A Database Index Is a Second, Ordered Copy of Some Columns, and Every Write Pays to Maintain It
The floor allows only 'primary keys are fast'. Selectivity, composite column order and covering indexes are all statements about this structure and cannot be written without it, and the write-cost half is what stops the learner from answering every later slow-query node with 'add an index'. It is also the node that makes an index-audit decision possible, which is the practical Apply question a real application eventually faces.
Questions this Concept answers
Why does an indexed lookup grow with the logarithm of the table's size rather than in proportion to it?
J
jeremy
Video
The Write, Bloat, and Locking Costs of Adding an Index in PostgreSQL
An index in PostgreSQL is a secondary, sorted data structure — typically a B-tree — maintained alongside the table, whose shallow branching depth converts a full sequential scan into a few page reads…