The core of Semantic Web programming is SPARQL [1]. To increase the productivity of Semantic Web programming, enhancing the reusability of SPARQL queries is key. Especially, the readability of code is essential for reusability. However, reformatting the SPARQL query is not a trivial issue. We need to parse the input SPARQL and output it appropriately to reformat it. Parsing a SPARQL query is a complex task because the SPARQL specification has a complex grammar, which is comprised of 173 rules in the format of EBNF. Thus, it is difficult to implement a parser in a conventional sequential programming language.
Several programming languages including JavaScript support Parsing Expression Grammar (PEG) [2] for describing definitions of a grammar in a declarative manner to generate a parser. Here, we aim to implement a formatter that fully supports SPARQL 1.1 by using PEG for command-line use and website development.
We used PEG.js (https://pegjs.org/) to generate a SPARQL 1.1 parser. 173 rules expressed in EBNF were extracted from SPARQL 1.1 Query Language specification [1] and translated into PEG rules for constructing abstract syntax trees (ASTs). An excerpt from the EBNF rules of SPARQL and the corresponding PEG rules for constructing the AST is shown in Figure 1. Notably, white spaces and comments in SPARQL are not included in EBNF rules. As comments are often inserted in SPARQL queries in practice, comments besides white spaces are also expressed in our PEG expressions. Comments are kept separate from the AST. The position from the beginning of the SPARQL input was obtained and included in the returned values.
The formatter was implemented in JavaScript, which accepts an AST of SPARQL and outputs each element of the SPARQL uniformly. The position of each element is compared with the positions of comments, so that the comments are added at the appropriate positions.
91 test queries were extracted from the SPARQL 1.1 Query Language [1] specifications, and 16 from SPARQL 1.1 Update [3]. All those queries were formatted and then manually curated for use as test cases.
The SPARQL formatter is published as an npm library, so that it is easy to install and use in a Node.js environment as the sparql-formatter command. The Docker version is also available and published on Docker Hub.
The JavaScript code of the parser and formatter has been bundled and made available on Content Delivery Network (CDN), so that users can incorporate the SPARQL formatter into their websites. Figure 2 shows a website using the SPARQL formatter. All the queries extracted from the SPARQL 1.1 specifications are available as example queries on the website.
The AST of a SPARQL query is represented internally in JSON and can be reused for purposes other than formatting. Figure 3 shows the AST obtained for an example SPARQL query, "SELECT * WHERE { ?s ?p ?o } LIMIT 10 ".
By generating a SPARQL parser via PEG.js, the SPARQL formatter was implemented. This work contributes to the increased reusability of SPARQL queries. The queries can be formatted in a uniform way, thus the readability will be increased. While other efforts to implement SPARQL parsers have been made in projects such as RSFStore-js [4] and SPARQL.js (https://github.com/RubenVerborgh/SPARQL.js), they could not format queries as intended. Notably, by adding a JSON-LD context to the obtained AST for a SPARQL, the query can be viewed as RDF, thus the SPARQL queries themselves can be treated as RDF resources. It may contribute to the FAIRification [5] of the Semantic Web queries. The PEG.js code developed for SPARQL 1.1 can be modified for extended specifications of SPARQL [6] or other subgraph matching problems on knowledge graphs [7].