Skip to content

Latest commit

 

History

History
61 lines (42 loc) · 2.89 KB

implementation.md

File metadata and controls

61 lines (42 loc) · 2.89 KB
Raw

Implementation Notes

This file contains implementation notes. It is intended for developers of TermIt or anyone interested in its internal structure and logic.

jsoup

Had to switch from standard Java DOM implementation to jsoup because DOM had sometimes trouble parsing HTML documents (meta tags in header). Jsoup, on the other hand, handles HTML natively and should be able to work with XML ( if need be) as well.

Validation

The application uses JSR 380 validation API. This provides a generic, easy-to-use API for bean validation based on annotations. Use it to verify input data. See User and its validation in BaseRepositoryService /UserRepositoryService. ValidationException is then handled by RestExceptionHandler and an appropriate response is returned to the client.

Storage

TermIt is preconfigured to run against a local GraphDB repository at http://locahost:7200/repositories/termit. This can be changed by updating application.yml.

User vs UserAccount

User is a domain class used for domain functions, mostly for resource provenance (author, last editor). It does not support password. UserAccount is used for security-related functions and supports password. Most parts of the application should use User.

JMX

A JMX bean called AppAdminBean was added to the application. Currently, it supports invalidation of application caches. The bean's name is set during Maven build. In case multiple deployments of TermIt are running on the same application server, it is necessary to provide different names for it. A Maven property with default value DEV was introduced for it. To specify a different value, pass a command line parameter to Maven, so the build call might look as follows:

mvn clean package -B -P production "-Ddeployment=DEV"

Fulltext Search

Fulltext search currently supports multiple types of implementation:

  • Simple substring matching on term and vocabulary label (default)
  • GraphDB with Lucene connector

Each implementation has its own search query which is loaded and used by SearchDao. In order for the more advanced implementation for Lucene to work, a corresponding Maven profile (graphdb) has to be selected. This inserts the correct query into the resulting artifact during build. If none of the profiles is selected, the default search is used.

Note that in case of GraphDB, corresponding Lucene connectors (label_index for labels and defcom_index for definitions and comments) have to be created as well.

RDFS Inference in Tests

The test in-memory repository is configured to be a RDF4J SAIL with RDFS inferencing engine. The repository is by default left empty (without the model) to facilitate test performance (inference in RDF4J is really slow). To load the TermIt model into the repository and thus enable RDFS inference, call the enableRdfsInference method available on both BaseDaoTestRunner and BaseServiceTestRunner.