Atomic ets

[Rick Pettit]

On Tue, Dec 13, 2005 at 06:45:53PM +0000, Sean Hinde wrote:
> 
> On 13 Dec 2005, at 17:35, Rick Pettit wrote:
> 
> >On Tue, Dec 13, 2005 at 05:30:26PM +0100, Tony Rogvall wrote:
> >>
> >>On 13 dec 2005, at 15.45, Thomas Lindgren wrote:
> >>
> >>>
> >>>Now that we are getting a multithreaded runtime
> >>>system, is this a good time to bring up ets? I am
> >>>having some issues here.
> >>
> >>For what it's worth I wrote a module called atomic.erl some time ago.
> >>It can be regarded as
> >>lightweight but it uses ets and monitor heavily to avoid
> >>synchronizing with a server process.
> >
> >Why not use global:trans/[234]?
> >
> The requirement was to avoid serialising all ets read/updates through  
> a gen_server. Global transactions require 2 gen_server calls (one to  
> lock/one to unlock) to save one call.
> 
> Tony's solution relies solely on ets access from the local process to  
> achieve locking. No other processes involved (except one to own the  
> public ets table).

Got it--sorry about that, should have paid closer attention.

-Rick