On Mon, May 28, 2012 at 8:15 AM, Waldecir Faria <fighter2929@hotmail.com> wrote:
> Good morning, I am doing a study about buffer management to improve the
> performance of one program that does heavy I/O operations. After looking and
> reading from different softwares' source codes/texts one friend suggested me
> to take a look at the PostgreSQL code. I already took a look at the
> PostgreSQL buffer management modules ( freelist.c and cia ) but now I am a
> bit confused how the buffer read/write works, I tried to see how PostgreSQL
> does to get, for example, a char array from one buffer. Looking at rawpage.c
> I think that I found a good example using the following function calls
> sequence starting at function get_raw_page_internal():
>
> StrategyGetBuffer->BufferAlloc->ReadBuffer_Common
> ->ReadBufferExtended->BufferGetPage-> memcpy page to buf
BufferGetPage() doesn't copy anything; it just takes the buffer number
and returns a pointer to the address of that buffer in memory. More
generally, that whole chain of function calls has to do with how a
page ends up inside PostgreSQL's buffer cache, not with how anything
on the page is actually decoded. Each buffer contains zero or more
tuples; each tuple contains multiple attributes. So after you get the
buffer you have to iterate over the tuples and then decode each tuple
to get the values that you want.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
