what happens in your php tests if you remove the sessionless call? does it work as expected?
On Thu, Oct 22, 2015 at 7:25 AM, Dylan Millikin <[email protected]> wrote: > >Where do you set the transaction model to MANUAL? I don't see that in > your requests > for your php tests or in the java test case? > > I don't. I actually never have, it was a bit of a misunderstanding we've > built up through the past few issues. I'm mostly responsible for it because > I assumed from reading the documentation (before checking the code) that > since we had to manually handle transactions in a session request, it > must've meant Transaction.onReadWrite() was set to MANUAL (clearly wrong). > > > I don't think it should matter - one socket should be able to handle > both sessionless > and in-session requests. > > Is it perhaps that the graph instance is created on a socket open so > somehow transaction handling is carried over between requests on that same > socket? (or something along those lines) > > Anyways, thanks for pointing out the WebSocketClient. I'll give that a go > and see if I can make a failing test, then I'll try and figure out what is > causing this. > > On Thu, Oct 22, 2015 at 12:03 PM, Stephen Mallette <[email protected]> > wrote: > > > I don't think it should matter - one socket should be able to handle both > > sessionless and in-session requests. If gremlin-driver hides too much > > stuff, you can do some tests with a strip down client I have for basic > > testing: > > > > > > > https://github.com/PommeVerte/incubator-tinkerpop/blob/beea4dc40ae5dae8e2c7f3c6094dbcc551478cc6/gremlin-driver/src/main/java/org/apache/tinkerpop/gremlin/driver/simple/WebSocketClient.java > > > > On Thu, Oct 22, 2015 at 5:53 AM, Dylan Millikin < > [email protected]> > > wrote: > > > > > Wow just reading over my email again helped me figure this one out. > > > Although the messages sent are the same, there's a distinct difference > > > between the two tests. > > > > > > The gremlin-driver test creates two clients that I'm guessing connect > to > > > gremlin-server via different sockets. > > > The gremlin-php test sends both sessionless and in session messages on > > the > > > same socket (same client). If I create two different clients the tests > > > pass. > > > > > > Is this something I can test with gremlin-driver? I'll go through the > api > > > to see if I can build a failing test. > > > > > > I don't know if this was by design but it's a bit of an issue as we > > usually > > > only create one client per thread and then mix sessionless and > in-session > > > requests with a single client. (The single client design is more or > less > > > standard in PHP). > > > > > > Anyways if it's by design and there's a good reason I guess I could > > modify > > > the driver to abstract multi client handling for both cases. > > > > > > On Thu, Oct 22, 2015 at 11:37 AM, Dylan Millikin < > > [email protected] > > > > > > > wrote: > > > > > > > Hey guys, > > > > > > > > I've written the following test that passes: > > > > > > > > > > > > > > https://github.com/PommeVerte/incubator-tinkerpop/blob/transaction-test/gremlin-server/src/test/java/org/apache/tinkerpop/gremlin/server/GremlinDriverIntegrateTest.java#L591-L615 > > > > > > > > Unfortunately the exact same test from my gremlin-php driver fails. > The > > > > second isOpen() test returns true. > > > > > > > > The gremlin-php messages sent are the following : > > > > > > > > > > > > > > > > > > application/json{"requestId":"77a459e3-32d9-4640-89cb-c2e28c67caf5","processor":"session","op":"eval","args":{"gremlin":"graphT.tx().open()","session":"6e1c97b3-3f69-4b08-aa7c-c137a47cf04b"}} > > > > > > > > > > > > > > > > > > application/json{"requestId":"576aec62-442a-4a21-bcac-cbac36a388e9","processor":"session","op":"eval","args":{"gremlin":"t.addV(\"name\",\"stephen\").next()","session":"6e1c97b3-3f69-4b08-aa7c-c137a47cf04b"}} > > > > > > > > > > > > > > > > > > application/json{"requestId":"e1dfae26-f115-43c3-9b44-cf2879375037","processor":"session","op":"eval","args":{"gremlin":"graphT.tx().commit()","session":"6e1c97b3-3f69-4b08-aa7c-c137a47cf04b"}} > > > > > > > > > > > > > > > > > > application/json{"requestId":"359e2f72-f6fd-4a91-a0bd-0dd8002465fb","processor":"session","op":"eval","args":{"gremlin":"graphT.tx().isOpen()","session":"6e1c97b3-3f69-4b08-aa7c-c137a47cf04b"}} > > > > ----RETURNS FALSE---- > > > > > > > > > > > > > > > > > > application/json{"requestId":"bf1a7d47-5819-4171-87a0-3be5adfe70e7","processor":"","op":"eval","args":{"gremlin":"graphT.traversal().V()"}} > > > > > > > > > > > > > > > > > > application/json{"requestId":"53f53620-4337-4ccf-8a85-6ab25d0f78e3","processor":"session","op":"eval","args":{"gremlin":"graphT.tx().isOpen()","session":"6e1c97b3-3f69-4b08-aa7c-c137a47cf04b"}} > > > > ----RETURNS TRUE---- > > > > > > > > And as far as I can tell these are the same as the ones sent from > > > > gremlin-driver test but the results differ. The only thing that > changes > > > > might be the serializer which should have no effect? > > > > > > > > If anyone has any idea as to where there could be an issue, feel free > > to > > > > shout out with ideas (as vague as they may be) I'm investigating this > > one > > > > in the dark so I'd be more that grateful for a little help focussing > my > > > > search. > > > > > > > > PS: results are the same with the ThreadLocal transaction consumers > > > patch. > > > > Although this shouldn't have an effect anyways I thought I would > point > > it > > > > out. > > > > > > > > Cheers, > > > > Dylan. > > > > > > > > > >
