I did it in the following way.

insert into romappingfinal(LAN,romappingfinalid,branch_code,
RO_CODE,main_ro_code,loan_end_date,ro_status,main_ro_status,
ro_transfer_status,main_ro_transfer_status,ro_name,main_ro_name)
select LAN,romappingstagingid as romappingfinalid,branch_code,
RO_CODE,main_ro_code,loan_end_date,ro_status,main_ro_status,
ro_transfer_status,main_ro_transfer_status,ro_name,main_ro_name
from viewromappingfinal
on duplicate key update LAN=values(LAN)

it’s working fine if i won’t use romappingfinalid(table primary key).

Suggest me how to use the INSERT ON UPDATE KEY using a table primary key but checking the condition on another primary key(LAN).

This is why it is always mentioned in cases count=count+1 and similar - in this case you always change the data. Really MySQL returns both rows matched and rows updated and it is rows matched you need to wants.

Generally Insert on dup key update is the method which was designed to make it most optimal though the question if it always is remains open

Also note the difference in this cases comes from amount of queries you have to run mainly - run benchmarks over network to see the difference better.

However, version 2 has a small issue.
If you try to update a record with the same data as in table, mysql ignore the query (affected rows => 0), and insert query is executed, even if the record exist in table.

For example, if I refresh twice the page the following query will be the same (if I refresh the page in the same SECOND), so affected rows = 0, thus INSERT INTO query is executed and an error will occur (duplicate entry cos’ of the primary index)

INSERT INTO forum_sessions (session_id, session_user_id, session_start, session_time, session_ip, session_page, session_logged_in, session_browser, session_mode) VALUES (’a321587721fdb3a86a49d761fb4b1a2e’, ”, ‘1216230846′, ‘1216230846′, ‘127.0.0.2′, ‘main’, ‘0′, ‘Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9) Gecko/2008052906 Firefox/3.0′, ‘html’)

Of course errors can be masked for users..

Can you provide me some tips for benchmarking? The generation time of the pages is almost the same in all 3 situations (I’ve tried all versions on my public site too.. ). localhost, I just refresh the page multiple times & read the smallest generation time.

Maybe I don’t feel any difference because the tables are small (4,870 rows, 1.4MB & 20-60 rows, < 100KB)

Cheers

Peter

The UPDATE rate is higher than the INSERT rate.

So.. peter, from your experience, what’s the FASTEST WAY from the following:

1) This method I’ve been using until found out about this article. First, we check if there is a record.. then.. see there.

mysql_query(”SELECT COUNT(*) as count FROM table WHERE user_id = ‘1′”);
$cnt = mysql_fetch_array();
$cnt = $cnt['count'];

if ($cnt == 0) {
mysql_query(”INSERT INTO table VALUES (’1′, ‘1′)”);
}
else {
mysql_query(”UPDATE table SET user_hits = user_hits+1 WHERE user_id = ‘1′”);
}

2) With inspiration from Peter Romianowski’s post. Haven’t tested, but I guess I’ve wrote the code ok.

mysql_query(”UPDATE table SET user_hits = user_hits+1 WHERE user_id = ‘1′”);
$affected_rows = mysql_affected_rows();

if ($affected_rows == 0) mysql_query(”INSERT INTO table VALUES (’1′, ‘1′)”);

3) Using INSERT .. ON duplicate KEY UPDATE ..

mysql_query(”INSERT INTO table VALUES (’1′, ‘1′) ON duplicate KEY UPDATE user_hits = user_hits+1 “);

So.. which one is recommended? I repeat.. my update rate is higher than the insert rate. I guess method 2. But I’m waiting other opinions.
Thanks.

The best place to ask the question is forums.mysqlperformanceblog.com ; and it is also best to provide exact statement and error message

INSERT INTO table SELECT * FROM table WHERE id=’2′ ON DUPLICATE KEY UPDATE id=LAST_INSERT_ID() + 1
I use LAST_INSERT_ID() coz im using auto increment.

Raven Riley