Package org.apache.tinkerpop.gremlin

Interface GraphProvider

All Known Implementing Classes:
AbstractFileGraphProvider, AbstractGraphProvider, GraphManager.ManagedGraphProvider
public interface GraphProvider
Those developing Gremlin implementations must provide a GraphProvider implementation so that the different test suites know how to instantiate their implementations. Implementers may choose to have multiple GraphProvider implementations to mix and match with multiple test suite implementations. For example, create one GraphProvider that has no indices defined and a separate GraphProvider that has indices. Then create separate test suite implementations for each GraphProvider. This approach will have the test suites executed once for each GraphProvider ensuring that the Graph implementation works under multiple configurations. Consider making these "extra" tests "integration tests" so that they don't have to be executed on every run of the build so as to save time. Run the "integration tests" periodically to ensure overall compliance.
Author:
Stephen Mallette (http://stephen.genoprime.com)

  • Field Details

    • CORE_IMPLEMENTATIONS

      static final Set<Class> CORE_IMPLEMENTATIONS
      Implementations from gremlin-core that need to be part of the clear process. This does not exempt providers from having to register their extensions to any of these classes, but does prevent them from having to register them in addition to their own.

  • Method Details

    • traversal

      default GraphTraversalSource traversal(Graph graph)
      Create a GraphTraversalSource from a Graph instance. The default implementation does not use GraphComputer so providers should override as necessary if their implementation is testing something that requires a different engine type, like GraphComputer.

    • traversal

      default GraphTraversalSource traversal(Graph graph, TraversalStrategy... strategies)
      Create a GraphTraversalSource from a Graph instance. The default implementation does not use GraphComputer so providers should override as necessary if their implementation is testing something that requires a different engine type, like GraphComputer.

      Implementations should apply strategies as necessary to the GraphTraversalSource before calling it's create method.

    • getGraphComputer

      default GraphComputer getGraphComputer(Graph graph)
      Create a GraphComputer from the Graph instance. The default implementation simply calls graph.compute().
      Parameters:
      graph - the graph to get the graph computer from
      Returns:
      a new graph computer

    • standardTestGraph

      default Graph standardTestGraph(Class<?> test, String testMethodName, LoadGraphWith.GraphData loadGraphWith)
      Creates a new Graph instance using the default Configuration from standardGraphConfiguration(Class, String, LoadGraphWith.GraphData).

    • openTestGraph

      default Graph openTestGraph(org.apache.commons.configuration2.Configuration config)
      Creates a new Graph instance from the Configuration object using GraphFactory. The assumption here is that the Configuration has been created by one of the newGraphConfiguration(String, Class, String, LoadGraphWith.GraphData) methods and has therefore already been modified by the implementation as necessary for Graph creation.

    • standardGraphConfiguration

      default org.apache.commons.configuration2.Configuration standardGraphConfiguration(Class<?> test, String testMethodName, LoadGraphWith.GraphData loadGraphWith)
      Gets the Configuration object that can construct a Graph instance from GraphFactory. Note that this method should create a Graph using the graphName of "standard", meaning it should always return a configuration instance that generates the same Graph from the GraphFactory.

    • clear

      default void clear(org.apache.commons.configuration2.Configuration configuration) throws Exception
      If possible (usually with persisted graph) clear the space on disk given the configuration that would be used to construct the graph. The default implementation simply calls clear(Graph, Configuration) with a null graph argument.

      Implementations should be able to accept an argument of null for the Configuration as well, and a proper handling is needed. Otherwise, a NullPointerException may be thrown.

      Throws:
      Exception

    • clear

      void clear(Graph graph, org.apache.commons.configuration2.Configuration configuration) throws Exception
      Clears a Graph of all data and settings. Implementations will have different ways of handling this. It is typically expected that Graph.close() will be called and open transactions will be closed. For a brute force approach, implementers can simply delete data directories provided in the configuration. Implementers may choose a more elegant approach if it exists.

      Implementations should be able to accept an argument of null for the Graph, in which case the only action that can be performed is a clear given the configuration. The method will typically be called this way as clean up task on setup to ensure that a persisted graph has a clear space to create a test graph.

      Implementations should be able to accept an argument of null for the Configuration as well, and a proper handling is needed. Otherwise, a NullPointerException may be thrown.

      Calls to this method may occur multiple times for a specific test. Develop this method to be idempotent.

      Throws:
      Exception

    • convertId

      default Object convertId(Object id, Class<? extends Element> c)
      Converts an identifier from a test to an identifier accepted by the Graph instance. Test that try to utilize an Element identifier will pass it to this method before usage. This method should be sure to be consistent in the return value such that calling it with "x" should always return the same transformed value.

    • convertLabel

      default String convertLabel(String label)
      Converts an label from a test to an label accepted by the Graph instance. Test that try to utilize a label will pass it to this method before usage.

    • newGraphConfiguration

      org.apache.commons.configuration2.Configuration newGraphConfiguration(String graphName, Class<?> test, String testMethodName, Map<String,Object> configurationOverrides, LoadGraphWith.GraphData loadGraphWith)
      When implementing this method ensure that a test suite can override any settings EXCEPT the "gremlin.graph" setting which should be defined by the implementer. It should provide a Configuration that will generate a graph unique to that graphName.
      Parameters:
      graphName - a unique test graph name
      test - the test class
      testMethodName - the name of the test
      configurationOverrides - settings to override defaults with.
      loadGraphWith - the data set to load and will be null if no data is to be loaded

    • newGraphConfiguration

      default org.apache.commons.configuration2.Configuration newGraphConfiguration(String graphName, Class<?> test, String testMethodName, LoadGraphWith.GraphData loadGraphWith)
      When implementing this method ensure that a test suite can override any settings EXCEPT the "gremlin.graph" setting which should be defined by the implementer. It should provide a Configuration that will generate a graph unique to that graphName.
      Parameters:
      graphName - a unique test graph name
      test - the test class
      testMethodName - the name of the test
      loadGraphWith - the data set to load and will be null if no data is to be loaded

    • loadGraphData

      void loadGraphData(Graph graph, LoadGraphWith loadGraphWith, Class testClass, String testName)
      Tests are annotated with a LoadGraphWith annotation. These annotations tell the test what kind of data to preload into the graph instance. It is up to the implementation to load the graph with the data specified by that annotation. This method also represents the place where indices should be configured according the the Graph implementation's API. Implementers can use the testClass and testName arguments to implement test specific configurations to their graphs.
      Parameters:
      graph - the Graph instance to load data into constructed by this GraphProvider
      loadGraphWith - the annotation for the currently running test - this value may be null if no graph data is to be loaded in front of the test.
      testClass - the test class being executed
      testName - the name of the test method being executed

    • getImplementations

      Set<Class> getImplementations()
      Get the set of concrete implementations of certain classes and interfaces utilized by the test suite. This method should return any implementations or extensions of the following interfaces or classes:

      The test suite only enforces registration of the following core structure interfaces (i.e. these classes must be registered or the tests will fail to execute):

      The remaining interfaces and classes should be registered however as failure to do so, might cause failures in the Groovy environment testing suite.

      Internally speaking, tests that make use of this method should bind in CORE_IMPLEMENTATIONS to the Set because these represent gremlin-core implementations that are likely not registered by the vendor implementations.

    • getWorkingDirectory

      default String getWorkingDirectory()
      Helper method for those build GraphProvider implementations that need a standard working directory for tests (e.g. graphs that persist data to disk). Typically, there is no need to override the default behavior of this method and if it is overridden, it is usually best to continue to use the TestHelper to produce the working directory as it will create the path in the appropriate build directories.
      Returns:
      UNIX-formatted absolute directory path

    • getTestListener

      default Optional<GraphProvider.TestListener> getTestListener()
      Returns a GraphProvider.TestListener implementation that provides feedback to the GraphProvider implementation. By default, this returns an empty listener.

    • getStaticFeatures

      default Optional<Graph.Features> getStaticFeatures()
      Gets a Graph.Features implementation that contains graph feature configuration that will never change from execution to execution of the tests given this current GraphProvider implementation. Implementing this method will allow the test suite to avoid creation of a Graph instance and thus speed up the execution of tests if that creation process is expensive. It is important that this static set of features be representative of what the Graph instance produced by this GraphProvider can actually do or else the cost of Graph instantiation will be incurred when it doesn't need to be. It is also important that this method be faster than the cost of Graph creation in the first place or there really won't be any difference in execution speed. In cases where some features are static and others are not, simply throw an UnsupportedOperationException from those features to let the test suite know that it cannot rely on them and the test suite will revert to using a constructed Graph instance.