REPLACE mechanically runs
INSERT. That may change the PRIMARY KEYs.
Here is something else you can do.
Suppose your table is called name_city and it looks like this:
CREATE TABLE name_city
id INT NOT NULL AUTO_INCREMENT,
name VARCHAR(30) NOT NULL,
city VARCHAR(30) NOT NULL,
PRIMARY KEY (id)
and you want to do the
LOAD DATA INFILE.
Here are your steps to update the city based on id:
Step 01) Create another table to use for the import
CREATE TABLE name_city_import LIKE name_city;
Step 02) Drop the name column from the import table
ALTER TABLE name_city_import DROP COLUMN name;
Step 03) Perform the import into
LOAD DATA INFILE 'somefile.csv' INTO TABLE name_city_import ...
Step 04) Perform an UPDATE JOIN
UPDATE name_city_import B
INNER JOIN name_city A USING (id)
SET A.city = B.city;
Step 05) Remove the import table
DROP TABLE name_city_import;
Give it a Try !!!