This document focuses on how queries are run on a graph managed by gaedo-blueprints. If one wants to have infos on specific features regarding graph queries, this is the right document. However, if one wants infos regarding how to write a query, regardless of the underlyin storage mechanism, or how dynamic finders are mapped on real QueryBuilder, general documentation of gaedo is better suited.

How is a query performed on graph - a broad view

To make things clear, explanations here are given in the best case, thats to say when StrategyType.beanBased is used. If the previous sentence has no meaning to you, it probably mean youre in that case, as the oppsite strategy (StrategyType.graphBased) is the non-default and should be selected by developper.

To make things simples, execution of a query on graph is quite simple and can be splitted in some easy steps

  1. Transform QueryExpression into a collection of VertexTest
  2. Choose some query roots from that collection of vertex and back-navigate to vertex corresponding to objects
  3. Perform all VertexTest on those vertices and, if they match all, add them to results iterable
  4. Browse results iterable and load objects on the fly

Now, lets detail these elements on some sample queries. For all those example queries, we will suppose (like in GraphPostFinderServiceTest), we have a 

// a standard IndexableGraphBackedFinderService with StrategyType.beanBased
FinderCrudService<Post, PostInformer> service;

which will be used to perform those sample queries

Find posts having a value

Find posts having a given literal value

First, lets start by a very simple example

service.find().matching(new QueryBuilder() {
    public QueryExpression createMatchingExpression(PostInformer informer) {
       return informer.getNote().equalsTo(2.0f);
    }
}).getAll()

The goal of this query is quite simple (and it could also be expressed using this dynamic finder : findAllByNoteEqualsTo(2.0f) which has the known advantage of readability), so lets detail this behaviour.

First, QueryExpression is visited by a BluePrintsQueryBuilder which will construct a GraphExecutableQuery using the predicates given by the QueryExpression. As the GraphExecutableQuery has a nice toString(), lets use it to see how this query translates at the blueprints level : 

com.dooapp.gaedo.blueprints.queries.executable.OptimizedGraphExecutableQuery [
    test= AndVertexTest 
       note EqualsTo 2.0
       classes CollectionContains class com.dooapp.gaedo.test.beans.Post, 
sort=com.dooapp.gaedo.finders.sort.SortingExpressionImpl@47570945]

Well, the sorting expression is useless here (as you may now, this version not yet implements sorting in graph). Anyway, the test is quite clear. let me repeat it : 

test= AndVertexTest 
    note EqualsTo 2.0
    classes CollectionContains class com.dooapp.gaedo.test.beans.Post,

This is quite clear : we look for objects having an edge named note having as value 2.0 and which classes collection contains com.dooapp.gaedo.test.beans.Post (a graphw ay to say target node can be casted to an instance of Post, no ?).

The graph can be big. Really big, I mean, so we clearly wont browse all vertices looking for the ones were searching. Hopefully, to avoid that (as the picture upper shows), properties are represented as edges. So were quite sure the results are linked to both the 2.0f literal and to the com.dooapp.gaedo.test.beans.Post literal. The ensemblist method would be here to intersect both sets. Unfortunatly, gaedo-blueprints does not yet provide such a feature. To emulate it, in a way, we check which of the set is the smallest, then test each of its elements for matching.

This way we obtain a list of vertices that will be loaded on iteration.

Find posts having a given object value

Now how to get posts written by a given author ? Suppose I have a User object (named user) obtained from a providential userService. To get all posts written by this author, its as simple as 

service.find().matching(new QueryBuilder() {
    public QueryExpression createMatchingExpression(PostInformer informer) {
       return informer.getAuthor().equalsTo(author);
    }
}).getAll()

Well, I give the dynamic query to you

Anyway, it translates as a GraphExecutableQuery which test is

AndVertexTest 
    author EqualsTo com.dooapp.gaedo.test.beans.User@b4deeb1c
    classes CollectionContains class com.dooapp.gaedo.test.beans.Post

How is the author translated into a vertex you guess ? Well, the userService does the job and gives the query a root vertex to count edges on.