Graph-Relational Model

Tabular Data

Actor Table

first name	last name	favorite food
John	Travolta	🍕
Samuel	Jackson	🍔
Uma	Thurman	🥦

• Entity type: implicit in the table name
• Identity of entities: either implicitly managed (error-prone) or explicit in some "key" column (beware of non-uniqueness of keys between tables)
• Rows: each row typically corresponds to an entity (or atom in general). Identity of each entity (row) may be made explicit by a key (local ID) column.
• Columns: The column name maps onto the relation type.
• Cell / Value in Table: this corresponds to the target element / actual value in the semantic triple.

Plain Python Data

[  
{'first_name': 'John',  'last_name': 'Travolta','favorite_food':'🍕'},  
{'first_name': 'Samuel','last_name':'Jackson',  'favorite_food':'🍔'},  
{'first_name': 'Uma',   'last_name': 'Thurman', 'favorite_food':'🥦'},  
]  

Semantic Triples (Facts)

[  
(ET.Person[:'p1'], RT.FirstName, 'John'),  
(ET.Person[:'p1'], RT.LastName, 'Travolta'),  
(ET.Person[:'p1'], RT.FavoriteFood, '🍕'),  
  
(ET.Person[:'p2'], RT.FirstName, 'Samuel'),  
(ET.Person[:'p2'], RT.LastName, 'Jackson'),  
(ET.Person[:'p2'], RT.FavoriteFood, '🍔'),  
  
(ET.Person[:'p3'], RT.FirstName, 'Uma'),  
(ET.Person[:'p3'], RT.LastName, 'Thurman'),  
(ET.Person[:'p3'], RT.FavoriteFood, '🥦'),  
]  

Zef Graph

!400
This is shown in very granular form here. For practical purposes, we usually have a more compressed view when visually displaying graphs.

Which Problems does the Graph-Relational Model Address?

Or why it is worth departing from the trodden paths?

	SQL-Based Relational DBs	Labeled Property Graphs	Graph-Relational
Schema	✅	Typically no strong support	✅
Data model decoupled from representation	❌	sometimes	✅
Supports fields("properties") on relations	❌	✅ (typically)	✅
no painful schema migrations	❌	✅	✅
Strong transactional model (ACID)	✅	❌ (commonly)	✅
good for thinking in tables	✅	❌	✅
good for thinking in graphs	❌ (not obvious to non-pros)	✅	✅
good for thinking in objects	❌ (ORMs have problems)	❌	✅
maps onto differential computation / streaming	3 / 5	1 / 5	5 / 5
conceptual soundness: based on formal logic	3 / 5	3 / 5	5 / 5
first class support for facts over time	❌	❌	✅

Comparison to Relational Model

• Thinking in terms of entities and relations allows you to focus more on the domain, not the representation of the data. It allows you to think at a higher level: types and sets instead of join tables and query patterns.

Related

• ZefDoc - Representing Temporally Qualified Information in ZefDB