November 2013 Summaries

The GraphConnect conference, hosted by Neo4j, brought together attendees from both sides of the Atlantic for a day of learning and hacking. In New York, over 150 attendees gathered for presentations on various graph-related topics, including data modeling, route finding, and analyzing career paths with College Miner. The keynote speaker, Peter Olson from Marvel Entertainment, discussed how graphs are used to represent the vast fictional content of the Marvel Universe. Meanwhile, in London, almost 200 attendees participated in two tracks of presentations, including talks on object graph mapping with Spring Data Neo4j 3 and adoption of a graph database in the insurance sector. The conference also featured GraphClinics, one-on-one consulting sessions with experts, and a special appearance by a TARDIS from Doctor Who.

The next major version of Neo4j, 2.0, is now feature-complete and available as a Release Candidate build. The release includes several key changes to the Cypher syntax, including support for patterns with properties in the MATCH clause, an OPTIONAL MATCH clause to handle optional details, and simplified syntax for MERGE on MATCH and ON CREATE. Additionally, Neo4j 2.0 introduces new features such as get-or-create operations for single nodes or relationships using MERGE, and changes to how Cypher handles null values. However, this release is not compatible with earlier 2.0.0 milestones and requires manual upgrade from previous milestone versions due to incompatible changes made to store files. Users are advised to carefully back up their data before attempting a manual upgrade.

Sarah Mei, a developer, faced challenges while managing highly connected data using document databases. These databases excel at storing single representations of aggregates but struggle with multiple views and connections between documents. Real-world use cases like TV show applications and social networks led to difficulties with the chosen data model and database, prompting a switch to relational databases for better modeling fidelity. However, this comes with performance issues due to query complexity. Graph databases offer a solution by embracing rich connections between domain entities. Two live graph data models, one for a TV show application and another for a social network, demonstrate how graph databases can handle such use cases efficiently. These models showcase patterns like shows, seasons, and episodes, as well as users writing reviews for individual episodes. Graph databases are gaining popularity, with Neo4j being widely used in production setups for over 10 years, and are now considered an uncontroversial choice for domains with richly connected data.

The MusicBrainz database is an open content music database that captures information about artists, their recorded works, and the relationships between them. It has around 1000 active users, nearly 800,000 artists, 75,000 record labels, over 1,200,000 releases, and more than 12,000,000 tracks. The data is identified by its MusicBrainz Identifier (MBID), which is a universal UUID for the music recording industry. The database's core data is licensed under CC0, placing it in the public domain. The relational model of MusicBrainz consists of several tables linked together to represent artists, labels, releases, and relationships between them. The database can be imported into Neo4j using the SQL2GRAPH tool, which creates a schema mapping from SQL tables to graph nodes and properties, and then exports the data in SQL format. This process involves three steps: creating a temporary table, exporting nodes and relationships as TSV files, and importing these files into Neo4j using the batch-import tool. The resulting import takes around 5 minutes on a small machine, with over 7 million nodes and 72 million relationships imported.