Re: Dumping enormous MYISAM table with BLOB based on WHERE clause

[Rick James]

More minor things:

> `previewfile` longblob,
> `preview_exists` tinyint(1) default '0',

Both are NULL, you could get rid of the existence flag and simply use
previewfile IS NULL

> `id` varchar(50) NOT NULL default '',
> `list_id` varchar(50) NOT NULL default '',
> PRIMARY KEY (`id`,`list_id`)

Seems strange because (1) 2-part PK and (2) VARCHAR(50). But there is nothing 'bad' about it in this context.

> PRIMARY KEY (`id`,`list_id`)
-->
PRIMARY KEY (`id`,`list_id`, serve_date)
PARTITION BY RANGE(serve_date)...
Note: You necessarily lose the UNIQUEness check on (id, list_id). If that is a problem, it will have to be handled in code. Hmmmm... It might be possible to move the two fields to another table and have them map to an AUTO_INCREMENT on this table. The extra table would not be partitioned and would allow for the UNIQUEness test; it would (presumably) be small enough so that you don't bother ever purging it. (We would need to check numbers, like average sound-file size, # new rows per day, etc., to see if that is reasonable.)