counting the number of edges between two nodes and storing that as a property in a new (summary) edge

[Olav Laudy]

Hi,

I have nodes with multiple edges between them. I can count them, however, since count is a reducing barrier, I'm unable to use the count in the creation of a new edge between the two nodes.

g.addV('object1')

g.addV('object2')

g.addV('object3')

g.V().hasLabel('object1').as('A')

  .V().hasLabel('object2').addE('obs1').from('A')

g.V().hasLabel('object1').as('A')

  .V().hasLabel('object2').addE('obs2').from('A')

g.V().hasLabel('object2').as('A')

  .V().hasLabel('object3').addE('obs1').from('A')

I would like to create two edges:

g.addE('summary').from('object1').to('object2').property('count',2)

g.addE('summary').from('object2').to('object3').property('count',3)

I can count:

g.V().hasLabel('object1').as('A').outE().as('B').count()

but after issuing the count, I'm unable to get hold of 'A' and 'B' due to the fact that count is a reducing barrier. 

Any suggestion is highly welcome!

[Kelvin Lawrence]

Others I am sure can come up with better ways but does something like this work for you?

gremlin> g.V(3).as('a').project('edges').by(outE().count()).addV('newnode').property('count',select('edges'))
==>v[54922]
gremlin> g.V(54922).valueMap()
==>[count:[61]]

[Olav Laudy]

Hi Kelvin,

Thank you so much for your suggestion. Your query creates and vertex, while I'm looking to create an edge. When I try to transform your query into an edge query, I run into the same issue as before: after the count is done, I can't 'reach' the vertex anymore.

Any ideas?

[Kelvin Lawrence]

Something like this then perhaps

Enter code  g.V(3).as('a').V('4').as('b').addE('summary').from('a').to('b').property('count',select('a').by(outE().count()))
e[54930][3-summary->4]

gremlin> g.E(54930).valueMap()

==>[count:67]here...

[Olav Laudy]

Hi,

Unfortunately, I don't get this to work: in your example, you id the nodes (g.V(3)); for me this needs to work with all the nodes I iterate over. If I do a g.V() twice, I get a Cartesian join which is huge in my real case and the counts do not relate correctly to the pair involved.

Maybe my example more clear:

g.addV('object').property('name',1)

g.addV('object').property('name',2)

g.addV('object').property('name',3)

g.V().has('object','name',1).as('A').V().has('object','name',2).addE('obs').from('A')

g.V().has('object','name',1).as('A').V().has('object','name',2).addE('obs').from('A')

g.V().has('object','name',2).as('A').V().has('object','name',3).addE('obs').from('A')

The outcome should be:

Link from name=1 to name=2, count=2

Link from name=2 to name=3, count=1

[Robert Dale]

g.V().

project('a','b').by().by(__.out('obs').groupCount()).as('group').

select('b').unfold().as('out').select(keys).

addE('summary').

from(select('group').select('a')).

property('summary', select('out').select(values))

Robert Dale

--
You received this message because you are subscribed to the Google Groups "Gremlin-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/1364f7bf-6c07-4e91-b0d3-dd9237485e07%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Olav Laudy]

Very impressive!!! Where do I learn that? How do I feel that logic?

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/1364f7bf-6c07-4e91-b0d3-dd9237485e07%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "Gremlin-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/gremlin-users/QEn7URER8aU/unsubscribe.
To unsubscribe from this group and all its topics, send an email to gremlin-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/CABed_4pGvDLvL%2B%2B2Yt%2BZQwafTO1zFBAO5_yS9f%2BCnysob8BVng%40mail.gmail.com.

[Kelvin Lawrence]

Hi Olav, glad to see Robert got you going. I missed your last post and was not understanding properly what you wanted to achieve. As to how you get to "feel" Gremlin - I think we would all (at least most of us anyway!) agree that it takes time and lots of experiments. before it starts to become intuitive and even then there is often a better way that someone else will come up with!

Hang in there - I know I learn something new about Gremlin just about every single day!

Kelvin

[Robert Dale]

Read everything Mr. Kuppitz has ever written.  ;-)

This is one of those where you have to put everything up front because you can't get reference it later or it's not optimal to.

For example, g.V().as('a').groupCount().by(out()).addE().from('a') <-- group is reducing barrier so can't get back to 'a'.

or where later you try .addE().property('summary', outE().count) <--  even if you filter on label 'obs', you'd still get ALL 'obs' edges to all vertexes (assuming you could have more than one) but even then you'd have to have an additional filter.

So, let's try to put everything required for the edge up-front. The groupCount() is the same as group().by().by(count()). This will group each out vertex with its count.

gremlin> g.V().

......1> project('a','b').by().by(__.out('obs').groupCount()).as('group')

==>[a:v[0],b:[v[2]:2]]

==>[a:v[2],b:[v[4]:1]]

==>[a:v[4],b:[]]

This is our 'group' that we can reference later.  It has the start vertex 'a', the out vertex 'b' key, and the count 'b' value. Everything we need.

I'm assuming you could have more than one out vertex.  AddE() does not work on multiple targets so you can't start with the starting vertex. Let's get to the 'out' vertexes as they will need to drive the addE().

 

gremlin> g.V().

......1> project('a','b').by().by(__.out('obs').groupCount()).as('group').

......2> select('b').unfold().as('out').select(keys)

==>v[2]

==>v[4]

Let's start adding that edge to the current vertex... 

addE('summary').

Here we need to get back to the original, starting vertex.  It's in the projected map 'group' as key 'a'.

from(select('group').select('a')).

Then add the property.  We reference back to the 'out' map and select the value which contains the count().

property('summary', select('out').select(values))

And as I saw in Kelvin's reply, this did take some experimentation to get right. I had to relearn some of the rules as I went along ;-)

HTH

Robert Dale

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/1364f7bf-6c07-4e91-b0d3-dd9237485e07%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

You received this message because you are subscribed to a topic in the Google Groups "Gremlin-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/gremlin-users/QEn7URER8aU/unsubscribe.

To unsubscribe from this group and all its topics, send an email to gremlin-users+unsubscribe@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/CABed_4pGvDLvL%2B%2B2Yt%2BZQwafTO1zFBAO5_yS9f%2BCnysob8BVng%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

--

You received this message because you are subscribed to the Google Groups "Gremlin-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/CAO%3DOAK%2BBFg1sX%2Bqy2LkZAfKB5Y3bQCLdRq6gF%3Dr1cMo6CJNmiw%40mail.gmail.com.

[Kelvin Lawrence]

A big plus one on read everything from Mt Kuppitz :-)

I have had several people suggest to me recently that the whole area of collections that a traversal builds up and also barrier steps is an area that people find confusing. Based on that feedback and this conversation as well  I will move coverage of that topic to the top of my todo list for things to add next to the  book.

Robert Dale

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/1364f7bf-6c07-4e91-b0d3-dd9237485e07%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

You received this message because you are subscribed to a topic in the Google Groups "Gremlin-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/gremlin-users/QEn7URER8aU/unsubscribe.

To unsubscribe from this group and all its topics, send an email to gremlin-user...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/CABed_4pGvDLvL%2B%2B2Yt%2BZQwafTO1zFBAO5_yS9f%2BCnysob8BVng%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "Gremlin-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.

[Stephen Mallette]

I'd advocate for not only reading Kuppitz but really digging into any Gremlin you don't understand at a glance and trying those traversals for yourself in the Gremlin Console. Make sure you understand what each step of the traversal is accomplishing and how it is transforming the data in traversers, mutating side-effects, etc. Change the steps, the parameters, the sample data and see what happens. These points are some of what I expect to get into with my "Gremlin's Anatomy" talk. 

As Kelvin mentioned I've noticed and heard that there isn't a lot of information available on collections. To that end, I'm working on a "Collections" recipe now that goes into this topic with some really nice examples. We'll have that as part of 3.2.8 and 3.3.2.

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/f669407a-1c35-475e-90b0-2bffe514df6d%40googlegroups.com.

[Olav Laudy]

Yes sir! that's exactly what I'm doing with each of the pearls I've been given on this mailinglist. 

Thank you and I really appreciate your responses!

[Olav Laudy]

o o, trouble in paradise:

Locally I'm running Gremlin 3.3.1 and the query works beautifully.

However, the real thing is supposed to run on AWS Neptune and (with the exact nodes/vertices and query as above) I receive:

"startup failed:\nScript32.groovy: 1: [Static type checking] - 

Cannot call org.apache.tinkerpop.gremlin.process.traversal.dsl.graph.GraphTraversal <org.apache.tinkerpop.gremlin.structure.Vertex, org.apache.tinkerpop.gremlin.structure.Edge>#from(org.apache.tinkerpop.gremlin.process.traversal.Traversal <org.apache.tinkerpop.gremlin.structure.Edge, org.apache.tinkerpop.gremlin.structure.Vertex>) with arguments [org.apache.tinkerpop.gremlin.process.traversal.dsl.graph.GraphTraversal <A extends java.lang.Object, E2 extends java.lang.Object>] \n @ line 1, column 1.\n   

g.V().hasLabel('object1').project('a','b')          

     .by()          

     .by(__.out('obs1').groupCount()).as('group')              

     .select('b').unfold().as('out').select(keys)              

     .addE('summary')                

     .from(select('group').select('a'))                

       .property('summary', select('out').select(values))

\n   ^\n\n1 error\n

Note I changed object->object1 and obs->obs1 to not interfere with existing labels. The part up to .addE works fine. The moment I use the addE part, even without the property added, I receive this error.

[Olav Laudy]

This is a comparison of the features enabled by Gremlin 3.3.1 and AWS Neptune:

Is there any feature that needs to be enabled for the above query to work?

(from https://docs.aws.amazon.com/neptune/latest/userguide/access-graph-gremlin-differences.html)

	Neptune	Gremlin	Difference
Graph Feature
Transactions	TRUE	FALSE	FALSE
ThreadedTransactions	FALSE	FALSE	TRUE
Computer	FALSE	TRUE	FALSE
Persistence	TRUE	TRUE	TRUE
ConcurrentAccess	TRUE	FALSE	FALSE
Variable Feature
Variables	FALSE	TRUE	FALSE
SerializableValues	FALSE	TRUE	FALSE
UniformListValues	FALSE	TRUE	FALSE
BooleanArrayValues	FALSE	TRUE	FALSE
DoubleArrayValues	FALSE	TRUE	FALSE
IntegerArrayValues	FALSE	TRUE	FALSE
StringArrayValues	FALSE	TRUE	FALSE
BooleanValues	FALSE	TRUE	FALSE
ByteValues	FALSE	TRUE	FALSE
DoubleValues	FALSE	TRUE	FALSE
FloatValues	FALSE	TRUE	FALSE
IntegerValues	FALSE	TRUE	FALSE
LongValues	FALSE	TRUE	FALSE
MapValues	FALSE	TRUE	FALSE
MixedListValues	FALSE	TRUE	FALSE
StringValues	FALSE	TRUE	FALSE
ByteArrayValues	FALSE	TRUE	FALSE
FloatArrayValues	FALSE	TRUE	FALSE
LongArrayValues	FALSE	TRUE	FALSE
Vertex Feature
MetaProperties	FALSE	TRUE	FALSE
DuplicateMultiProperties	FALSE	TRUE	FALSE
AddVertices	TRUE	TRUE	TRUE
RemoveVertices	TRUE	TRUE	TRUE
MultiProperties	TRUE	TRUE	TRUE
UserSuppliedIds	TRUE	TRUE	TRUE
AddProperty	TRUE	TRUE	TRUE
RemoveProperty	TRUE	TRUE	TRUE
NumericIds	FALSE	TRUE	FALSE
StringIds	TRUE	TRUE	TRUE
UuidIds	FALSE	TRUE	FALSE
CustomIds	FALSE	FALSE	TRUE
AnyIds	FALSE	TRUE	FALSE
Vertex Property Feature
UserSuppliedIds	FALSE	TRUE	FALSE
AddProperty	TRUE	TRUE	TRUE
RemoveProperty	TRUE	TRUE	TRUE
NumericIds	TRUE	TRUE	TRUE
StringIds	TRUE	TRUE	TRUE
UuidIds	FALSE	TRUE	FALSE
CustomIds	FALSE	FALSE	TRUE
AnyIds	FALSE	TRUE	FALSE
Properties	TRUE	TRUE	TRUE
SerializableValues	FALSE	TRUE	FALSE
UniformListValues	FALSE	TRUE	FALSE
BooleanArrayValues	FALSE	TRUE	FALSE
DoubleArrayValues	FALSE	TRUE	FALSE
IntegerArrayValues	FALSE	TRUE	FALSE
StringArrayValues	FALSE	TRUE	FALSE
BooleanValues	TRUE	TRUE	TRUE
ByteValues	TRUE	TRUE	TRUE
DoubleValues	TRUE	TRUE	TRUE
FloatValues	TRUE	TRUE	TRUE
IntegerValues	TRUE	TRUE	TRUE
LongValues	TRUE	TRUE	TRUE
MapValues	FALSE	TRUE	FALSE
MixedListValues	FALSE	TRUE	FALSE
StringValues	TRUE	TRUE	TRUE
ByteArrayValues	FALSE	TRUE	FALSE
FloatArrayValues	FALSE	TRUE	FALSE
LongArrayValues	FALSE	TRUE	FALSE
Edge Feature
AddEdges	TRUE	TRUE	TRUE
RemoveEdges	TRUE	TRUE	TRUE
UserSuppliedIds	TRUE	TRUE	TRUE
AddProperty	TRUE	TRUE	TRUE
RemoveProperty	TRUE	TRUE	TRUE
NumericIds	FALSE	TRUE	FALSE
StringIds	TRUE	TRUE	TRUE
UuidIds	FALSE	TRUE	FALSE
CustomIds	FALSE	FALSE	TRUE
AnyIds	FALSE	TRUE	FALSE
Edge Property Feature
Properties	TRUE	TRUE	TRUE
SerializableValues	FALSE	TRUE	FALSE
UniformListValues	FALSE	TRUE	FALSE
BooleanArrayValues	FALSE	TRUE	FALSE
DoubleArrayValues	FALSE	TRUE	FALSE
IntegerArrayValues	FALSE	TRUE	FALSE
StringArrayValues	FALSE	TRUE	FALSE
BooleanValues	TRUE	TRUE	TRUE
ByteValues	TRUE	TRUE	TRUE
DoubleValues	TRUE	TRUE	TRUE
FloatValues	TRUE	TRUE	TRUE
IntegerValues	TRUE	TRUE	TRUE
LongValues	TRUE	TRUE	TRUE
MapValues	FALSE	TRUE	FALSE
MixedListValues	FALSE	TRUE	FALSE
StringValues	TRUE	TRUE	TRUE
ByteArrayValues	FALSE	TRUE	FALSE
FloatArrayValues	FALSE	TRUE	FALSE
LongArrayValues	FALSE	TRUE	FALSE

[Robert Dale]

Try casting it.

.from((Vertex) select('group').select('a'))  

The method signature changed in 3.3.1 so you don't have to do that.

Robert Dale

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/638f1987-4cf0-4e06-84d6-d5e3b3aabc12%40googlegroups.com.

[Jean-Baptiste Musso]

Thanks for sharing this Robert, very nice explanation!

That recipe is very useful since precomputing edges between vertices or properties on vertices is often required for performance requirement, especially when traversing big graphs. Will mentally bookmark this.

Jean-Baptiste

To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/CABed_4pFd7Qn4rni-isZvL2Q6szkGtsXE-tsGr_auOJtZNKE_A%40mail.gmail.com.

[Olav Laudy]

Hi,

I'm sorry I'm so blind here:

This is my (simplified) query: 

g.V().hasLabel('object1').project('a','b')          

     .by()          

     .by(__.out('obs1').groupCount()).as('group')              

     .select('b').unfold().as('out').select(keys)              

     .addE('summary')                

     .from(select('group').select('a'))         

which I change to:

g.V().hasLabel('object1').project('a','b')          
     .by()          

     .by(__.out('obs1').groupCount()).as('group')              

     .select('b').unfold().as('out').select(keys)              
     .addE('summary')                

     .from( (Vertex) select('group').select('a'))     

Now gives me:

Error encountered evaluating script

Jean-Baptiste

Robert Dale

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/1364f7bf-6c07-4e91-b0d3-dd9237485e07%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

You received this message because you are subscribed to a topic in the Google Groups "Gremlin-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/gremlin-users/QEn7URER8aU/unsubscribe.

To unsubscribe from this group and all its topics, send an email to gremlin-user...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/CABed_4pGvDLvL%2B%2B2Yt%2BZQwafTO1zFBAO5_yS9f%2BCnysob8BVng%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "Gremlin-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/CAO%3DOAK%2BBFg1sX%2Bqy2LkZAfKB5Y3bQCLdRq6gF%3Dr1cMo6CJNmiw%40mail.gmail.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "Gremlin-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.

[Daniel Kuppitz]

Casting to Vertex is invalid. It should either be cast to Traversal<Edge, Vertex> (which will probably fail, too) or to the non-generic version Traversal. The latter would be my best bet.

g.V().hasLabel('object1').

  project('a','b').

    by().

    by(__.out('obs1').groupCount()).as('group').

  select('b').unfold().as('out').

  select(keys).

  addE('summary').                

    from((Traversal) select('group').select('a'))

However, I don't see the need for groupCount() and project() in your query. Why don't you just use a much simpler query?

g.V().hasLabel('object1').as('a').

  local(out('obs1').dedup()).

  addE('summary').from('a')

Maybe I missed something (and yes, I'm too lazy to reread the whole thread :)). Just in case you actually want to use the group-counts:

g.V().hasLabel('object1').as('a').

  map(out('obs1').groupCount()).

  unfold().as('x').select(keys).

  addE('summary').

    from('a').

  property('count', select('x').select(values))

This query will create the summary edges and store the groupCount values as a property on the respective edge.

Cheers,

Daniel

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/1ed93e8d-5b62-4ab1-a410-00a76c0ab3ce%40googlegroups.com.

[Stephen Mallette]

> Casting to Vertex is invalid. It should either be cast to Traversal<Edge, Vertex> (which will probably fail, too) 

The API simply doesn't support addV(Traversal) prior to 3.3.1. I'm pretty sure Neptune is 3.3.0 - you have to be sure the versions align. Perhaps in TinkerPop 4.x we can do a better job of dealing with situations like this. Better messaging at the least - smarter handling of incompatibilities at the best.

To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/CA%2Bf9seVrVAHF7Maykgc%2B2ZiPtHO9RVtP5eNyc57GGvBJNL4ppA%40mail.gmail.com.

[Olav Laudy]

Brilliant! This did the trick:

g.V().hasLabel('object1').as('a').

  map(out('obs1').groupCount()).

  unfold().as('x').select(keys).

  addE('summary').

    from('a').

  property('count', select('x').select(values))

For completeness and searchability:

How to create a new edge that counts the number of edges between two nodes. It's basically a pre-compilation or a summary edge.

Start with:

g.addV('object1').property('name',1)

g.addV('object1').property('name',2)

g.addV('object1').property('name',3)

g.V().has('object1','name',1).as('A').V().has('object1','name',2).addE('obs1').from('A')

g.V().has('object1','name',1).as('A').V().has('object1','name',2).addE('obs1').from('A')

g.V().has('object1','name',2).as('A').V().has('object1','name',3).addE('obs1').from('A')

Apply the query above and receive:

['Edge:',
 {'id': 'd4b08f52-530e-628a-3955-76a4ef8bdd99',
  'inV': '1eb08f52-45e1-b633-af8d-a32dc812bb10',
  'inVLabel': 'object1',
  'label': 'summary',
  'outV': '72b08f52-456f-aa41-4a10-e55c5782b998',
  'outVLabel': 'object1',
  'properties': {'count': 2}},
 'Edge:',
 {'id': '52b08f52-5312-afa8-e7a2-f103893efdc8',
  'inV': '6ab08f52-4650-3975-45e5-629ab72a3266',
  'inVLabel': 'object1',
  'label': 'summary',
  'outV': '1eb08f52-45e1-b633-af8d-a32dc812bb10',
  'outVLabel': 'object1',
  'properties': {'count': 1}}]