordination.