[chaley]

I would do something like this. I didn't try to run it, but see below.

	# I assume you have an old API db reference. Get a new_api reference
	ndb = db.new_api
	book_to_id_map = {}
	for i, title, last_modified, isbn in extracted_ids:
		# Get the existing identifiers for the book
		identifiers_for_book = ndb.field_for('identifiers', i)
		# Add/replace the ISBN identifer
		identifiers_for_book['isbn'] = isbn
		# Save the (probably) updated list of identifiers
		book_to_id_map[i] = identifiers_for_book
	# Set all the books' identifier values
	ndb.set_field('identifiers', book_to_id_map)

As a test I ran this in the template tester. It worked. The first 4 books with more than one identifier ended up including "isbn:grumblebunny"

python:
def evaluate(book, context):
	db = context.db.new_api
	ids = list(db.search("identifiers:#>2"))
	book_to_id_map = {}
	for id_ in ids[0:4]:
		identifiers_for_book = db.field_for('identifiers', id_)
		identifiers_for_book['isbn'] = 'grumblebunny'
		book_to_id_map[id_] = identifiers_for_book
	db.set_field('identifiers', book_to_id_map)
	return 'foo'