Graph Database Connection

Table of Contents

Description
Where you can use them
Options
Converting a Neo4j connection
Compatibility
Capabilities

Description

A graph database connection connects to a graph database which speaks Cypher or Gremlin. The graph database type determines how Hop connects and which statements it generates:

Type Plugin Connection

Type	Plugin	Connection
Neo4j	Neo4j	Bolt. Automatic configuration with the `neo4j://` scheme on port 7687, the defaults of a Neo4j Connection.
Memgraph	Neo4j	Bolt. A direct `bolt://` connection on port 7687 without encryption. Memgraph only offers routing in its high availability setup.
Amazon Neptune	Neo4j	Bolt. A `bolt://` connection on port 8182 with encryption, which Neptune requires.
FalkorDB	FalkorDB	The Redis protocol, on port 6379. One connection works with one graph, which FalkorDB creates when it doesn’t exist.
Apache AGE	Apache AGE	PostgreSQL with the AGE extension, on port 5432, through the PostgreSQL JDBC driver which comes with Hop. One connection works with one AGE graph, which Hop can create when it doesn’t exist.
Gremlin	Gremlin	The Apache TinkerPop driver, on port 8182: Apache TinkerPop Gremlin Server, JanusGraph or the Gremlin endpoint of Amazon Neptune.

Neo4j

Bolt. Automatic configuration with the neo4j:// scheme on port 7687, the defaults of a Neo4j Connection.

Memgraph

Neo4j

Bolt. A direct bolt:// connection on port 7687 without encryption. Memgraph only offers routing in its high availability setup.

Amazon Neptune

Neo4j

Bolt. A bolt:// connection on port 8182 with encryption, which Neptune requires.

FalkorDB

The Redis protocol, on port 6379. One connection works with one graph, which FalkorDB creates when it doesn’t exist.

Apache AGE

PostgreSQL with the AGE extension, on port 5432, through the PostgreSQL JDBC driver which comes with Hop. One connection works with one AGE graph, which Hop can create when it doesn’t exist.

Gremlin

The Apache TinkerPop driver, on port 8182: Apache TinkerPop Gremlin Server, JanusGraph or the Gremlin endpoint of Amazon Neptune.

Graph database connections are serialized as JSON in the Hop metadata folder under

metadata/graph-database-connection/

When a Neo4j connection and a graph database connection have the same name, the Neo4j connection is used, so existing projects keep working exactly as before.

Where you can use them

Feature Neo4j, Memgraph, Neptune FalkorDB, Apache AGE Gremlin

Feature	Neo4j, Memgraph, Neptune	FalkorDB, Apache AGE	Gremlin
Graph output transform	Yes	Yes	Yes: nodes and relationships are upserted with Gremlin traversals. A node is identified by its primary properties and gets one label.
Validating against the graph model in Graph output	Yes	Yes	The use of the model only: Gremlin servers don’t list their indexes.
Cypher output transform	Yes	Yes	No: it writes Cypher. Use Graph output with a graph model.
Graph query transform, Graph script action	Yes	Yes	Yes, with Gremlin scripts instead of Cypher. Parameters become script bindings.
Returning a graph from the Graph query transform	Yes	Yes	Yes
Cypher builder transform	Yes	Yes	No: it builds Cypher
Graph index, Graph constraint actions	Neo4j and Memgraph	Indexes on FalkorDB. On Apache AGE, create PostgreSQL indexes on the label tables.	No: create indexes with the schema tools of the server
Check graph database connections action	Yes	Yes	Yes
Get graph schema transform	Yes: Neo4j’s schema procedures, Memgraph’s SHOW SCHEMA INFO, a sample on Neptune	Yes: the labels from the database, the properties from a sample	Yes, from a sample of the vertices and edges
Graph vector search transform	Neo4j and Memgraph	FalkorDB	No
Execution logging (`HOP_GRAPH_LOGGING_CONNECTION`), Get graph logging info transform, Graph logging perspective, graph database execution information location	Yes	Yes	No: they are built on Cypher

Graph output transform

Yes

Yes: nodes and relationships are upserted with Gremlin traversals. A node is identified by its primary properties and gets one label.

Validating against the graph model in Graph output

Yes

The use of the model only: Gremlin servers don’t list their indexes.

Cypher output transform

Yes

No: it writes Cypher. Use Graph output with a graph model.

Graph query transform, Graph script action

Yes

Yes, with Gremlin scripts instead of Cypher. Parameters become script bindings.

Returning a graph from the Graph query transform

Yes

Cypher builder transform

Yes

No: it builds Cypher

Graph index, Graph constraint actions

Neo4j and Memgraph

Indexes on FalkorDB. On Apache AGE, create PostgreSQL indexes on the label tables.

No: create indexes with the schema tools of the server

Check graph database connections action

Yes

Get graph schema transform

Yes: Neo4j’s schema procedures, Memgraph’s SHOW SCHEMA INFO, a sample on Neptune

Yes: the labels from the database, the properties from a sample

Yes, from a sample of the vertices and edges

Graph vector search transform

Neo4j and Memgraph

FalkorDB

Execution logging (HOP_GRAPH_LOGGING_CONNECTION), Get graph logging info transform, Graph logging perspective, graph database execution information location

Yes

No: they are built on Cypher

Amazon Neptune is only tested through its Bolt protocol for reading and writing data. Neptune has no username and password authentication: the username and password of the connection aren’t used. Requests aren’t signed with AWS Signature Version 4, so IAM database authentication isn’t supported: connect to a Neptune cluster with IAM authentication disabled, over TLS.

Options

Neo4j, Memgraph and Amazon Neptune

The options are the same as the options of a Neo4j Connection. They are grouped in four tabs:

Tab Options

Tab	Options
Connection	Server or IP address, database name, Bolt port, browser port, username and password.
Protocol	Automatic configuration, protocol (URL scheme), routing, routing policy, encryption and trusting all certificates. Every on/off option has a variable alternative which, when set, overrides it, for example `'${NEO4J_ENCRYPTION}'`.
Advanced	Connection liveness check timeout, maximum connection lifetime, maximum connection pool size, connection acquisition timeout, connection timeout and maximum transaction retry time, all in milliseconds.
Manual URLs	URLs which replace the ones built from the other options, for example `bolt://localhost:7687`.

Connection

Server or IP address, database name, Bolt port, browser port, username and password.

Protocol

Automatic configuration, protocol (URL scheme), routing, routing policy, encryption and trusting all certificates. Every on/off option has a variable alternative which, when set, overrides it, for example '${NEO4J_ENCRYPTION}'.

Advanced

Connection liveness check timeout, maximum connection lifetime, maximum connection pool size, connection acquisition timeout, connection timeout and maximum transaction retry time, all in milliseconds.

Manual URLs

URLs which replace the ones built from the other options, for example bolt://localhost:7687.

FalkorDB

Server or IP address, port, graph name, username, password and whether to use TLS. Leave the username empty for the default user and the password empty when the server doesn’t ask for one.

Gremlin

Servers or IP addresses (several servers of a cluster separated by commas), port, traversal source (g by default), username, password, whether to use TLS, and the serializer: GraphBinary for current servers, GraphSON 3 or GraphSON 2 for older ones. The JanusGraph types are registered with every serializer, so JanusGraph vertices and edges come back like those of any other server.

Apache AGE

Server or IP address, port, database name, username, password, graph name and whether to create the graph when it doesn’t exist. AGE graph names have at least 3 characters.

Use the Test button to verify the connection.

Converting a Neo4j connection

Existing Neo4j connections keep working as they are: there is no need to convert them. To convert one, open it in the metadata perspective and click Convert…. Convert all… converts all the Neo4j connections of the project at once. A Neo4j connection is skipped when a graph database connection with the same name exists already. This saves graph database connections of type Neo4j with the same names and settings and removes the Neo4j connections. Pipelines, workflows, the graph database execution information location and the execution logging find connections by name, so they keep working without changes. Older versions of Hop don’t know graph database connections, so only convert when every Hop installation that uses the project can read them.

Compatibility

The transforms and actions generate Cypher. The database type determines which index and constraint statements are generated, and the Graph index and Graph constraint dialogs only offer what the database supports:

Feature	Neo4j	Memgraph	Amazon Neptune	FalkorDB	Apache AGE
Node indexes	Yes, named	Yes, unnamed: identified by label and properties	No, Neptune manages its indexes	Yes, unnamed: identified by label and properties	No, AGE indexes are PostgreSQL indexes
Relationship indexes	Yes	Yes, on a single property	No	Yes	No
Unique and existence constraints on nodes	Yes	Yes. Existence constraints are on a single property.	No	No	No
Node key constraints and relationship constraints	Yes	No	No	No	No
Transactions	Yes	Yes, but index and constraint changes run on their own	Yes	No: every statement is committed on its own	Yes

Feature

Neo4j

Memgraph

Amazon Neptune

FalkorDB

Apache AGE

Node indexes

Yes, named

Yes, unnamed: identified by label and properties

No, Neptune manages its indexes

Yes, unnamed: identified by label and properties

No, AGE indexes are PostgreSQL indexes

Relationship indexes

Yes

Yes, on a single property

Yes

Unique and existence constraints on nodes

Yes

Yes. Existence constraints are on a single property.

Node key constraints and relationship constraints

Yes

Transactions

Yes

Yes, but index and constraint changes run on their own

Yes

No: every statement is committed on its own

Yes

More differences to keep in mind:

FalkorDB: dates, times and binary values are passed to FalkorDB as ISO and Base64 strings.
Gremlin: a vertex has one label, Graph Output uses the first label of a node. The binding names id, label, key and value are reserved by Gremlin servers: don’t use them as parameter names. Amazon Neptune doesn’t run scripts with bindings: on Neptune, use Graph Output, which doesn’t need them. Dates and times are stored as UTC dates, the date type every Gremlin server supports. Graph Output looks nodes up by their keys: create indexes on them on the server for large graphs. There are no transactions: every request is committed on its own. Scripts return a row per result: a map, from project(), elementMap() or valueMap(), gives a field per key, anything else one field called result.
Apache AGE: a node has one label, and label names can’t contain spaces. Cypher statements that return values need to name them: RETURN * isn’t supported.
Notifications of the database, like deprecation warnings or the query planner hints of Memgraph, are advice about a statement which succeeded. The Cypher output, Graph output and Graph query transforms log them, warnings at the basic log level and informational notifications at the detailed log level, and never fail on them.
Cypher you write yourself, for example in the Graph query transform, has to use the syntax of the database.

Capabilities

Each graph database type tells Hop what it supports. The editor of a graph database connection shows this on the Supports line under the graph database type, and its tooltip lists every capability with yes or no and the query language. On the command line, hop-conf prints the same as JSON, see Hop Conf:

--graph-database-types lists every installed graph database type with its capabilities. No connection is needed.
--graph-connection-info=<name> prints the type and capabilities of a connection. Its settings, like the server or the password, are not printed.
--graph-connection-test=<name> tests a connection, like the Test button.

The capabilities of the graph database types in this version of Hop:

Capability Neo4j Memgraph Amazon Neptune FalkorDB Apache AGE Gremlin

Capability	Neo4j	Memgraph	Amazon Neptune	FalkorDB	Apache AGE	Gremlin
`cypher`: the database runs Cypher	Yes	Yes	Yes	Yes	Yes	No
`supportingNodeIndexes`	Yes	Yes	No	Yes	No	No
`supportingRelationshipIndexes`	Yes	Yes	No	Yes	No	No
`supportingIndexes`: node or relationship indexes	Yes	Yes	No	Yes	No	No
`supportingConstraints`	Yes	Yes	No	No	No	No
`supportingVectorIndexes`	Yes	Yes	No	Yes	No	No
`supportingVectorSearch`: searching a vector index for the nearest nodes	Yes	Yes	No	Yes	No	No
`supportingRelationshipVectorSearch`: searching a vector index on relationships for the nearest relationships	Yes	Yes	No	Yes	No	No
`supportingSchemaIntrospection`: reading the labels, relationship types and properties	Yes	Yes	Yes	Yes	Yes	Yes
`supportingSchemaChangesInTransactions`: index and constraint changes run in a transaction	Yes	No	Yes	Yes	Yes	Yes
`supportingShortestPath`: the `shortestPath()` function	Yes	No	No	No	No	No

cypher: the database runs Cypher

Yes

supportingNodeIndexes

Yes

supportingRelationshipIndexes

Yes

supportingIndexes: node or relationship indexes

Yes

supportingConstraints

Yes

supportingVectorIndexes

Yes

supportingVectorSearch: searching a vector index for the nearest nodes

Yes

supportingRelationshipVectorSearch: searching a vector index on relationships for the nearest relationships

Yes

supportingSchemaIntrospection: reading the labels, relationship types and properties

Yes

supportingSchemaChangesInTransactions: index and constraint changes run in a transaction

Yes

supportingShortestPath: the shortestPath() function

Yes

The capability names are the names of the methods of the graph database dialect without is, for example isSupportingVectorIndexes() becomes supportingVectorIndexes. Capabilities added in later versions of Hop or by other plugins appear in the editor and the JSON without changes.