Tuesday, October 24

Author: Peter Eisentraut

More robust collations with ICU support in PostgreSQL 10

Eisentraut's PlanetPostgreSQL, PostgreSQL
In this article, I want to introduce the ICU support in PostgreSQL, which I have worked on for PostgreSQL version 10, to appear later this year. Sorting Sorting is an important functionality of a database system. First, users generally want to see data sorted. Any query result that contains more than one row and is destined for end-user consumption will probably want to be sorted, just for a better user experience. Second, a lot of the internal functionality of a database system depends on sorting data or having sorted data available. B-tree indexes are an obvious example. BRIN indexes have knowledge of order. Range partitioning has to compare values. Merge joins depend on sorted input. The idea that is common to these different techniques is that, roughly speaking, if you have sorted (more…)

PostgreSQL 10 identity columns explained

Eisentraut's PlanetPostgreSQL, PostgreSQL
For PostgreSQL 10, I have worked on a feature called "identity columns". Depesz already wrote a blog post about it and showed that it works pretty much like serial columns: CREATE TABLE test_old ( id serial PRIMARY KEY, payload text ); INSERT INTO test_old (payload) VALUES ('a'), ('b'), ('c') RETURNING *; and CREATE TABLE test_new ( id int GENERATED BY DEFAULT AS IDENTITY PRIMARY KEY, payload text ); INSERT INTO test_new (payload) VALUES ('a'), ('b'), ('c') RETURNING *; do pretty much the same thing, except that the new way is more verbose. ;-) So why bother? Compatibility The new syntax conforms to the SQL standard. Creating auto-incrementing columns has been a notorious area of incompatibility between different SQL implementations. Some have lately (more…)