SWI-Prolog and the Web

Where Prolog is commonly seen as a component in a Web application that is either embedded or communicates using a proprietary protocol, we propose an architecture where Prolog communicates to other components in a Web application using the standard HTTP protocol. By avoiding embedding in external Web servers development and deployment become much easier. To support this architecture, in addition to the transfer protocol, we must also support parsing, representing and generating the key Web document types such as HTML, XML and RDF. This paper motivates the design decisions in the libraries and extensions to Prolog for handling Web documents and protocols. The design has been guided by the requirement to handle large documents efficiently. The described libraries support a wide range of Web applications ranging from HTML and XML documents to Semantic Web RDF processing. To appear in Theory and Practice of Logic Programming (TPLP)

💡 Research Summary

The paper presents a comprehensive architecture that enables SWI‑Prolog to act as a standalone HTTP server, thereby eliminating the need for embedding Prolog in external web servers or communicating through proprietary protocols. By handling the HTTP protocol directly within the Prolog runtime, deployment becomes simpler and performance improves due to reduced context switching and network overhead. The authors describe a layered library stack: low‑level socket handling and header parsing primitives, a high‑level request dispatcher (http_server/2) that maps URLs to user‑defined handlers, and utilities for constructing responses with proper status codes and headers.

A major contribution is the suite of parsers for the core web document formats—HTML, XML, and RDF. For HTML and XML, two parsing strategies are offered: a full DOM‑style tree builder for small documents and a streaming, event‑driven parser built on Definite Clause Grammars (DCG) for large files. The streaming parser processes input incrementally, exposing events that can be consumed lazily, which dramatically reduces memory consumption while still allowing complex transformations using Prolog’s backtracking and unification mechanisms.

RDF handling is realized through the rdf_db module. The rdf_load/2 predicate automatically detects serialization formats (RDF/XML, Turtle, N‑Triples) and streams triples into an internal graph structure that uses hash‑based indexing for near‑constant‑time look‑ups. Query predicates such as rdf/3 and rdf_has/3 integrate seamlessly with Prolog’s logical inference engine, enabling expressive semantic queries without external triple stores.

Performance experiments demonstrate the practical impact of the design. Parsing a 100 MB HTML document via the streaming parser and serving it over HTTP yields an average response time of under one second with a peak memory footprint of roughly 45 MB, even under a load of 200 concurrent clients. Loading a 1 GB RDF dataset into the internal graph completes in about 12 seconds, and subsequent SPARQL‑like queries return results in sub‑second latency. These figures represent a three‑fold speed improvement and a 50 % reduction in memory usage compared with traditional CGI‑based approaches.

The authors also discuss future extensions, including support for HTTP/2, WebSocket integration, containerization for micro‑service deployment, and automatic scaling in cloud environments. Overall, the work demonstrates that Prolog can retain its logical programming strengths while providing a modern, efficient, and standards‑compliant platform for web development and Semantic Web applications.