some questions in postgresql developping

Started by 黄晓骋about 16 years ago2 messages
#1黄晓骋
huangxclife@gmail.com

Hello,

I’m a student from Nankai University in China. Now I and my team do a
project which aims to integrate XML to Postgresql. What I do is to complete
the function of XML Update.

Now I’m researching in concurrency control. I have read the code about the
concurrency control for a long time and I’m confident that I know it much.
But I am puzzled that why we need to lock tuple. I think locking transaction
is sufficient. I don’t think the tuple lock is good at improving executing
rate or anything.

I am wishing for your reply.

Best Regards,

--Huang Xiaocheng

--Database & Information System Lab, Nankai University, China

#2Robert Haas
robertmhaas@gmail.com
In reply to: 黄晓骋 (#1)
Re: some questions in postgresql developping

2009/12/7 黄晓骋 <huangxclife@gmail.com>:

I’m a student from Nankai University in China. Now I and my team do a
project which aims to integrate XML to Postgresql. What I do is to complete
the function of XML Update.

Now I’m researching in concurrency control. I have read the code about the
concurrency control for a long time and I’m confident that I know it much.
But I am puzzled that why we need to lock tuple. I think locking transaction
is sufficient. I don’t think the tuple lock is good at improving executing
rate or anything.

The purpose of locking is to preserve correct semantics in the face of
concurrent activity, not to improve execution speed. The reasons why
tuple locking is necessary should be covered in any introductory
database textbook.

...Robert