Analytics Engineering with SQL and dbt Building Meaningful Data Models at Scale (Rui Pedro Machado, Helder Russa)（Z-Library）

【One-Line Pitch】 A practical field guide to building trustworthy, scalable data models with SQL and dbt, aimed at analysts, data engineers, BI developers, and data scientists who want to move from ad-hoc scripts to a collaborative, testable transformation platform. 【Book Arc】 - **Opening (~0%–10%)**: Frames the shift from warehouses to lakes and introduces the analytics engineer role, dbt's place in the stack, and how the book is organized across six chapters. - **Early (~10%–30%)**: Traces the evolution of data management—from Inmon and Kimball through Hadoop, Redshift, and cloud platforms—and explains why SQL and stored procedures for ETL/ELT hit their limits. - **Early–Middle (~30%–50%)**: Covers data modeling fundamentals: normalization and its trade-offs, dimensional modeling with star and snowflake schemas, Data Vault concepts, and the pain of monolithic SQL scripts. - **Middle (~50%–70%)**: Moves into SQL practice for analytics, including common table expressions, window functions, distributed SQL processing, and hands-on data manipulation with DuckDB, Polars, and FugueSQL. - **Late (~70%–90%)**: Introduces dbt itself—its design philosophy, data flow, dbt Cloud setup with BigQuery and GitHub, the Cloud UI and IDE, and the structure of a dbt project using the Jaffle Shop example. - **Ending (~90%–100%)**: Focuses on modular data models, testing transformations, and generating documentation so that models are reusable, verifiable, and understandable by others. 【Key Takeaways】 - **The analytics engineer bridges data engineering and analytics** (Early): this role designs storage, builds pipelines, and collaborates with data scientists, making reliable insights possible without owning the entire platform. - **Normalization is for OLTP, not analytics** (Middle): transactional systems optimize for fast writes and integrity, while analytics systems need read-optimized, denormalized structures that support historical analysis. - **Star and snowflake schemas trade simplicity for integrity** (Middle): star schemas simplify queries; snowflake schemas reduce redundancy but require more joins—choose based on dataset size and relationship complexity. - **Monolithic SQL scripts don't scale as a practice** (Middle): without version control, dependency management, or idempotency, large SQL files become fragile and unreusable, which is the problem dbt is designed to solve. - **dbt brings software engineering discipline to SQL** (Late): project structure, YAML configuration, sources, models, and tests turn transformations into maintainable, collaborative assets. - **Testing and documentation are first-class concerns** (Ending): building tests for transformations and generating data documentation ensures models stay trustworthy as they evolve. - **SQL remains the core language of analytics engineering** (Middle): CTEs, window functions, and distributed SQL patterns are essential skills, with DuckDB, Polars, and FugueSQL as practical tools for local and scalable manipulation. 【Reading Tips】 - Skim the historical evolution in Chapter 1 if you already know the data landscape; slow down on the analytics engineer role and its responsibilities. - Deep-read the data modeling chapter—normalization, star/snowflake, and Data Vault concepts are the conceptual foundation for everything dbt does later. - Treat the SQL chapter as a working reference: run the DuckDB and Polars examples yourself rather than just reading them. - When you reach the dbt chapters, follow the Jaffle Shop setup end-to-end; the project structure, YAML files, and tests only click when you build them. - Don't skip the testing and documentation sections—they are what separate a hobby project from a production transformation platform. 【Coverage Limits】 The excerpts cover the book's structure, historical context, data modeling concepts, SQL tooling, and dbt project setup, but do not include detailed code from later dbt chapters or advanced deployment topics. Specific chapter-level depth beyond the table of contents and sampled sections is not fully represented.