Copy field into column before Replace Into - mysql

Question

I have a database which stores user-published articles. The owner can modify their article at any time.
I do want to add a backup feature, in case the user accidentally deletes the content of their article or something else goes wrong when they update it.
For this reason, I have the content column which stores the content of the article, as well as a backup_content which is intended to keep a copy of the content before the last update.
The user has a "Restore" button which is meant to replace the new content with the backup. Very much like an "Undo" feature.
My prepared statement to insert/update an article is as follows:
REPLACE INTO custom_pages (name, banner_url, full_url, backup_content, content, updated_on) VALUES (?, ?, ?, content, ?, CURRENT_TIMESTAMP);
Here, I tried putting the previous value of content in backup_content and then changing content with the new value. Doing so sets the backup_content to NULL however.
I've seen a few answers on SO on how to achieve a copy, but those answers seem to apply strictly for update and insert, and don't seem to work in Replace queries. I'd prefer one statement over two, and that's where I'm having trouble.
Is there any way to achieve such copy in a single Replace statement?

Answer 1

I would also place my support behind Gordon Linoff's suggestion that you create a continuous update history via triggers and one-to-many related tables.
However, if a significant architectural change is not practical for you right now, you can achieve what you are attempting with INSERT INTO...ON DUPLICATE KEY UPDATE instead of the older REPLACE INTO feature.
Using REPLACE INTO...SELECT FROM may result in more than one access against the table's index, but INSERT INTO...ON DUPLICATE KEY UPDATE should hit it only once.
Since name has a unique index, the presumption is that you never attempt to UPDATE, and instead always execute an INSERT which copies the old value to backup_content.
-- Inserting a row which does not yet exist..
INSERT INTO custom_pages (name, banner_url, full_url, content)
VALUES ('uniquename', 'http://example.com', 'http://example.com', 'this is the original content');
-- In practice, you use this format:
-- uniquename already exists, so update necessary fields
INSERT INTO custom_pages (name, banner_url, full_url, content)
VALUES ('uniquename', 'http://example.com', 'http://example.com', 'this is new content')
ON DUPLICATE KEY UPDATE
-- Update from the VALUES() list
banner_url = VALUES(banner_url),
-- Set backup_content to old content BEFORE updating
-- content from VALUES()
backup_content = content,
content = VALUES(content),
updated_on = NOW();
Using this method, you would never use the first INSERT statement without its ON DUPLICATE KEY clause. Instead, always use the second one; rows that don't exist by unique key will be created, those that already exist will be updated.
Here it is in action: http://sqlfiddle.com/#!9/2f687/1

Answer 2

I think you should re-think your data structure. If you want to preserve history, then use a separate table not column. Something like custom_pages_history. You would remove the backup_content column from your table and instead rely on the history table.
Then, define a trigger on inserts and updates to insert a row into the history table.
The advantages of this approach are:
You have complete history of all the articles.
The changes will be timestamped.
A user can go back to any earlier version of the article, if desired.
This doesn't directly answer your question about replace. Instead of replace you would do an update from the history table.