Create 2024-10-08-partition-planning.md

_posts/2024-10-08-partition-planning.md

@@ -0,0 +1,38 @@
+---
+layout: writing
+title: "Postgres: planning with partitions"
+tags: [writing, pg]
+last_updated: 2024-10-08
+---
+# Postgres: planning with partitions
+
+Oct 8, 2024
+
+I've been meaning to write about this topic for a while so here it is.  The main goal is
+to explain how the number of partitions a table has affects the planning time of a query
+referencing only that table, both when the query's `WHERE` condition can be used to
+perform [partition pruning](https://www.postgresql.org/docs/current/ddl-partitioning.html#DDL-PARTITION-PRUNING)
+and when it can't.  I will mention the names of some functions in the Postgres source
+code that are involved in the planning of such queries so that anyone who is interested
+to further explore can jump into the source code to understand this better.
+
+Let's get right into it.  We'll use the following partitioned table:
+
+```
+create table foo (a int) partition by range (a);
+create table foo_1 partition of foo for values from (1, 1000);
+create table foo_2 partition of foo for values from (1000, 2000);
+...
+create table foo_N partition of foo for values from ((N - 1) * 1000, N * 1000);
+```
+
+When a user sends a query like this:
+
+```
+select * from foo where a = 1;
+```
+
+Partitions `foo_1` to `foo_N` are not looked at until the query reaches
+the planner function `query_planner()`.  `query_planner()` is a function that is
+responsible for deciding how to scan the individual tables and also how join if
+there is more than one table in the query.