Service - Graph Handler
Overview
Service that manages a graph database, intended to be deployed any number of times, each instance managing one graph within a project. Most project relationships will be stored in graphs, other services can duplicate subsets of relationship data or element data if they choose or can use the graphs as their store.
As the project grows relationship data can be extended to allow for analysis of relationships, weighting same type connections etc..
Use standardized labeling of nodes to allow graphs to be merged or connected later.
Repository
https://bitbucket.org/stb_working/graph-handler/src/master/
Neo4j graph
Scales well with a large number of node labels / relationship types.
DynamoDB tables
Standard Config Table Per Service
Configuration tags
...
NodeSchema
{
nodeLabel: "xx",
schema: {
identifier: false, // optional (default false), can be used to check if a label being added is an identifier which needs to also set all identifier properties, if true implicitly adds labelCanBeRemoved: false (not immutable/elementCanBeRemoved/allPropertiesImmutable because want case where label is identifier but node can be removed, or properties can be changed)
immutable: false, // optional (default false), if true implicitly adds labelCanBeRemoved: false, elementCanBeRemoved: false, allPropertiesImmutable: true
labelCanBeRemoved: false, // optional (default false), if false cannot remove this label from a node once it has been set
labelCanBeAdded: true, // optional (default true), if false cannot add this label to an existing node
elementCanBeRemoved: false, // optional (default false), if false cannot delete a node when this label exists
allPropertiesImmutable: false, // optional (default false), if true all properties are forced to being immutable, including any properties not set in this schema, if false then check per property setting
restrictProperties: false, // optional (default false), if set can only add properties in this schema
restrictRelationships: false, // optional (default false), if set can only add relationships in this schema (maybe only outgoing relationships?)
properties: {
xxx: {
identifier: false, // optional (default false), if true implicitly adds immutable: true
immutable: false, // optional (default false), if set cannot be changed/removed/added except on creation
},
..
},
relationshipsOutgoing: { // (maybe only outgoing relationships?)
xxx: {
immutable: false, // optional (default false), if set relationships of this type connected to a node with this label cannot be removed after added. Not yet implemented, this would be a second check for relationship connected to this node label only, if relationship schema sets immutable: true then this setting is ignored
},
..
},
}
}
- settings per node label
- if multiple labels are supplied in a query for a node (eg if have multiple identifier labels) we combine (or iterate through) all relationship/property settings when validating query
RelationshipSchema
{
relationshipType: "xx",
schema: {
immutable: false, // optional (default false), if true implicitly adds elementCanBeRemoved: false, allPropertiesImmutable: true
elementCanBeRemoved: false, // optional (default false), if false cannot remove this relationship once created
allPropertiesImmutable: false, // optional (default false), if true all properties are forced to being immutable, including any properties not set in this schema, if false then check per property setting
restrictProperties: false, // optional (default false), if set can only add properties in this schema
properties: {
xxx: {
identifier: false, // optional (default false), if true implicitly adds immutable: true
immutable: false, // optional (default false), if set cannot be changed/removed/added except on creation
},
..
},
}
}
- global settings for relationship types
Query language
Focus on using Cypher.
Querying data in graph database
Planning all Put/Update/Delete/CRUD queries to pass through an API/Graph Handler Lambda functions.
Read/Get queries could query the graph directly using eg IAM constraints/throttling, but currently unable to lock the EC2 on a private server and allow Lambda's outside of the VPC to access the database, so all requests will pass through Graph Handler Lambda functions until infrastructure allows for a better design.
Node/relationship schema
setup
Each service sends schema to graph handler in InitialSetup. If we do this directly by put into Dynamo then the access to do this is deleted along with the IntialSetup Lambdas. We could alternatively have a Lambda in the GraphHandler that does it but we would want to remove that once complete.
Graph actions like setting constraints would need to be handled by a Lambda.