Xponents

Geographic Place, Date/time, and Pattern entity extraction toolkit along with text extraction from unstructured data and GIS outputters.

View the Project on GitHub

Building Xponents

After checkout, first build Core API (opensextant-xponents-core) Then you can begin working with the extractor SDK (opensextant-xponents)

  1. Setup project.
  # Set up Core API with independent reference data
  ./setup.sh

  # Set the main SDK API with the Gazetteer data unique to this project:
  cd solr && ./build.sh meta
  1. Setup Core API to build and test:
  (cd ./Core && mvn install)
  1. Build a full Gazetteer from scratch.

./solr/README.md captures all the mechanics of building the Gazettter from checkout. The SDK is dependent on having one of the folders ./solr source or ./xponents-solr built. Since you are reading thie BUILD documentation you’ll want to try building the ./solr folder from source.

  1. Build SDK:

There are some outstanding build tasks related to gazetteer metadata assembly that need automation. But in general, you would make use of the ./solr/build.sh script and consult the README there.

  mvn install  


  # To try out demos, tests, etc.  This step invokes `mvn install` on the desired projects.
  ant build build-examples

Examples sub-project makes use of Core, the main SDK, and the OpenSextant XText project. XText libraries are published in Maven Central.

  1. Distribution and Packaging: ant dist

Producing API documentation updates: infrequently, only for major releases.

 # target ./doc folders already exist as folders known to git.
 cp -r target/apidocs/ doc/sdk-apidocs
 cp -r Core/target/apidocs/ doc/core-apidocs

 # When done, commit changes, as API docs appear on GitHub public site.
  1. Test. Leverage the Examples projects to more fully test your build.

  2. Final Release. With the properly compiled, tested packaging then create a distribution in full.

 ./script/dist.sh 

 # Note - this script call ant  package-dist which just packages the pre-built, tested libraries.
  1. Publish Java Libraries to Maven Central as below

  2. Build, Test and Publish Docker Images as described in ./Examples/Docker/

All Together

Put it all together it might look like this, with notes about size of interim data:


  ./setup.sh

  # About here you head into ./solr and produce the Gazetteer accoring to those build notes.

  # Java modules
  # ----------------------
  # Ant script below automates these Maven routines. So to build there are options
  # Depending on what you are doing.
  # Option -- Maven to install
  # (cd ./Core && mvn install)
  # mvn install 
  # (cd ./Examples && mvn install)

  ant compile 
  # Once tested and good -- publish a release to Maven Central 

  # Use Ant to wrap around Maven to site, package and prep distro
  ant dist

  # Test a bit, right?

  # Complete distribution
  # ----------------------
  # Produce the  full Xponents SDK, then build docker images from that.
  # Raw distro is 4 GB;  Docker image is 4.8 GB,  Offline Docker image is 5.0 GB
  # SubTotal - 15 GB
  ./script/dist.sh
  ./script/dist-docker.sh
  
  # Prior to doing anything with Sonar make sure it is accessible. 
  # If using Examples/Docker/Sonarqube -- go there now and run `docker-compose up -d`.  Configured port is 9900
  ./script/dist-docker-offline.sh  http://localhost:9900  $SONAR_TOKEN

Code Practices

Three (3) or more different software quality practices are in play here:

This project overall converted from using checkstyle and findbugs to just using Sonarqube. See the deployment of Sonarqube in Examples/Docker/Sonarqube

For historical purposes here is a POM snippet for using findbugs, which is no longer maintained. In a similar reduction of dependencies, checkstyle plugin functions are replaced by Sonar scanning:


<project>
  ...    
  <pluginManagement>
        <plugin>
          <groupId>org.codehaus.mojo</groupId>
          <artifactId>findbugs-maven-plugin</artifactId>
          <version>3.0.5</version>
        </plugin>
  </pluginManagement>
  <plugins>
      <!-- run explicitly with: mvn findbugs:check -->
      <plugin>
        <groupId>org.codehaus.mojo</groupId>
        <artifactId>findbugs-maven-plugin</artifactId>
        <configuration>
          <xmlOutput>true</xmlOutput>
          <!--<threshold>High</threshold> -->
        </configuration>
      </plugin>
      
      <!-- run explicitly with: mvn checkstyle:check -->
      <plugin>
        <groupId>org.apache.maven.plugins</groupId>

        <artifactId>maven-checkstyle-plugin</artifactId>
        <configuration>
          <configLocation>checkstyle.xml</configLocation>
          <propertyExpansion>checkstyle.indentChars=4</propertyExpansion>
          <suppressionsLocation>checkstyle-suppressions.xml</suppressionsLocation>
          <consoleOutput>true</consoleOutput>
          <failOnViolation>false</failOnViolation>
        </configuration>
      </plugin>
  </plugins>
</project>

For Docker Scan and Snyk usage, you will need to review those emerging services. They seem to work in an insecure mode, but have issues with secure operation behind firewalls. With a vulnerability report in hand you can adjust Maven POM or other aspects before publishing.

  # Get a vulnerability report
  snyk auth
  snyk test  --insecure


  # Ditto, using Docker
  docker scan opensextant:xponents-offline-3.5

Sizing

Overall a complete build of gazetteer and production docker images will consume 45 GB. Docker images are compressable to about 2.0 GB. The breakdown is below:

Level of Effort.

First time to acquire all the software and build could be a day: You would pull in 4 GB of compressed files, unpack them all, configure software, etc.
The Second time to walk through this all usually takes 2-4 hours, on a Macbook or Linux server with 4 CPU, for example.

If this sounds like a bit much, then leverage the artifacts in Maven Central or Docker Hub.

Maven Publishing

  cd ./Core
  mvn clean deploy -P release
  cd ..
  mvn clean deploy -P release