Database mind map

SQL-style queries and relational patterns.

Core idea: Shape rows in this order: filter → relate → aggregate → rank.

Mnemonic: Filter → join → group → rank · Cost: Index the query path

Shape: Filter & join

Signal: The answer needs selected rows plus related data from another key space.

Move: Filter early; choose inner or left join by whether unmatched rows must survive.

Examples: Customers Without Orders, Employees & Managers, Invalid Records

Signal: Many rows must collapse into one answer per key.

Move: Group by the result grain, aggregate values, then filter groups with HAVING.

Examples: Duplicate Emails, Department Counts, Daily Totals

Signal: Rows keep their identity but need rank, neighbor, or running state.

Move: Partition, order, and compute over the frame without collapsing the source rows.

Examples: Nth Salary, Top Per Department, Running Total

Signal: Reads must stay fast and multi-step writes must preserve invariants.

Move: Enforce identity, index selective predicates, and commit dependent mutations atomically.

Examples: Delete Duplicates, Pagination Index, Inventory Transfer