Table “books” and “books_tmp” is identical.

“books” table is production state.
“books_tmp” is sandbox/testing/pre-production state.

Data comes in csv files >> dumped to “books_tmp” >> checked, sorted, etc. >> good data moved to “books”, row by row.

Before moved, “books_tmp” data is checked for duplicates in “books”, using ISBN (unique).
Upon inserted into “books”, (auto increment) id is checked for duplicate, if so, update the id.

This is the query I use (MySQL):
INSERT INTO books
SELECT * FROM books_tmp
WHERE books_tmp.ISBN NOT IN (SELECT ISBN FROM books)
ON DUPLICATE KEY UPDATE books.id = LAST_INSERT_ID(books.id)