Mysql - drop empty columns to update only set values - php

Question

I am using mysql+php. The php provides interface to import XLSX table of goods into Mysql table(s). I am using a temporary table created with CREATE TABLE LIKE i.e. empty clone of live table for user evaluation before it is merged with live table
I am using INSERT INTO finalTable SELECT * FROM importTable ON DUPLICATE KEY UPDATE .... to avoid INSERT for every record. The fields to be updated on duplicate key are determined by php with SHOW COLUMNS i.e. all columns from the temporary table.
My problem is that the XLSX does not necessarily contain ALL the fields(name, price, category) i.e. it may be just an update of existing records. The current method will update ALL fields no matter if they were set or not during the import i.e. if the XLSX contained only price update, the rest of fields is null and would overwrite current values.
I tought to alter the temporary table by removing columns(not rows!) that are null for all rows. This way the SHOW COLUMNS would return only update-able columns. If a fields needs to be zeroed, it would be easy to set a special value e.g. "!X!" and use single UPDATE sentence to do so.
Is there a method for this or do you have any better suggestion(I am open to abando ON DUPLICATE KEY too)?

Answer 1

I'm a bit confused by your question.
I am using INSERT ... ON DUPLICATE KEY UPDATE to avoid INSERT for every record.
Like it says, it inserts or updates, you don't avoid anything. If you want to avoid inserts use INSERT IGNORE.
If I were you, I'd merge the tables in two steps.
First step, the insert (just the necessary ones):
INSERT INTO finalTable (col1, col2, ...)
SELECT i.col1, i.col2
FROM importTable i
LEFT JOIN finalTable f ON i.ID = f.ID
WHERE f.ID IS NULL;
Second step, the update:
I don't understand, why you want to delete columns. Unnecessary step which might take a while, and most importantly, your problem of updating with NULL still persists if just a few rows in a column are NULL in your import table. So, here's the solution.
UPDATE finalTable f
INNER JOIN importTable i ON f.ID = i.ID
SET f.col1 = COALESCE(i.col1, f.col1),
f.col2 = COALESCE(i.col2, f.col2),
...;
The trick in the second step is, to use COALESCE(). This function returns the first of its parameters which is not null. So if you have a column which is null in your import table, the value in the final table stays as it is.
UPDATE:
If you insist on having just one statement, you can of course do
INSERT INTO final f
SELECT * FROM import i
ON DUPLICATE KEY UPDATE
SET f.col1 = COALESCE(i.col1, f.col1),
f.col2 = COALESCE(i.col2, f.col2),
...;