Re: [Evolution-hackers] Camel mmap summary ideas, proposal for a meeting

[Philip Van Hoof <spam pvanhoof be>]

On Wed, 2006-07-12 at 17:19 -0400, Jeffrey Stedfast wrote:
> On Wed, 2006-07-12 at 20:05 +0200, Philip Van Hoof wrote:
 
> > -- Oh by the way. There's a huge amount of "Re: " strings in my memory
> > at this moment. We can probably improve that too ;-) --
> 
> Yea, I've been wanting to overlap strings in such a way as to allow "Re:
> my subject string" and "my subject string" to re-use the same memory.
> Not sure the complexity of this would be worth the (likely very small)
> memory savings tho

Just some idea dumping ...

Using GSList is going to introduce 8 bytes per node (the next and the
data pointer, right?) on x86. So not a good option imo.

If you store an offset-pointer you might get away with a few per such
chunk. For example ([offset] is a pointer to for example a GString str
pointer and [x] is a 7 bit encoded integer, like the pstring length):

[offset]ingtestRe: 

Then "Re: testing" would be stored like [7][3][0] and the component that
at some moment in time wants the string (for example the ETable) will
get a copy of the string from a method that using [offset] reconstructs
the string. That component will of course have to g_free() it itself.

> but at worst, using a single strtab would be nice even if "Re: ..." and
> "..." didn't overlap.


-- 
Philip Van Hoof, software developer at x-tend 
home: me at pvanhoof dot be 
gnome: pvanhoof at gnome dot org 
work: vanhoof at x-tend dot be 
http://www.pvanhoof.be - http://www.x-tend.be