Chapter 2

Data Models and Query Languages

2.1 Relational Model vs Document Model
2.2 Query Languages for Data
2.3 Graph-Like Data Models

Relational Model vs Document Model

The relational model (SQL) has dominated since the 1980s, but document databases (NoSQL) gained traction for schema flexibility and better data locality. Document model works well when data has a tree structure (one-to-many). Relational model wins for many-to-many relationships and joins. The impedance mismatch between relational tables and application objects drove the document model's popularity.

Query Languages for Data

SQL is declarative — you specify what you want, not how to get it. This lets the optimizer choose the best execution strategy. Contrast with imperative approaches like IMS/CODASYL. MapReduce is neither fully declarative nor imperative; it's based on map and reduce functions that must be pure. MongoDB added the aggregation pipeline as a more ergonomic alternative.

Graph-Like Data Models

When many-to-many relationships are central to your data, graph models are natural. Two main approaches:

Property graph model (Neo4j): vertices + edges, each with properties. Queried with Cypher.
Triple stores (RDF): subject-predicate-object triples. Queried with SPARQL.

Graph models excel at traversing relationships with variable depth — something relational joins handle awkwardly.