Back to search

Cornell University

Domain-Specific Languages for Ad Hoc Data Processing

Abstract

dc:description.abstract

Ad hoc data is everywhere. There is a plethora of data formats in use, which, due to the specificity of their domain, lack the processing tools that we might otherwise take for granted. We call such data formats, and the data that they represent, ad hoc. Ad hoc data is usually stored in file systems and can be organized into larger structures that we call filestores, collections of files and folders along with the properties between them. This dissertation aims to support the usage of ad hoc filestores. We design domain-specific languages for processing filestores with increasingly complex requirements, building on previous work on PADS and Forest. These existing systems accept declarative specifications of ad hoc data formats, as single files and filestores respectively. From these specifications, they generate tools for loading, storing, and validating the data. Unfortunately, Forest does not adequately deal with large filestores, or cost control in general. Nor does it offer support for correctly managing concurrent operations, which are common in file systems. This dissertation offers solutions to these problems. We first introduce Incremental Forest, a domain-specific language and system that enables incremental processing of filestores. This language offers a new mechanism, a delay construct, for explicitly controlling the costs of loading and storing in filestores. Incremental Forest comes with a customizable cost model, which guarantees that a wide class of costs monotonically decrease as delays increase. Our next system, Transactional Forest, does away with the delays from Incremental Forest, opting to use an entirely new interface language and abstraction, which offer automatic incrementality. Additionally, Transactional Forest leverages this abstraction, a zipper, to provide simple, provably correct serializable transactions using optimistic concurrency control. Finally, the Zipper File System goes beyond designing a domain-specific language, targeting the file system itself to provide deeper control with respect to other users. The Zipper File System uses the ideas from Transactional Forest to provide serializable transactions in a zipper-based file system. It also comes with a translation from POSIX that theoretically allows standard applications to be run on our file system, without changes. Taken together, these systems use domain-specific languages to enable users to efficiently and correctly manage ad hoc filestores in concurrent settings.

Degree

thesis:*
Name thesis:degree_name
Ph. D., Computer Science
Level thesis:degree_level
Doctor of Philosophy
Discipline thesis:degree_discipline
Computer Science
Grantor
Cornell University
Year dc:date.issued
2020

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • DiLorenzo, Jonathan
Chair dc:contributor.chair
  • Foster, Nate
Committee members dc:contributor.committeemember
  • Kleinberg, Robert David
  • Kozen, Dexter
  • Schneider, Fred Barry

Subjects

dc:subject × 6

Rights

dc:rights
Statement dc:rights
  • Attribution 4.0 International
Language dc:language.iso
en

Identifiers

dc:identifier.*
Dc Identifier Bibid
13312067
Dc Identifier Other
ProQuest Submission ID: 12376
ProQuest Publication ID: 28260847

Chain of custody

source
Harvested from
Cornell University
Base URL
ecommons.cornell.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

DiLorenzo, Jonathan. Domain-Specific Languages for Ad Hoc Data Processing. Doctor of Philosophy thesis, Cornell University, 2020. https://hdl.handle.net/1813/103357