Sum paths in weighed graph - gremlin

Question

I have a bi-directional graph, similar to this: https://gremlify.com/6zxjsbstb5f, where out edges have a weighted property.
There is a closeness relationship between articles, the weight total of all paths between 2 articles
So far I've been able to get the paths between articles, but the weighting is only the weight value of the unique path. I would like the aggregate (sum) weight of all paths between the starting article from the set returned by: repeat(outE().inV().simplePath()).until(hasLabel('article'))
g.V('70679').
repeat(outE().inV().simplePath()).
until(hasLabel('article')).as('a').
path().as('p').
map(unfold().coalesce(values('weight'),constant(0)).sum()).as('weighting').
select('weighting', 'p')
Steps to create the sample graph (taken from Gremlify)
g.addV('article').as('1').
addV('brand').as('2').
addV('article').as('3').
addV('category').as('4').
addV('zone').as('5').
addV('article').as('6').
addV('article').as('7').
addE('zone').from('1').to('5'). property('weight', 0.1).
addE('category').from('1').to('4').property('weight', 0.5).
addE('brand').from('1').to('2').property('weight', 0.8).
addE('article').from('2').to('6').
addE('article').from('2').to('1').
addE('article').from('2').to('3').
addE('zone').from('3').to('5').property('weight', 0.1).
addE('category').from('3').to('4').property('weight', 0.3).
addE('brand').from('3').to('2').property('weight', 0.4).
addE('article').from('4').to('1').
addE('article').from('4').to('3').
addE('article').from('5').to('6').
addE('article').from('5').to('7').
addE('article').from('5').to('1').
addE('article').from('5').to('3').
addE('zone').from('6').to('5').property('weight', 0.1).
addE('brand').from('6').to('2').property('weight', 0.6).
addE('zone').from('7').to('5').property('weight', 0.1)
I've been able to get this query which is close to what we require, where 8630 is an article Id in the graph
g.V('8630')
.repeat(outE().inV().simplePath())
.until(hasLabel('article')).as('foundArticle')
.path()
.map(unfold().coalesce(values('weight'), constant(0)).sum()).as('pathWeight')
.group().by(select('foundArticle').id()).as('grouping')
This produces results similar to:
[
{
"8634": [0.1, 0.5, 0.8]
},
{
"8640": [0.1, 0.8]
},
{
"8642": [0.1]
}
]
More desirable would be a result set similar to:
[
{
"8634": 1.4
},
{
"8640": 0.9
},
{
"8642": 0.1
}
]

Answer 1

Just to make it easier I gave each article a custom ID. The ID A1 corresponds to the example output you showed for ID 8630.
g.addV('article').as('1').property(id,'A1').
addV('brand').as('2').
addV('article').as('3').property(id,'A2').
addV('category').as('4').
addV('zone').as('5').
addV('article').as('6').property(id,'A3').
addV('article').as('7').property(id,'A4').
addE('zone').from('1').to('5'). property('weight', 0.1).
addE('category').from('1').to('4').property('weight', 0.5).
addE('brand').from('1').to('2').property('weight', 0.8).
addE('article').from('2').to('6').
addE('article').from('2').to('1').
addE('article').from('2').to('3').
addE('zone').from('3').to('5').property('weight', 0.1).
addE('category').from('3').to('4').property('weight', 0.3).
addE('brand').from('3').to('2').property('weight', 0.4).
addE('article').from('4').to('1').
addE('article').from('4').to('3').
addE('article').from('5').to('6').
addE('article').from('5').to('7').
addE('article').from('5').to('1').
addE('article').from('5').to('3').
addE('zone').from('6').to('5').property('weight', 0.1).
addE('brand').from('6').to('2').property('weight', 0.6).
addE('zone').from('7').to('5').property('weight', 0.1)
The query you had produced, was actually very close to having the result you wanted. I just added a second by step to the group to sum up the values.
g.V('A1').
repeat(outE().inV().simplePath()).
until(hasLabel('article')).as('foundArticle').
path().
map(unfold().coalesce(values('weight'), constant(0)).sum()).as('pathWeight').
group().
by(select('foundArticle').id()).
by(sum()).
unfold()
Which yields
{'A2': 1.4}
{'A3': 0.9}
{'A4': 0.1}
I think your query can also be simplified. If I come up with something simpler I will add it to this answer,
UPDATED
Here's a version of the query that uses sack and avoids needing to collect the path and post process it.
g.withSack(0).
V('A1').
repeat(outE().sack(sum).by(coalesce(values('weight'),constant(0))).
inV().simplePath()).
until(hasLabel('article')).
group().
by(id()).
by(sack().sum()).
unfold()
which again yields
{'A2': 1.4}
{'A3': 0.9}
{'A4': 0.1}