On Tue, 28 Jun 2005, Rasmus Christian Kaae wrote:
>If you are running Windows you may use a similar memory-mapped file (see the
>Win32API for details).
>
>
>Does anyone know which is fastest -- Using a memory-mapped file (or
>/dev/shm) in comparison with sqlite's internal memory mapped tree?
Haven't tested it, so take with a pinch of salt.
</disclaimer>
My guess is that the SQLite internel memory only representation will be
quicker, as there is less data transformation being done. The SQLite disk
image must be written in such a way that is n-bit and bi-endian clean, so
that the file can be read on any platform. This doesn't come for free.
The in-memory database, however, just uses native pointers and datatypes
of the processor, so is much quicker to manipulate and traverse.
If you're not keeping the data, the in-memory is almost certainly the best
way to go.
If in doubt, test it.
As to the original question, I can't say I've ever used SQLite that way,
but only because I want to keep my data as is.
>
>--
>Best regards / Med venlig hilsen
>Rasmus Christian Kaae - [EMAIL PROTECTED] or [EMAIL PROTECTED]
>Web: www.kalchas.dk www.eurojumping.com www.hestebasen.com
>Office: E4-119, Aalborg University.
>Mobile: +45 28 72 04 13
>
>
>
>Quoting Ben Clewett <[EMAIL PROTECTED]>:
>
>> If you are using Linux, you can use the /dev/shm. This is a memory
>> resident file system. I use this and find it about 200 times faster
>> than writing to disk. The only problem is that this is erased when the
>> server reboots. I hope this is of some use :)
>
--
/"\
\ / ASCII RIBBON CAMPAIGN - AGAINST HTML MAIL
X - AGAINST MS ATTACHMENTS
/ \
