Changes between Version 2 and Version 3 of Ticket #33885, comment 47
- Timestamp:
- 05/25/2017 10:42:19 AM (7 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #33885, comment 47
v2 v3 37 37 * Future versions of MySQL and MariaDB will default to `innodb_large_prefix` [https://mariadb.com/kb/en/mariadb/xtradbinnodb-server-system-variables/#innodb_large_prefix being `ON` by default]. This won't directly avoid temporary tables & filesorts, but it will generally improve query performance the way @matt_fw eluded to earlier. 38 38 * For anyone that has added an index to their `meta_value` columns (not uncommon for sites that do complex `JOIN`s), you'll want to keep a similar eye on your customizations. 39 * Because of how rare & weird it is to `ORDER BY meta_key`, I do think it's OK to notoptimize the meta database tables for his behavior.40 * I do not think it's OK for our meta database tables to all have an ineffective `meta_key` index. Thankfully, even when their lengths are misaligned, the `meta_key` index does still function as intended for all queries other than `ORDER BY meta_key`.39 * Because of how rare & weird it is to `ORDER BY meta_key`, I do think it's OK to ''not'' optimize the meta database tables for his behavior. 40 * Thankfully, even when their lengths are misaligned, the `meta_key` index does still function as intended for all queries other than `ORDER BY meta_key`, so the `meta_key` index is not wholly ineffective - only for `ORDER BY` queries. 41 41 * @dd32's transient approach in [https://core.trac.wordpress.org/ticket/33885#comment:44 #44] is a sound work-around for this isolated problem, for now. 42 42 * For later, we should consider what a redux of this meta-box experience looks like.