Set up and query composite databases

The examples featured in this section make use of the two Cypher clauses: USE and CALL {}.

Graph set-up

The following set-up is required to recreate the examples on this page:

Create a standard database movies2022

CREATE DATABASE movies2022

Create a composite database cineasts

CREATE COMPOSITE DATABASE cineasts

Create database alias cineasts.latest for a local database in a composite database

CREATE ALIAS `cineasts`.`latest`
  FOR DATABASE movies2022

Create database alias cineasts.upcoming for a remote database in a composite database

CREATE ALIAS `cineasts`.`upcoming`
  FOR DATABASE upcoming
  AT 'neo4j+s://location:7687'
  USER neo4j
  PASSWORD 'password'

For more information about composite databases and database aliases in composite databases, see Managing composite databases, and Managing database aliases in composite databases.

Graph selection

Queries submitted to a composite database may contain several USE clauses that direct different parts of the query to different constituent graphs.

Each constituent graph is named after the alias that introduces it into the Composite database.

Query a single graph

Example 1. Reading and returning data from a single graph

USE cineasts.latest
MATCH (movie:Movie)
RETURN movie.title AS title

The USE clause at the beginning of the query selects the cineasts.latest graph for all the subsequent clauses. MATCH is performed on that graph.

Query multiple graphs

Example 2. Reading and returning data from two graphs

USE cineasts.latest
MATCH (movie:Movie)
RETURN movie.title AS title
  UNION
USE cineasts.upcoming
MATCH (movie:Movie)
RETURN movie.title AS title

The first part of the UNION query selects the cineasts.latest graph and the second part selects the cineasts.upcoming graph.

Dynamic graph access

Queries can also select constituent graphs dynamically, using the form USE graph.byName(graphName).

Example 3. Reading and returning data from dynamically selected graphs

UNWIND ['cineasts.latest', 'cineasts.upcoming'] AS graphName
CALL {
  USE graph.byName(graphName)
  MATCH (movie:Movie)
  RETURN movie
}
RETURN movie.title AS title

In the example above, the part of the query accessing graph data, MATCH (movie:Movie), is wrapped in a sub-query with a dynamic USE clause. UNWIND is used to get the names of our graphs, each on one row. The CALL {} sub-query executes once per input row. In this case, once selecting cineasts.latest, and once selecting cineasts.upcoming.

Listing graphs

The built-in function graph.names() returns a list containing the names of all constituent graphs on the current Composite database.

Example 4. The graph.names() function

UNWIND graph.names() AS graphName
RETURN graphName

+---------------------+
| graphName           |
+---------------------+
| "cineasts.latest"   |
| "cineasts.upcoming" |
+---------------------+

The names returned by this function can be used for dynamic graph access.

Example 5. Reading and returning data from all graphs

UNWIND graph.names() AS graphName
CALL {
  USE graph.byName(graphName)
  MATCH (movie:Movie)
  RETURN movie
}
RETURN movie.title

Query result aggregation

Example 6. Getting the earliest release year of all movies from all graphs

UNWIND graph.names() AS graphName
CALL {
  USE graph.byName(graphName)
  MATCH (movie:Movie)
  RETURN movie.released AS released
}
RETURN min(released) AS earliest

The sub-query returns the released property of each movie, from each constituent graph. The RETURN at the end of the main query aggregates across the full result to calculate the global minimum.

Correlated subqueries

This query finds all movies in cineasts.upcoming that are to be released in the same month as the longest movie in cineasts.latest.

Example 7. Correlated subquery

CALL {
  USE cineasts.latest
  MATCH (movie:Movie)
  RETURN movie.releasedMonth AS monthOfLongest
    ORDER BY movie.runningTime DESC
    LIMIT 1
}
CALL {
  USE cineasts.upcoming
  WITH monthOfLongest
  MATCH (movie:Movie)
  WHERE movie.releasedMonth = monthOfLongest
  RETURN movie
}
RETURN movie

The first part of the query finds the movie with the longest running time from cineasts.latest, and returns its release month. The second part of the query finds all movies in cineasts.upcoming that fulfill our condition and returns them. The sub-query imports the monthOfLongest variable using WITH monthOfLongest, to make it accessible.

Updates

Composite database queries can perform updates to constituent graphs.

Example 8. Constituent graph update

USE cineasts.upcoming
CREATE (:Movie {title: 'Dune: Part Two'})

Updates can only be performed on a single constituent graph per transaction.

Example 9. Multi-graph update will fail

UNWIND graph.names() AS graphName
CALL {
  USE graph.byName(graphName)
  CREATE (:Movie {title: 'The Flash'})
}

Writing to more than one database per transaction is not allowed.

Limitations

Queries on Composite databases have a few limitations.

Graph accessing operations

Consider a Composite database query:

UNWIND graph.names() AS graphName
CALL {
  USE graph.byName(graphName)
  MATCH (movie:Movie)
  RETURN movie
}
RETURN movie

Here the outer clauses, i.e. the UNWIND, the CALL itself, and the final RETURN, appear in the root scope of the query, without a specifically chosen graph. Clauses or expressions in scopes where no graph has been specified must not be graph-accessing.

The following Composite database query is invalid because [p=(movie)-→() | p] AS paths is a graph-accessing operation in the root scope of the query:

UNWIND graph.names() AS graphName
CALL {
  USE graph.byName(graphName)
  MATCH (movie:Movie)
  RETURN movie
}
RETURN [p=(movie)-->() | p] AS paths

See examples of graph-accessing operations:

RETURN 1 + 2 AS number
WITH node.property AS val

Nested `USE` clauses

An inner scope must use the same graph as its outer scope:

USE cineasts.latest
MATCH (n)
CALL {
  USE cineasts.upcoming
  MATCH (m)
  RETURN m
}
RETURN n, m

Nested subqueries must use the same graph as their parent query.
Attempted to access graph cineasts.upcoming
"    USE cineasts.upcoming"
     ^

Sub-queries without a USE clause can be nested. They inherit the specified graph from the outer scope.

CALL {
  USE cineasts.upcoming
  CALL {
    MATCH (m:Movie)
    RETURN m
  }
  RETURN m
}
RETURN m

Cypher runtime

When a query is submitted to a Composite database, different parts of the query may run using different runtimes. Clauses or expressions in scopes where no graph has been specified run using the slotted runtime. Parts of the query directed to different constituent graphs are run using the default runtime for that graph, or respect the submitted Cypher query options if specified.

Built-in graph functions

Graph functions are located in the namespace graph. The following table describes these functions:

Table 1. Built-in graph functions
Function	Explanation
`graph.names()`	Provides a list of names of all constituent graphs on the current Composite database.
`graph.byName(graphName)`	Used with the `USE` clause to select a constituent graph by name dynamically. This function is supported only with `USE` clauses.
`graph.propertiesByName(graphName)`	Returns a map containing the properties associated with the given graph.

For more information, see Graph functions in the Cypher Manual.