SOAP - Problem with separate threads

[Dave]

Hi Andre,

Thanks a lot for the detailed response. I'd actually thought of using  
a queue to hold the requests, it seems like a better approach anyway,  
however I didn't do it like that because I couldn't figure out how to  
return the results from the request.

In my SOAP library, all requests are handled via an "ExecuteCommand"  
function. This function builds the SOAP request from parameters  
supplied, sends the HTML request, decodes the response and passes  
back the data received to the calling function. The "ExecuteCommand"   
calls an internal function called "SendRPCRequest" to handle all HTML  
requests.

Given the following pseudo code:

function ExecuteCommand theParameters

put BuildSOAPXML(theParameters) into myRequestData..
put AddRequestToQueue(myRequestData) into myTransactionID
repeat while CheckRequestComplete(myTransactionID,myResponse) = false
wait .25 seconds
end repeat

(Decode and Pass back SOAP response to caller)

The thread that handles the SOAP requests would work like this:

put GetRequestFromQueue() into myReqest
put SendRequest() into myResponse
get AddResponseToQueue(myResponse)

Is this how you saw it working? How do you suggest I pass back the  
Response to the calling function?

Thanks again.
All the Best
Dave

On 23 Oct 2007, at 17:42, Andre Garzia wrote:

> Dave,
>
> never, ever, use a wait command while network routines may be working.
> The wait call will wreck them. If the problem is entering a race
> condition, like, when one call is being placed, the other can't work.
> This happens because libURL will queue requests to the same server, so
> if you call the same server two times in a row without allowing it to
> deal with the first request, you receive an "error previous request
> not completed".
>
> The idea of using a flag local variable is the best option, but do
> that without the wait call. Work like this, if the flag is true,
> postpone your operation to same later time. like
>
> if isInUseFlag then
>   -- call in use, postpone..
>   send "doTheRequestAgain" to me in 2 secs
> else
>   -- do the request...
>   put false into isInUseFlasg
> end if
>
> The hardest part is making a function that can re-request something
> because if you're using send in time, you can't gather what were the
> original parameters for the call by examining the pendingMessages.
>
> If I were in your shoes, I'd work like this:
>
> 1) create a transaction queue. This would be an array where each web
> request would be a member. Each array member stores all the info
> needed to do a SOAP request.
>
> 2) I'd had a loop that would pick this array and work on the first
> member, after dealing with this member, I'd remove it from the array.
>
> 3) Pooling the server then is just a matter of a simple send in time
> that adds an element to the end of the array.
>
> 4) any function that would force a SOAP request would simply add an
> element to the end of this array.
>
> This way all your requests will work in order without the need of a
> locking mechanism or a wait call. By having a 'callback' stored along
> the data for each array member, you make your code as multi-taksing as
> it can be in Rev.
>
> Since all your calls are on the same place (the array pooling
> function) you will not enter race conditions and your debugging will
> be easier.
>
> People with CS background will recognize this as nothing more than a
> datastructure known as First-in First-Out queue. By making functions
> like:
>
> addToQueue pQueue, pElement -- add Element to Queue at last poisition.
> takeElemFromQueue pQueue -- returns the first Element from queue  
> and removes it.
>
> You have created a generic queue function library that can be worked
> in other projects.
>
> Cheers
> andre
> _______________________________________________
> use-revolution mailing list
> use-revolution at lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your  
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-revolution