Comments on: Multiple row operations in MySQL / PHP

By: bonita

bonita — Tue, 20 Oct 2009 02:15:29 +0000

hi! what is the best way of inserting multiple entries into one table?

For example: 10 employees to be inserted in one click with their name, email, address.. etc…

Please help..thanks.

By: Shantanu Oak

Shantanu Oak — Wed, 15 Jul 2009 15:48:27 +0000

Thanks for the tip.
I have one question.

>> At 2560 rows inserted, it took the loop
What if I need to send tens of thousands of records?
Will the server have any issues?

By: thomas

thomas — Sat, 06 Jun 2009 19:57:56 +0000

hi i have 3 tables like a ,b c
where a contains employee detail , b contains details of item consumed by each employee and c contains different types of items, one user can have many items

so table b will have more than one entries of same id

i want to show in a pattern like below in webpage

employee 1– item1 — item2 — item 3 –item4
employee 2– item1 — ‘ ‘— item 3 –item4
employee — item1 — item2 — item 3 –item4– item5

i want to fetch the details in above pattern

i get result in query as belowe

employee1 – item1
employee1-item2
–

—

how could i fetch the details in the patern i want

By: Eran Galperin

Eran Galperin — Sat, 16 May 2009 14:45:38 +0000

Plain SQL of course, I make sure to properly filter and escape user input

By: Arik Fraimovich

Arik Fraimovich — Sat, 16 May 2009 11:07:41 +0000

@Eran & @George – thank you for the information. Going to rewrite some queries now

@Eran – “ON DUPLICATE…” in Zend Framework: so you just passing SQL in such cases or made your own wrapper?

By: George

George — Fri, 15 May 2009 07:26:46 +0000

@Arik Fraimovich: REPLACE and ON DUPLICATE KEY UPDATE are actually quite different. REPLACE actually deletes the row if it is inserted and then re-inserts it. This means that REPLACE will change any AUTOINCREMENT fields you have in your row, thus wasting IDs and even braking table relationships.

By: Eran Galperin

Eran Galperin — Thu, 14 May 2009 22:26:14 +0000

Yes, definitely the server architecture should be taken into consideration. In this case I would think that sending one query would have even more performance benefits, as remote connections are more costly due to higher latency.

You can always mix-and-match approaches, for example splitting the operations into 6 major query instead of one giant to preserve memory if needed. The performance boost over 2500~ queries would still be significant.

By: Andre

Andre — Thu, 14 May 2009 22:13:07 +0000

here’s a question. In a scenario where you have a seperate database server removed from the web application, would it be better to go for the use more memory approach and pass one large SQL statement?

I think it depends on the server architecture to determine when to use the multiple sql versus one sql solution.

By: Zippy

Zippy — Thu, 14 May 2009 20:49:22 +0000

I’m still relative new to maximizing my db queries, so I really enjoyed this article. Thanks for the response. I can think of a few instances where switching to the methods suggested would benefit.

By: Eran Galperin

Eran Galperin — Thu, 14 May 2009 20:21:03 +0000

@arik:
1. REPLACE is different – if the row exists already it deletes it before insertion. This is not useful for maintaining relationships between tables, and it can cause faster index fragmentation (as rows are constantly deleted and inserted).
2. Unfortunately no… for INSERT and UPDATE it uses just the basic syntax.

@andre:
I benchmarked it using PHP, pasted the results into an excel sheet and made a graph of it just tinkered a little with the colors. Too bad excel doesn’t have anti-aliasing…

@zippy:
The problem with using prepared statements for those scenarios is that the size of the query itself is dynamic. If your statement is prepared for one row at a time, it will still execute many times instead of one.
The overhead might be slightly smaller (as the statement is prepared and cached after one run), but it will still under perform badly against a single query.