From 8df90408be68beed8e8cfef771367e17a30b6792 Mon Sep 17 00:00:00 2001
From: Shlomi Noach <2607934+shlomi-noach@users.noreply.github.com>
Date: Mon, 28 Oct 2024 16:23:01 +0200
Subject: [PATCH] using subselect. Adding comments

Signed-off-by: Shlomi Noach <2607934+shlomi-noach@users.noreply.github.com>
---
 go/mysql/flavor_mysql.go | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/go/mysql/flavor_mysql.go b/go/mysql/flavor_mysql.go
index 0cfa41ad548..b840c3cca36 100644
--- a/go/mysql/flavor_mysql.go
+++ b/go/mysql/flavor_mysql.go
@@ -413,6 +413,20 @@ const BaseShowTables = `SELECT t.table_name,
 `
 
 // TablesWithSize80 is a query to select table along with size for mysql 8.0
+// Note the following:
+//   - `TABLES`.`TABLE_NAME` has `utf8mb4_0900_ai_ci` collation.  `INNODB_TABLESPACES`.`NAME` has `utf8mb3_general_ci`.
+//     We normalize the collation to get better query performance (we force the casting at the time of our choosing)
+//   - InnoDB has different table names than MySQL does, in particular for partitioned tables. As far as InnoDB
+//     is concerned, each partition is its own table.
+//   - We use a `UNION ALL` approach to handle two distinct scenarios: tables that are partitioned and those that are not.
+//     Since we `LEFT JOIN` from `TABLES` to `INNODB_TABLESPACES`, we know we already do full table scan on `TABLES`. We therefore
+//     don't mind spending some extra computation time (as in `CONCAT(t.table_schema, '/', t.table_name, '#p#%') COLLATE utf8mb3_general_ci`)
+//     to make things easier for the JOIN.
+//   - We utilize `INFORMATION_SCHEMA`.`TABLES`.`CREATE_OPTIONS` column to tell if the table is partitioned or not. The column
+//     may be `NULL` or may have multiple attributes, one of which is "partitioned", which we are looking for.
+//   - In a partitioned table, InnoDB will return multiple rows for the same table name, one for each partition, which we successively SUM.
+//     We also `SUM` the sizes in the non-partitioned case. This is not because we need to, but because it makes the query
+//     symmetric and less prone to future edit errors.
 const TablesWithSize80 = `SELECT t.table_name,
 		t.table_type,
 		UNIX_TIMESTAMP(t.create_time),
@@ -420,7 +434,7 @@ const TablesWithSize80 = `SELECT t.table_name,
 		SUM(i.file_size),
 		SUM(i.allocated_size)
 	FROM information_schema.tables t
-	LEFT JOIN information_schema.innodb_tablespaces i
+	LEFT JOIN (SELECT name, file_size, allocated_size FROM information_schema.innodb_tablespaces WHERE name LIKE CONCAT(database(), '/%')) i
 	ON i.name = CONCAT(t.table_schema, '/', t.table_name) COLLATE utf8mb3_general_ci
 	WHERE
 		t.table_schema = database() AND IFNULL(t.create_options, '') NOT LIKE '%partitioned%'
@@ -434,7 +448,7 @@ SELECT t.table_name,
 		SUM(i.file_size),
 		SUM(i.allocated_size)
 	FROM information_schema.tables t
-	LEFT JOIN information_schema.innodb_tablespaces i
+	LEFT JOIN (SELECT name, file_size, allocated_size FROM information_schema.innodb_tablespaces WHERE name LIKE CONCAT(database(), '/%')) i
 		ON i.name LIKE (CONCAT(t.table_schema, '/', t.table_name, '#p#%') COLLATE utf8mb3_general_ci)
 	WHERE
 		t.table_schema = database() AND t.create_options LIKE '%partitioned%'