'after' in gen_server

Rick Pettit <>
Thu Mar 23 22:35:21 CET 2006


On Thu, Mar 23, 2006 at 04:26:22PM -0500,  wrote:
> I want to do it from all callbacks.  Basically I want to make it so if 
> the server is idle for X amount of time, then that means something is 
> wrong and I need to do something to figure out what is wrong.

Then have your callbacks set a gen_server "idle timeout" by way of the
gen_server timeout mechanism, and service timeouts in handle_info/2.

It is as simple as passing a timeout value in the return value from your
callbacks--Module:init/1, Module:handle_call/3, Module:handle_cast/2, etc.

-Rick

> On Mar 23, 2006, at 11:04 AM, Rick Pettit wrote:
> 
> >On Thu, Mar 23, 2006 at 10:42:04AM -0500,  wrote:
> >>I am reworking a bit of code into gen_server pattern.  In this code I
> >>have a typical loop construct.  In this I want to perform an action if
> >>no messages have been receive in a certain amount of time, to do this 
> >>I
> >>simply have after sometimeout ->.  Is there any equivalence of this in
> >>gen_server?
> >
> >If I understand you correctly, all you need to do is set a timeout when
> >returning from one of the gen_server callback functions.
> >
> >For example, instead of returning {ok,State} from Module:init/1, return
> >{ok,State,TimeoutMs}. You can do this for all the gen_server callback
> >routines (at least handle_call/handle_cast, etc).
> >
> >This effectively sets a timer which will expire if no message is 
> >received
> >by the gen_server before TimeoutMs has elapsed. If/when the timer does
> >expire, you receive a 'timeout' message in Module:handle_info/2.
> >
> >-Rick
> >
> >P.S. Since gen_server abstracts the server receive loop, I (as a 
> >general rule
> >     of thumb) *never* call receive in a gen_server callback module.
> >
> >>It has been suggested that I use a timer and record when the last
> >>message has come in and when the timer signals check against that
> >>message.  This seems a poor solution.  Another suggestions was to use 
> >>a
> >>gen_fsm with a timeout, but this seem a bit much just to get an 
> >>'after'
> >>mechanism. Any other suggestions?  Perhaps my disregard of gen_fsm is 
> >>a
> >>bit hasty?
> >
> 



More information about the erlang-questions mailing list