Publication Date

Fall 2021

Degree Type

Master's Project

Degree Name

Master of Science (MS)


Computer Science

First Advisor

Chris Pollett

Second Advisor

Katerina Potika

Third Advisor

Ben Reed


WARC Database, node, Javascript, Linear Hash Table


WARC files are central to internet preservation projects. They contain the raw resources of web crawled data and can be used to create windows into the past of web pages at the time they were accessed. Yet there are few tools that manipulate WARC files outside of basic parsing. The creation of our tool WARC-KIT gives users in the Node.js JavaScript environment, a tool kit to interact with and manipulate WARC files.

Included with WARC-KIT is a WARC parsing tool known as WARCFilter that can be used standalone tool to parse, filter, and create new WARC files. WARCFilter can also, create CDX index files on the WARC files, parse existing CDX files, or even generate webgraph datasets for graph analysis algorithms. Aside from WARCFilter, WARC-KIT includes a custom on disk database system implemented with an underlying Linear Hash Table data structure. The database system is the first of its kind as a JavaScript only on disk document store. The overall main application of WARC-KIT is that it allows users to create custom indices upon collections of WARC files. After creating an index on a WARC collections, users are then query their collection using the GraphQL query language to retrieve desired WARC records.

Experiments with WARCFilter on a WARC dataset composed of 238,000 WARC records demonstrates that utilizing CDX index files speeds WARC record filtering around ten to twenty times faster than raw WARC parsing. Database timing tests with the JavaScript Linear Hash Table database system displayed twice as fast insertion and retrieval operations than a similar Rust implemented Linear Hash Table database. Experiments with the overall WARC-KIT application on the same 238,000 WARC record dataset exhibited consistent query times for different complex queries.