Dr Jim Webber, Neo4 Chief Scientist, discusses the mainstream adoption of graphs in infrastructure software. Databases that have traditionally supported and are natively-designed for other models of data have added graph capabilities to their offering. These non-native operators and layers provides syntactic sugar to enable graph-like traversals.
What’s missing? Under the covers it’s still the same data structures with the same algorithmic cost (B-tree style indexes, for example). In the case of building graph operators on document stores, you have n*log(n) cost of traversing the graph. This databases which are native for document storage and retrieval simply don’t perform how we expect and can even quite conceivably lead to corrupted graphs (such as dangling edges).
We’re super excited to see greater awareness of the capabilities of graph thinking, but anyone serious about graphs should use a native graph database like Neo4j.