Why this is in my collection
From the publisher:
A practical guide for accelerating your data science, data analytics, and data engineering workflows
Highlights
- The analytical database as a library, not a server — DuckDB runs in-process, so the infrastructure layer effectively disappears; what remains is pure structure (SQL over columnar data), a live demonstration that the platform can shrink to almost nothing while the model keeps all its value.
- Query files where they lie — Aubury and Letcher lean hard on reading CSV, Parquet, and JSON directly; the files stay the source of truth and the database becomes a queryable view over them — exactly the vault architecture where markdown+YAML is the master and DuckDB is the lens.
- SQL is the durable interface — a decades-old declarative language riding on a brand-new engine; the boring, proven layer persists while execution engines churn underneath, the technology-selection lesson in miniature.
- Removing friction changes behavior — when an analytical query is one line with no cluster, no service, and no upload, you actually run it; personal-scale data engineering becomes real, which is the precondition for treating your own knowledge and content as data.
- Analytics embedded in the working environment — DuckDB inside Python notebooks and scripts rather than a warehouse you visit; the tool comes to the workflow, inverting the platform-centric default of the cloud era.
- Local-first as a legitimate architecture — most datasets that matter to an individual or team fit on one machine; the book quietly deflates the assumption that serious analytics requires distributed systems.
Highlights on this page are generated with the help of AI.