Contributing a new query operation actor to the Comunica repository


On this page

    This guide focuses on all the required steps for contributing a new query operation actor to Comunica. Concretely, we will focus on implementing a custom actor for the SPARQL REDUCED operator.

    Once you have followed this guide and actually want to contribute, have a look at our contribution guide.

    1. Requirements

    You will need the following to follow this guide:

    • git
    • Node.js (version 8.0 or higher)
    • Yarn package manager
    • Any kind of editor that be used to edit JavaScript files (We recommend WebStorm)
    • A GitHub account

    2. Cloning the repo

    Since you do not have access to the Comunica repository by default, you will have to fork the Comunica repo first.

    1. Go to https://github.com/comunica/comunica
    2. Make sure you are logged into your GitHub account.
    3. Click on the "Fork" button.

    After this, a copy of the Comunica repo will be available for your account in which you can make all the changes you want.

    Next, we will clone your fork to the local file system as follows:

    $ git clone https://github.com/<my-username>/comunica.git
    

    Make sure you replace <my-username> with your GitHub username.

    As a final setup step, we can install all dependencies as follows:

    $ cd comunica
    $ yarn install
    

    This will install the dependencies of all modules. After that, all Comunica packages are available in the packages/ folder and can be used in a development environment. All pre-built Comunica engines and configs are available in the engines/ folder such as querying with Comunica SPARQL (engines/query-sparql).

    A good git practise is to develop on feature branches. For this, branch from the master as follows:

    $ git checkout -b feature/my-feature
    

    Replace my-feature with a short name (without spaces) of the feature you want to implement.

    If you fix a bug, you can name your branch something like fix/my-fix.

    If you want to make sure that everything has been installed correctly, navigate to engines/query-sparql, and try out a simple query from the command line:

    $ cd engines/query-sparql
    $ node bin/query.js https://fragments.dbpedia.org/2016-04/en \
      'SELECT * WHERE { ?s ?p ?o } LIMIT 100'
    

    If this command produces valid output, your development environment has been successfully setup.

    Let's navigate back to the repo root, so we're ready for the next step:

    $ cd ../..
    

    3. Creating a new package

    The Comunica monorepo contains a large collection of packages in the packages/ directory. This contains different types of packages: actors, mediators and buses.

    Learn more about the core architecture and SPARQL actors and buses.

    For each type of package, we provide a generator tool to initialize a template repo. For this, you can use the generator-comunica project (a Yo generator).

    To install this generator, start a new terminal session outside of the Comunica repo directory, and execute the following commands:

    $ npm i -g yo
    $ git clone git@github.com:comunica/generate-comunica.git
    $ cd generate-comunica
    $ npm install
    $ npm link
    

    This will expose the comunica:bus, comunica:mediator, comunica:actor, and comunica:actor-query-operation generators for initializing projects of the respective types. comunica:actor-query-operation is a special type of the comunica:actor generator that has been preconfigured to the query-operation bus, which we will make use of in this guide. If you want to create an actor on another bus than query-operation, you will have to invoke comunica:actor instead.

    In this case, we want to create an actor on the query-operation bus for the REDUCED query operation. As such, we can execute the generator as follows in the repo root:

    $ yo comunica:actor-query-operation
    ? The SPARQL Algebra type name of the operator (lowercase) reduced
    ? The SPARQL Algebra interface name Reduced
    ? Actor name (without actor-bus- prefix, lowercase) reduced-my
    ? The full readable name of the actor Reduced My
    ? The component base name of the actor (without Bus part) ReducedMy
    ? A description of the actor A comunica Reduced My Query Operation Actor.
    ? The component context prefix caqorm
       create packages/actor-query-operation-reduced-my/components/Actor/QueryOperation/ReducedMy.jsonld
       create packages/actor-query-operation-reduced-my/components/components.jsonld
       create packages/actor-query-operation-reduced-my/components/context.jsonld
       create packages/actor-query-operation-reduced-my/lib/ActorQueryOperationReducedMy.ts
       create packages/actor-query-operation-reduced-my/test/ActorQueryOperationReducedMy-test.ts
       create packages/actor-query-operation-reduced-my/.npmignore
       create packages/actor-query-operation-reduced-my/index.ts
       create packages/actor-query-operation-reduced-my/package.json
       create packages/actor-query-operation-reduced-my/README.md
    

    After answering the required question, a new package will be initialized at packages/actor-query-operation-reduced-my/.

    In order to link the dependencies of this new package, make sure to run yarn install again in the monorepo root. You will see some compilation errors, which you can ignore, as your new actor has not been implemented yet.

    4. Implementing your actor

    In this step, we will implement our actor in packages/actor-query-operation-reduced-my/lib/ActorQueryOperationReducedMy.ts.

    The generated class extends from ActorQueryOperationTypedMediated, which abstracts away many of the commonly required tasks for operators. This class requires you to override two methods: testOperation and runOperation. These two methods correspond to the test and run phases that will be called by mediators.

    4.1. Test phase

    Since the ActorQueryOperationTypedMediated class already implements the test phase by checking if the incoming operation is a REDUCED operation, we can just implement testOperation as follows:

      public async testOperation(pattern: Algebra.Reduced, context: IActionContext): Promise<TestResult<IActorTest>> {
        return passTestVoid();
      }
    
    If you want to make your actor only handle specific types of this operation, you can add additional checks in here. If you want to fail the test in certain cases, you will have to return failTest(`a message describing the failure`).

    4.2. Run phase

    The runOperation method will contain the actual logic for evaluation the REDUCED operator.

    Before we start, change the return type of this method from Promise<IQueryOperationResult> to Promise<IQueryOperationResultBindings>, because this method will always return bindings as query result.

    The first step of implementing the REDUCED actor, requires evaluating the sub-operation that this REDUCED operation exists over.

    For example, REDUCED can be applied over the following BGP:

    SELECT REDUCED * WHERE {
      ?s ?p <http://dbpedia.org/resource/Belgium>.
      ?s ?p ?o.
    }
    

    As such, we first have to evaluate this BGP first (or whatever other sub-operator is defined).

    This sub-operation is stored in the input field of our pattern. By using the query operation mediator (this.mediatorQueryOperation), we can evaluate this sub-operation. The sub-operator can be evaluated by the mediator as follows:

    // Delegate resolving the input operation to the mediator.
    const output = ActorQueryOperation.getSafeBindings(await this
          .mediatorQueryOperation.mediate({ operation: pattern.input, context }));
    

    Since the REDUCED operator is very loosely defined in the SPARQL specification, it is valid to filter nothing from the results, and just return the child operator's results as-is.

    As such, we can return the following:

    return {
      type: 'bindings',
      bindingsStream: output.bindingsStream,
      metadata: output.metadata,
    };
    
    Have a look at the other query operation actors if you want to do something more complex with the output's bindingsStream.

    5. Unit-testing your actor

    Since testing is very important in Comunica, the generator will automatically generate some unit tests for your class in packages/actor-query-operation-reduced-my/test/ActorQueryOperationReducedMy-test.ts.

    Since we don't actually do anything in our actor, all default unit test should already pass. Check this by executing in the repo root:

    yarn run test ActorQueryOperationReducedMy-test.ts
    

    Here, it is important that every class in your package reaches a code coverage of 100%. Therefore, if you have a different actor implementation, you may have to add additional unit tests to check different cases.

    6. Configuring your actor

    If you want to make it so that your actor is enabled by default in Comunica SPARQL, then you'll have to make sure it is present in the default config.

    For this, first add your package as a dependency in engines/query-sparql/package.json:

    {
      ...
      "dependencies": {
        ...
        "@comunica/actor-query-operation-reduced-my": "^1.0.0"
      }
      ...
    }
    
    When creating a new actor, you can leave the version fixed at "^1.0.0". This version will be incremented automatically upon each new Comunica release.

    Next, we have to configure the actor by replacing the existing REDUCED actor in the default config file engines/config-query-sparql/config/query-operation/actors/query/reduced.json:

    {
      "@context": [
        "https://linkedsoftwaredependencies.org/bundles/npm/@comunica/runner/^3.0.0/components/context.jsonld",
    
        "https://linkedsoftwaredependencies.org/bundles/npm/@comunica/actor-query-operation-reduced-my/^3.0.0/components/context.jsonld"
      ],
      "@id": "urn:comunica:default:Runner",
      "@type": "Runner",
      "actors": [
        {
          "@id": "urn:comunica:default:query-operation/actors#reduced",
          "@type": "ActorQueryOperationReducedMy",
          "mediatorQueryOperation": { "@id": "urn:comunica:default:query-operation/mediators#main" }
        }
      ]
    }
    
    When adding non-query-operator actors, you may have to include your actor in a different config set.

    7. Testing with Comunica SPARQL

    Before we make our pull request, we have to make sure that our actor actually works in practise.

    For this, we have to make sure our TypeScript is properly compiled to JavaScript, and that our configuration file has been compiled:

    $ yarn run build # Compile typescript and the components files at the ROOT OF THE REPO
    $ cd engines/query-sparql
    $ yarn run prepare # Compiles config
    
    You can also just run yarn install again from the root package, which will take care of all of this, and more.

    After that, we should now be able to execute Comunica SPARQL from the command line with a given REDUCED query:

    $ node bin/query.js https://fragments.dbpedia.org/2016-04/en \
      'SELECT REDUCED * WHERE { ?s ?p ?o } LIMIT 100'
    

    8. Creating a pull request

    Once everything has been tested, we can commit our code and create a pull request.

    First, add the changed files, and commit your code.

    $ git add packages/actor-query-operation-reduced-my \
      engines/query-sparql/config \
      engines/query-sparql/package.json
    $ git commit -m "Add my custom reduced operator" 
    
    Before making the commit, make sure you are not any unneeded files. You can use git status for this.

    Several pre-commit checks will be done, such as linting and unit testing. Should any of these checks fail, your commit will not be done, and you have to retry again after fixing the problems.

    Also make sure to check in your new package if there are any TODOs remaining, such as in the README.md file.

    Once your commit is done, you can push your changes to your fork:

    $ git push origin feature/my-feature
    

    The only thing that's left to do is making the pull request from your branch to the Comunica master branch at https://github.com/comunica/comunica/pulls. Once you've opened the pull request, several automated checks will be run, and someone will have a look at your contribution very soon!