You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@tinkerpop.apache.org by "stephen mallette (JIRA)" <ji...@apache.org> on 2016/10/08 11:55:20 UTC
[jira] [Comment Edited] (TINKERPOP-1490) Provider a Future based
Traversal.async(Function) terminal step
[ https://issues.apache.org/jira/browse/TINKERPOP-1490?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15557840#comment-15557840 ]
stephen mallette edited comment on TINKERPOP-1490 at 10/8/16 11:54 AM:
-----------------------------------------------------------------------
{{future}} with an extension of {{FutureTask}} seems good to me, though i don't think we should ignore {{CompletableFuture}}. {{Traversal}} should have:
{code}
void promise(CompletableFuture future) // perhaps we just start a thread to iterate the traversal (i.e no thread pool)
void promise(CompletableFuture future, ExecutorService executor) // in this case we provide a thread pool to execute in
{code}
{{CompletableFuture}} provides a lot more flexibility for handling exceptions, starting async tasks in different executors, and clarifies situations with lots of callback chaining. It makes it trivial to do stuff like:
{code}
promiseOutNames = new CompletableFuture()
promiseInNames = new CompletableFuture().thenAccept(f -> transformResult(f))
promiseBothNames = promiseInNames.thenCombine(promiseOutNames, (in,out) -> combineInOut(in, out)).
thenRunAsync(f -> notifySomethingOfBoth(f), gremlinServerExecutorService)
g.V(1).out().values("name").promise(promise1)
g.V(1).in().values("name").promise(promise2)
{code}
So the above would traverse in/out vertices in parallel across two separate threads, combine the results and then make an async call to notify some service about the combined results.
was (Author: spmallette):
{{future}} with an extension of {{FutureTask}} seems good to me, though i don't think we should ignore {{CompletableFuture}}. {{Traversal}} should have:
{code}
void promise(CompletableFuture future) // in this case, perhaps we just start a thread to iterate the traversal in gremlinServerExecutorService (i.e no thread pool)
void promise(CompletableFuture future, ExecutorService executor) // in this case we provide a thread pool to execute in
{code}
{{CompletableFuture}} provides a lot more flexibility for handling exceptions, starting async tasks in different executors, and clarifies situations with lots of callback chaining. It makes it trivial to do stuff like:
{code}
promiseOutNames = new CompletableFuture()
promiseInNames = new CompletableFuture().thenAccept(f -> transformResult(f))
promiseBothNames = promiseInNames.thenCombine(promiseOutNames, (in,out) -> combineInOut(in, out)).
thenRunAsync(f -> notifySomethingOfBoth(f), gremlinServerExecutorService)
g.V(1).out().values("name").promise(promise1)
g.V(1).in().values("name").promise(promise2)
{code}
So the above would traverse in/out vertices in parallel across two separate threads, combine the results and then make an async call to notify some service about the combined results.
> Provider a Future based Traversal.async(Function<Traversal,V>) terminal step
> ----------------------------------------------------------------------------
>
> Key: TINKERPOP-1490
> URL: https://issues.apache.org/jira/browse/TINKERPOP-1490
> Project: TinkerPop
> Issue Type: Improvement
> Components: language-variant, process
> Affects Versions: 3.2.2
> Reporter: Marko A. Rodriguez
>
> [~mbroecheler] had the idea of adding a {{Traversal.async()}} method. This is important for not only avoiding thread locking on a query in Gremlin, but also, it will allow single threaded language variants like Gremlin-JavaScript to use callbacks for processing query results.
> {code}
> Future<List<String>> result = g.V().out().values("name").async(Traversal::toList)
> {code}
> {code}
> Future<List<String>> result = g.V().out().name.async{it.toList()}
> {code}
> {code}
> g.V().out().values('name').async((err,names) => {
> // I don't know JavaScript, but ...
> return list(names);
> })
> {code}
> ...
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)