TINKERPOP-1865 Run Gremlin.Net tests with GraphSON 3.0

[FlorianHockmann]

https://issues.apache.org/jira/browse/TINKERPOP-1865

The tests simply failed because of the traversal used in ScenarioData.GetEdges(). I didn't dig deep into why it failed, but it had something to do with the deserialization of the keys of the received dictionary. I also couldn't get the traversal used there to work with gremlin-python, so I simply replaced it by two simpler traversals. This increases the runtime a bit as the second traversal now has to be executed for every edge, but I think that it improves the readability a lot.

Unfortunately, I had to add 3 more tests to the list of ignored ones. One failed due to the missing T deserializer (TINKERPOP-1866) and two failed because the data received from the server didn't match the expected data:

• g_V_storeXaX_byXoutEXcreatedX_countX_out_out_storeXaX_byXinEXcreatedX_weight_sumX: We receive a g:Set of 8 values here, including 4 duplicates. This is deserialized to a HashSet by Gremlin.Net which means that the 8 values are reduced to 4 unique values, but the scenario expects a collection of 8 values.
• g_VX1X_hasXlabel_personX_mapXmapXint_ageXX_orderXlocalX_byXvalues_decrX_byXkeys_incrX: Here we receive a map with g:Int32 keys (at least when my understanding of the g:Map type is correct), but the scenario expects string keys.

To me, this looks like mistakes in the scenarios, but it would be good if someone could verify that. I tried to modify the scenarios accordingly, but then both gremlin-python and gremlin-javascript failed for the scenario with the g:Set containing duplicates. So either the deserialization of g:Set is flawed in those two GLVs or my understanding of sets is wrong.

[jorgebay]

I've rebased this branch to get the fixes for P.within() from #817.

Also, I've switched to use a lambda (now that we support it :) ) to obtain all the edges for the scenario data, that way we use the same traversal as Python. It would be nice to backport the same traversal for tp32 but it's not necessary for now.

About g_VX1X_hasXlabel_personX_mapXmapXint_ageXX_orderXlocalX_byXvalues_decrX_byXkeys_incrX, I've changed the lambda to return a Map<string, int>.

Regarding g_V_storeXaX_byXoutEXcreatedX_countX_out_out_storeXaX_byXinEXcreatedX_weight_sumX, its still ignored as the returned data doesn't match, it returns:

{ '@type': 'g:Set',
  '@value': 
   [ { '@type': 'g:Int64', '@value': 1 },
     { '@type': 'g:Int64', '@value': 1 },
     { '@type': 'g:Int64', '@value': 0 },
     { '@type': 'g:Int64', '@value': 0 },
     { '@type': 'g:Int64', '@value': 0 },
     { '@type': 'g:Int64', '@value': 2 },
     { '@type': 'g:Double', '@value': 1 },
     { '@type': 'g:Double', '@value': 1 }

Which is deserialized into a Set<object> containing only the different values (4 values in total), so it's a scenario design issue (not related to Gremlin.NET).

@FlorianHockmann I hope you don't mind that I've directly pushed to the dev branch, feel free to revert the commits if needed.

[FlorianHockmann]

Also, I've switched to use a lambda (now that we support it :) ) to obtain all the edges for the scenario data,

That lambda really makes the traversal much easier to read and understand :)

For g_V_storeXaX_byXoutEXcreatedX_countX_out_out_storeXaX_byXinEXcreatedX_weight_sumX,
I'd say that we should change the scenario to only expect the 4 unique values and then fix gremlin-python and gremlin-javascript as they seem to deserialize sets incorrectly at the moment. (Otherwise they had to fail for this test at the moment.) But I'm not sure if we want to do all that in this PR.

@FlorianHockmann I hope you don't mind that I've directly pushed to the dev branch, feel free to revert the commits if needed.

Certainly not, thanks for helping me with this PR!

[jorgebay]

I agree, we should fix g_V_storeXaX_byXoutEXcreatedX_countX_out_out_storeXaX_byXinEXcreatedX_weight_sumX outside of this pr.

Having the gherkin based tests running on GraphSON3 for Gremlin.NET is a huge improvement, VOTE +1

[FlorianHockmann]

Ok, I created TINKERPOP-1927 to fix the scenarios.

VOTE +1

[spmallette]

VOTE +1

[spmallette]

I think that I need this one merged to properly test out TINKERPOP-1866 as T serialization only works right on GraphSON 3.0