Shape Expressions (ShEx) was de ned as a human-readable and concise language to describe and validate RDF. In the last years, the usage of ShEx has grown and more functionalities are being demanded. One such functionality is to ensure interoperability between ShEx schemas and domain models in programming languages. In this paper, we present ShEx-Lite, a tabular based subset of ShEx that allows to generate domain object models in di erent object-oriented languages. Although the current system generates Java and Python, it o ers a public interface so anyone can implement code generation in other programming languages. The system has been employed in a work ow where the shape expressions are used both to de ne constraints over an ontology and to generate domain objects that will be part of a clean architecture style.
Since the appearance of Shape Expressions Language [
Commons License Attribution 4.0 International (CC BY 4.0). 3 https://www.um.es/web/hercules 4 https://www.weso.es/shex-lite/
The ShEx language is quite powerful and allows for a high degree of expressiveness like disjunctions, negations, etc. which make generating domain object models from shapes a non-trivial task. However the ShEx-lite language is simpler and it permits mainly tabular based schemas based on constraints of type PROPERTY TYPE CARDINALITY referred to as simple triple constraints. For instance, Fig. 1 shows an schema example that de nes the properties of a Person model. In this particular case a Person is de ned as a property :name of type xsd:string and cardinality 1 (Default one) as well as a second property :knows of type @:Person and cardinality 0-n, which represents the set of people known by the current Person.
Person.shexl 1 # Prefixes ... 2 : Person { 3 : name xsd: string ; 4 : knows @: Person * 5 }
Person.java 1 // Imports ... 2 public class Person { 3 private String name ; 4 private List <Person > knows ; 5 // Constructor ... 6 // Getters and Setters ...
7 }
Once de ned the input of ShEx-Lite it is easier to explain how the system generated domain object models. Basically, the comunication with the system is done through a CLI tool that is provided. In this tool the users can de ne several options but the one that is in the scope of this paper is --java-pkg=STRING which triggers the java code generation and generates the target object in the speci ed package.
For example, for the input java -jar shexlc.jar --java-pkg=demo person.shexl where the person.shexl le corresponds to the schema de ned at Fig. 1 ShEx-Lite generates a single java class with the code that appears at the Person.java le, also in Fig. 1.
From this process a number of questions raise, such as how the mapping
process between the constraint types in the schemas and the object-oriented
programming language is done, or what happens if the schema expresses
something that the programing language is not able to represent as xed cardinalities
or repeated properties. The way ShEx-Lite solves this issue is by delegation [
For example, by default, Java and Python code generation is built in with ShEx-Lite but the JavaCodeGenerator runs some checks that the PythonCodeGenerator does not and viceversa. If any incompatibility between the schema error [ E014 ]: feature not available --> input_incorrect_schema_big_schema_2 . shexl :15:24
| 15 | schema : name | asdf : string ^ this prefix has no mapping in java and the target language is found by the corresponding code generator, ShEx-Lite will let the user know by means of error or warning messages such as the once shown in Fig. 2. 3
ShEx-Lite is available as open source software at GitHub5. It has been designed
in accordance with the concept \compiler as an API", born with the Roslyn
compiler [
{ Parse. The components of this module aim at reading the source le and performing the syntax validation. The syntax validation of the schemas checks that the schemas de ned follow the ShEx-Lite syntax. If this is not the case, the compiler lets the user know about the problem and possible solutions.
{ Sema. At this stage the compiler checks that types are correct and that the invocations and references that occur in the schemas are de ned. Also, during this process, if any error or warning is detected, the compiler will notify the user about the problem and possible solutions. { IRGen. This module is the one actually generating target code (Java, Python, Any...). In order to allow adding other languages in the future, ShEx-Lite delegates the speci c language checks and mappings and therefore it o ers an interface that other language generators will implement. Each one of the code generators is responsible of checking that the constraints represented by the schemas are able to be expressed in the corresponding language. If the schema meets the requierements of the corresponding language, it is the speci c code generator the one that produces the code. 4
In this paper we have presented ShEx-lite, a subset of ShEx that allows to generate domain model objects in di erent Object-Oriented languages like Python and Java. The system proposed is being used in the Hercules project6, and as a future work we're planning to incorporate the possiblity to read shape expressions from tabular formats like CSV7.
Aknowledgements. The HERCULES Semantic University Research Data Project is backed by the Ministry of Economy, Industry and Competitiveness with a budget of 5.462.600,00 euros with an 80% of co nancing from the 20142020 ERDF Program. This work has also been partially funded by the Spanish Ministry of Economy and Competitiveness (Society challenges: TIN2017-88877R).
7 https://github.com/dcmi/dcap