utfmb4 database

[Underdog-01]

Description

With a database that has been installed as ie. urf8mb4 the SMF 2.1 installation creates all tables as 3-Byte UTF8 and does not conform to the default database setting/structure.

Steps to reproduce

1. Create utf8mb4_unicode_ci database
2. Change database type in Settings.php file to utf8mb4 (set $db_mb4 variable as true)
3. Run the SMF 2.1 installer
4. Use a database client (ie. phpmyadmin) to view table structure of the database tables that SMF 2.1 created

Environment (complete as necessary)

• Version/Git revision: Current
• Database Type: MySQL / MariaDB
• Database Version: MySQL 8.0.12 / MariaDB 10.3.10
• PHP Versions: PHP 5.6 - PHP 7.2.11

Additional information/references

The settings variable should have the behavior of forcing the SMF 2.1 installer to create tables as utf8mb4. In my opinion it is possible to read the database default value and use that instead of relying on the variable in Settings.php. In either case the installation process fails to create the tables as utf8mb4.

[albertlast]

The setup process didn't read the settings file.

The user journey behind this setting is:

• that you install your board in mysql with mb3
• than change by using phpmyadmin all tables to mb4
• change the settings to mb4
• run the entity converter

From my side(i'm the guy how introduce the mb4 option in settings) is actual implementation enough.

[Underdog-01]

In my opinion it would be better to install as 4-byte initially where the database default setting is mb4.
The entity converter is a great option either way but I would implement adhering to the db default.

What about mod installation behavior?
If the forum has already been converted to mb4 & a mod uses smcFunc to create its tables, are those tables initially mb4 or must the entity converter be run yet again?

[albertlast]

Since this feature only sugar,
without any needs to be ther (like you self notice you can safe mb4 information in a mb3 enviorment when you convert it)
When you mean it introdruce to many issue,
we can removed and you had to use postgres as rdbms when you realy want a mb4 env.

[Underdog-01]

Actually I see a problem with sorting when it uses html entities instead of the proper character code only native to utf8mb4 unicode. Database queries will not be sorted properly when dealing with ie. names, topic titles, etc. when those use 4-byte characters that have been converted to 3-byte decimal entities.

[albertlast]

this is the mysql way of live.

[Underdog-01]

This is perhaps the pre - 2010 mysql way of life.

[MissAllSunday]

@albertlast do you think this is a valid issue? Why do you think installing in mb4 when the database is already set to it isn't a feasible feature to add?

[albertlast]

Nope,
in my eyes in the existing solution enough.

[MissAllSunday]

Cool, next question would be, how common is to encounter an already created DB set to mb4/ full utf8?

If this is something we can easily detect when installing/upgrading and SMF runs smoothly on either 3 or 4 bytes then I would prefer to add support for full utf8 when the db is already ready for it.

[albertlast]

from my pov,
you will see this on release.

when realy, something could be added later on.

[sbulen]

I would bet that virtually 100% of all testing thus far has been in utf8, not utf8mb4, since our install script hard-codes to utf8:
https://github.com/SimpleMachines/SMF2.1/blob/e5b77f4133d81524fda48ad5ad7452ce7032fd77/other/install.php#L1177

utf8mb4 really is the way to go. In 2.1.x, I think we should mirror how we supported utf8 in 2.0.x. :

• allow a utf8mb4 option upon install
• provide a utf8mb4 conversion utility under forum maintenance (for upgraded forums)
• provide an entities conversion utility

The problem is it's completely untested. So the question is WHEN...

I recommend this be in the Final release at the earliest... Maybe even a follow on.

[MissAllSunday]

Yep, I also think we should go with mb4 and if there is an easy way for us to take advantage of an already set up DB then we should use it.

So yeah, at minimum we should offer support for installing SMF on mb4 when its available.

[chopinesque]

Something similar was done for Drupal 3 years ago

https://www.drupal.org/project/utf8mb4_convert
https://docs.acquia.com/acquia-cloud/manage/database/utf8mb4/