This is the manual for one unit of a Spanish vocational certification in database administration, with the structure that imposes: introduction, development, summary, exercises. Three chapters: storage objects and fragmentation; the language itself — create, alter, drop, select, insert, update, delete, aggregation, subqueries; and transactions and concurrency.
What I took from it
The third chapter. The part covering what happens when two processes write to the same thing at once, and what a transaction actually guarantees, is what separates someone who writes queries from someone who writes queries against a production system. It is dull material and it is where the failures live that surface once a month, at eleven at night, with nobody able to reproduce them.
Out of it came a fixed question: when a process writes into an ERP database from outside, what matters is not only whether the statement is correct but what happens if it runs twice at once or gets cut off halfway. The difference between optimistic and pessimistic control, which the book explains calmly, decides whether a month-end close hangs or duplicates.
The second thing is the data dictionary. Asking the database what tables, columns and constraints it really has, instead of believing the diagram someone drew three years ago. They almost never match, and the database is the one that is right.
What I left out
The rest. The language section is correct and available in any free manual, and the certification format forces it to define terms nobody needs defined. Anyone wanting to learn queries will learn more writing them against a real database; this book is useful to someone sitting the exam, or to someone who wants chapter three.