<div dir="ltr">Along these lines, I've been trying to figure out a reliable way to tell if a call to dispatch is necessary. Suppose I have a handler that can be called alone or could be nested inside another handler. My handler is going to do something async so I need to make sure it's been dispatched. Since it's about to do something very fast and async there's no need to move that work to another thread, I just want to run it locally so **if** it needs dispatching I'd call "dispatch(SameThreadExecutor.INSTANCE, ::doSomeStuff)". The handler wrapping this might do the same (dispatch, but right back to the IO thread). Thus, when my handler gets called it **might** have already been dispatched but it's not sure.<div><br></div><div>Calling "Exchange.isInIoThread()" isn't reliable as the wrapping handler may have dispatched via SameThreadExecutor so being on the IO thread does not mean a dispatch in needed. OTOH, as I understand it (not sure, please correct me if I'm wrong) "isDispatched" isn't generally reliable as that gets unset as soon as you've left the IO thread so if the wrapper dispatched using any executor besides SameThreadExecutor my handler doesn't need to dispatch but isDispatched will be false.</div><div><br></div><div>So, I think what is reliable is that you need to dispatch if and only if "isInIoThread() && !isDispatched()". Is that correct?</div></div><br><div class="gmail_quote"><div dir="ltr">On Thu, Jan 26, 2017 at 4:10 PM Stuart Douglas <<a href="mailto:sdouglas@redhat.com">sdouglas@redhat.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">In general there is never any real need to dispatch back to an IO<br class="gmail_msg">
thread, although if you perform any non blocking IO ownership of the<br class="gmail_msg">
exchange can end up back in the IO thread when the operation is<br class="gmail_msg">
complete.<br class="gmail_msg">
<br class="gmail_msg">
Stuart<br class="gmail_msg">
<br class="gmail_msg">
On Thu, Jan 26, 2017 at 12:53 AM, Bill O'Neil <<a href="mailto:bill@dartalley.com" class="gmail_msg" target="_blank">bill@dartalley.com</a>> wrote:<br class="gmail_msg">
> Hey Stuart, your last response reminded me of of a question I had hopefully<br class="gmail_msg">
> its not unrelated. I was curious how and when you should dispatch back to<br class="gmail_msg">
> the IO thread. For example lets say we have a simple handler that does a SQL<br class="gmail_msg">
> query that is composed with the AccessLogHandler.<br class="gmail_msg">
><br class="gmail_msg">
> AccessLogHandler -> BlockingHandler -> CustomSqlHandler In this approach I<br class="gmail_msg">
> believe the final access log statement will be logged from the worker thread<br class="gmail_msg">
> not the initial IO thread.<br class="gmail_msg">
><br class="gmail_msg">
> Is it possible / recommended to kick back to an IO thread once blocking is<br class="gmail_msg">
> complete?<br class="gmail_msg">
><br class="gmail_msg">
> AccessLogHandler -> BlockingHandler -> CustomSqlHandler -><br class="gmail_msg">
> DispatchBackToIOThread? This way the final logging of AccessLogHandler is<br class="gmail_msg">
> handled in the IO thread.<br class="gmail_msg">
><br class="gmail_msg">
> I'm not very familiar with the best ways to mix the nonblocking and blocking<br class="gmail_msg">
> handlers.<br class="gmail_msg">
><br class="gmail_msg">
><br class="gmail_msg">
><br class="gmail_msg">
> On Wed, Jan 25, 2017 at 12:24 AM, Stuart Douglas <<a href="mailto:sdouglas@redhat.com" class="gmail_msg" target="_blank">sdouglas@redhat.com</a>><br class="gmail_msg">
> wrote:<br class="gmail_msg">
>><br class="gmail_msg">
>> The handler is executed<br class="gmail_msg">
>> io.undertow.server.Connectors#executeRootHandler, which means that<br class="gmail_msg">
>> when the call stack returns the exchange will be ended.<br class="gmail_msg">
>><br class="gmail_msg">
>> Conceptually this is similar to how dispatching to a thread pool<br class="gmail_msg">
>> works, when you call dispatch(HttpHandler) the exchange will be ended<br class="gmail_msg">
>> when the call stack returns, even though you are no longer on the IO<br class="gmail_msg">
>> thread (unless you call dispatch again).<br class="gmail_msg">
>><br class="gmail_msg">
>> Stuart<br class="gmail_msg">
>><br class="gmail_msg">
>><br class="gmail_msg">
>> On Wed, Jan 25, 2017 at 3:57 PM, Oliver Dain <<a href="mailto:oliver@analyticspot.com" class="gmail_msg" target="_blank">oliver@analyticspot.com</a>><br class="gmail_msg">
>> wrote:<br class="gmail_msg">
>> > I have some code for handling file uploads. It uses FormDataProcessor<br class="gmail_msg">
>> > and<br class="gmail_msg">
>> > tries to do everything asynchronously. I call "FormDataParser.parse"<br class="gmail_msg">
>> > passing<br class="gmail_msg">
>> > in another handler. I'll call that handler the OnFormDataAvailable. The<br class="gmail_msg">
>> > OnFormDataAvailable handler checks to see if it's on the IO thread. If<br class="gmail_msg">
>> > it<br class="gmail_msg">
>> > is, it calls dispatch. Either way, once we're sure we're dispatched that<br class="gmail_msg">
>> > handler calls yet another handler.<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > What I've seen is that my OnFormDataAvailable handler (the one called by<br class="gmail_msg">
>> > parse()) is not on the IO thread so it doesn't need to call dispatch.<br class="gmail_msg">
>> > And<br class="gmail_msg">
>> > yet, the exchange gets ended before the handler it calls is even close<br class="gmail_msg">
>> > to<br class="gmail_msg">
>> > complete.<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > I've found a fix, but I don't understand why it's necessary.<br class="gmail_msg">
>> > Specifically,<br class="gmail_msg">
>> > if OnFormDataAvailable is not on the IO thread when its invoked it calls<br class="gmail_msg">
>> > the<br class="gmail_msg">
>> > 0-argument version of "HttpServerExchange.dispatch()" (which you've told<br class="gmail_msg">
>> > me<br class="gmail_msg">
>> > in another conversation shouldn't ever be necessary). If I do that,<br class="gmail_msg">
>> > everything is fine.<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > For completeness, here's the complete code:<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > public class FormDataParsingHandler {<br class="gmail_msg">
>> > private static final Logger log =<br class="gmail_msg">
>> > LoggerFactory.getLogger(FormDataParsingHandler.class);<br class="gmail_msg">
>> > private static final FormParserFactory formParserFactory =<br class="gmail_msg">
>> > FormParserFactory.builder().build();<br class="gmail_msg">
>> > public static final AttachmentKey<FormData> FORM_DATA_ATTACHMENT_KEY =<br class="gmail_msg">
>> > AttachmentKey.create(FormData.class);<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > /**<br class="gmail_msg">
>> > * The only public method - this is what gets exposed as the<br class="gmail_msg">
>> > HttpHandler.<br class="gmail_msg">
>> > */<br class="gmail_msg">
>> > public CompletableFuture<FormData> parseForm(HttpServerExchange<br class="gmail_msg">
>> > exchange)<br class="gmail_msg">
>> > {<br class="gmail_msg">
>> > <a href="http://log.info" rel="noreferrer" class="gmail_msg" target="_blank">log.info</a>("audio file upload request received.");<br class="gmail_msg">
>> > FormDataParser parser = formParserFactory.createParser(exchange);<br class="gmail_msg">
>> > if (parser == null) {<br class="gmail_msg">
>> > log.warn("No parser found that can handle this content type.<br class="gmail_msg">
>> > Headers<br class="gmail_msg">
>> > were: {}", exchange.getRequestHeaders());<br class="gmail_msg">
>> > throw new UserVisibleException("No parser for the given content<br class="gmail_msg">
>> > type.", ResponseCodes.BAD_REQUEST);<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > CompletableFuture<FormData> toComplete = new CompletableFuture<>();<br class="gmail_msg">
>> > try {<br class="gmail_msg">
>> > parser.parse(new OnFormDataAvailable(toComplete));<br class="gmail_msg">
>> > } catch (Exception e) {<br class="gmail_msg">
>> > log.error("Error parsing form data:", e);<br class="gmail_msg">
>> > throw wrapAsUnchecked(e);<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > exchange.addExchangeCompleteListener((ex, nextListener) -> {<br class="gmail_msg">
>> > // Must close the parser so it can free any temporary files that<br class="gmail_msg">
>> > were<br class="gmail_msg">
>> > created.<br class="gmail_msg">
>> > try {<br class="gmail_msg">
>> > parser.close();<br class="gmail_msg">
>> > } catch (IOException e) {<br class="gmail_msg">
>> > log.error("Error closing the FormDataParser. Request was handled<br class="gmail_msg">
>> > successfully but temporary files may not "<br class="gmail_msg">
>> > + "have been cleaned up.", e);<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> > nextListener.proceed();<br class="gmail_msg">
>> > });<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > return toComplete;<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > // The FormDataParser calls an HttpHandler when it's complete so we<br class="gmail_msg">
>> > add a<br class="gmail_msg">
>> > silly handler here that does nothing but<br class="gmail_msg">
>> > // complete this method's future when the form data is available.<br class="gmail_msg">
>> > private static class OnFormDataAvailable implements HttpHandler {<br class="gmail_msg">
>> > private final CompletableFuture<FormData> toComplete;<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > private OnFormDataAvailable(CompletableFuture<FormData> toComplete)<br class="gmail_msg">
>> > {<br class="gmail_msg">
>> > this.toComplete = toComplete;<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > @Override<br class="gmail_msg">
>> > public void handleRequest(HttpServerExchange exchange) throws<br class="gmail_msg">
>> > Exception<br class="gmail_msg">
>> > {<br class="gmail_msg">
>> > // Before we complete the future we have to re-dispatch or we'll<br class="gmail_msg">
>> > fall<br class="gmail_msg">
>> > off the end of this method and Undertow<br class="gmail_msg">
>> > // will complete the exchange on our behalf.<br class="gmail_msg">
>> > FormData data = exchange.getAttachment(FormDataParser.FORM_DATA);<br class="gmail_msg">
>> > if (exchange.isInIoThread()) {<br class="gmail_msg">
>> > log.debug("Was on the IO thread. Re-dispatching.");<br class="gmail_msg">
>> > exchange.dispatch(SameThreadExecutor.INSTANCE, () -><br class="gmail_msg">
>> > afterDistpach(data));<br class="gmail_msg">
>> > } else {<br class="gmail_msg">
>> > // THIS THE MYSTERY LINE. WHY IS THIS NEEDED?<br class="gmail_msg">
>> > exchange.dispatch();<br class="gmail_msg">
>> > afterDistpach(data);<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > private void afterDistpach(FormData data) {<br class="gmail_msg">
>> > if (data == null) {<br class="gmail_msg">
>> > toComplete.completeExceptionally(<br class="gmail_msg">
>> > new UserVisibleException("Parsing of data failed.",<br class="gmail_msg">
>> > ResponseCodes.BAD_REQUEST));<br class="gmail_msg">
>> > } else {<br class="gmail_msg">
>> > toComplete.complete(data);<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> > }<br class="gmail_msg">
>> ><br class="gmail_msg">
>> > --<br class="gmail_msg">
>> > CTO, Analytic Spot<br class="gmail_msg">
>> > 44 West Broadway #222<br class="gmail_msg">
>> > Eugene, OR 97401<br class="gmail_msg">
>> > <a href="http://analyticspot.com" rel="noreferrer" class="gmail_msg" target="_blank">analyticspot.com</a> • <a href="tel:(425)%20296-6556" value="+14252966556" class="gmail_msg" target="_blank">425-296-6556</a><br class="gmail_msg">
>> > <a href="http://www.linkedin.com/in/oliverdain" rel="noreferrer" class="gmail_msg" target="_blank">www.linkedin.com/in/oliverdain</a><br class="gmail_msg">
>> ><br class="gmail_msg">
>> > _______________________________________________<br class="gmail_msg">
>> > undertow-dev mailing list<br class="gmail_msg">
>> > <a href="mailto:undertow-dev@lists.jboss.org" class="gmail_msg" target="_blank">undertow-dev@lists.jboss.org</a><br class="gmail_msg">
>> > <a href="https://lists.jboss.org/mailman/listinfo/undertow-dev" rel="noreferrer" class="gmail_msg" target="_blank">https://lists.jboss.org/mailman/listinfo/undertow-dev</a><br class="gmail_msg">
>><br class="gmail_msg">
>> _______________________________________________<br class="gmail_msg">
>> undertow-dev mailing list<br class="gmail_msg">
>> <a href="mailto:undertow-dev@lists.jboss.org" class="gmail_msg" target="_blank">undertow-dev@lists.jboss.org</a><br class="gmail_msg">
>> <a href="https://lists.jboss.org/mailman/listinfo/undertow-dev" rel="noreferrer" class="gmail_msg" target="_blank">https://lists.jboss.org/mailman/listinfo/undertow-dev</a><br class="gmail_msg">
><br class="gmail_msg">
><br class="gmail_msg">
</blockquote></div><div dir="ltr">-- <br></div><div data-smartmail="gmail_signature"><div dir="ltr"><div style="font-size:small"><div style="font-size:13px;line-height:19.5px">CTO, Analytic Spot</div><div style="font-size:13px;line-height:19.5px">44 West Broadway #222</div><div style="font-size:13px;line-height:19.5px">Eugene, OR 97401<br></div><div style="font-size:13px;line-height:19.5px"><a href="http://analyticspot.com/" style="z-index: 0;">analyticspot.com</a> <span style="color:rgb(127,127,127);font-family:'helvetica neue';font-size:11px;line-height:normal">• </span>425-296-6556</div></div><div style="font-size:small"><span style="line-height:19.5px"><a href="http://www.linkedin.com/in/oliverdain" style="z-index: 0;">www.linkedin.com/in/oliverdain</a></span></div></div></div>