nichtich
commented
8 months ago Tomaszuk et a.l. (2019) did a survery of property graph database systems and serialization formats. The support of different graph features in database systems can be summarized as following:
- node labels
- optional: Neo4J, JanusGraph, Amazon Neptune, InfoGrid, Memgraph, AgensGraph
- repeatable: Neo4j, Amazon Neptune, InfoGrid, Memgraph, AgensGraph
- edge labels:
- optional: none
- repeatable: AgensGraph
- edges
- undirected: OrientDB, ArangoDB, JanusGraph, TigerGraph, InfiniteGraph, InfoGrid, Sparksee, VelicityDB
- multiple between same nodes: all but InfoGrid
- multiple between same nodes with same edge label(s): all but TigerGraph and InfoGrid
- properties
- multiple values with same key: all but InfoGrid, Sparksee, and HGraphDB
- null-values: OrientDB, ArangoDB, InfiniteGraph, InfoGrid, Sparksee, Memgraph, VelocityDB, HGraphDB
In addition there are differences in allowed labels and property keys and in supported data types.
Translated to profiles this could be e.g.
node-labels:0(none) |1(required, one) |?(optional, one) |+(required, many) |*(optional, many)edge-labels:0(none) |1(required, one) |?(optional, one) |+(required, many) |*(optional, many)node-label-pattern: regexedge-label-pattern: regexedge-direction:directed|undirected|mixededge-set:nodes(one edge between two nodes at most) |labels(restricted by label) |value(restricted by label and properties)property-value:one|set|list|mixed(does any system support lists instead of sets?)data-types: set ofstring,null, ...data-type-int-max,data-type-int-min...
A profile is a set of rules restricting the set of possible property graphs. Examples:
Profiles help to compare capabilities of property graph formats and databases.
A profile can be applied as filter to modify a graph (e.g. rewrite node identifiers to numbers) or as validator.
Possible argument
--profile(repeatable),--validateor--warn.