How to output append reason to similarities in Gremlin

Question

I have the following simple Graph:

User -- Likes --> Item

I'm finding top 10 similar users to a user u using the following Gremlin code:

u.out('Likes').in('Likes').filter([u]).groupCount.cap.orderMap(T.decr)[0..10].map()

This outputs something like that:

==>{userid=1}
==>{userid=5}
==>{userid=10}
==>{userid=15}

I would like the output to be more informative and have additional information such as rank in the sorted map and items (itemid) shared with the original user, something like this:

==>{userid=1, rank=0, reason_items={1,2,3,5}}
==>{userid=5, rank=1, reason_items={1,2,10}}
==>{userid=10, rank=2, reason_items={1,2,4}}
==>{userid=15, rank=3, reason_items={1,2}}

An efficient gremlin-groovy code sample would be nice!

Thank you.

Faber · Accepted Answer · 2015-03-05T15:21:37.653

By appending the appropriate transform closure to your query:

rank = 0; itemsU1 = [] as Set; u1.out('Likes').aggregate(itemsU1).in('Likes')
      .filter{it != u1}.groupCount.cap.orderMap(T.decr)
      .transform{[id:it.id, rank:rank++, reason_item_ids:itemsU1.intersect(it.out('Likes').toSet()).collect{it.id}]}

… you can obtain:

==>{id=User6, rank=0, reason_item_ids=[Item1, Item5]}
==>{id=User4, rank=1, reason_item_ids=[Item1, Item2]}
==>{id=User2, rank=2, reason_item_ids=[Item1]}
==>{id=User5, rank=3, reason_item_ids=[Item5]}
==>{id=User3, rank=4, reason_item_ids=[Item2]}

for the following sample graph:

g = new TinkerGraph()

u1 = g.addVertex('User1')
u2 = g.addVertex('User2')
u3 = g.addVertex('User3')
u4 = g.addVertex('User4')
u5 = g.addVertex('User5')
u6 = g.addVertex('User6')

i1 = g.addVertex('Item1')
i2 = g.addVertex('Item2')
i3 = g.addVertex('Item3')
i4 = g.addVertex('Item4')
i5 = g.addVertex('Item5')

g.addEdge(u1,i1,'Likes')
g.addEdge(u1,i2,'Likes')
g.addEdge(u1,i5,'Likes')
g.addEdge(u2,i1,'Likes')
g.addEdge(u2,i4,'Likes')
g.addEdge(u3,i2,'Likes')
g.addEdge(u4,i1,'Likes')
g.addEdge(u4,i2,'Likes')
g.addEdge(u4,i3,'Likes')
g.addEdge(u5,i4,'Likes')
g.addEdge(u5,i5,'Likes')
g.addEdge(u6,i1,'Likes')
g.addEdge(u6,i4,'Likes')
g.addEdge(u6,i5,'Likes')

Daniel Kuppitz · Answer 2 · 2015-03-06T00:30:05.023

Given Faber's sample graph, you can do this:

u = u1; m = [:].withDefault {[]}; rank = 0; key = null
u.out('Likes').as('item').in('Likes').except([u]).as('user').select().groupBy {
  key = it.getColumn('user')
} {
  m[key] << it.getColumn('item').id
} {
  it.size()
}.cap().orderMap(T.decr)[0..10].transform {[
  'userid'         : it.id,
  'rank'           : rank++,
  'reason_item_ids': m[it]
]}

No need for nested traversals within .transform().

How to output append reason to similarities in Gremlin

2 Answers2