+1 I and a few of my former Lucidworks colleagues know the pain of the synchronous client very well.
Thanks for the first step toward an improved design Dat and David Smiley! Marcus On Sun, Aug 2, 2020 at 07:47 Atri Sharma <a...@apache.org> wrote: > I am +1 to this approach. Some thoughts inline. > > How would query timeout be respected in this approach? > >> > > The default approach might be configured to throw >> UnsupportedOperationException, or perhaps might simply use an Executor to >> get it done in an obvious way (assuming we can get ahold of an Executor >> somewhere?). >> > > Would that mean that we use an Executor to execute a single thread? > > > >> > CompletableFuture, and which merely takes the SolrRequest parameter and > nothing else. Alternatively the client could supply a CompletableFuture > parameter that Solr will call complete() on etc. but that seems a bit less > natural to the notion of a method that returns it's results, albeit with a > wrapper. > > I would think that we allow users to specify their callback. One of the > advantages of AsyncListener is that it a custom implementation can allow > users to handle the behaviour of timeout and other events. We should retain > that behaviour. > -- > Regards, > > Atri > Apache Concerted > -- Marcus Eagan
