Recommendation nodes with edge filter

Users Dgraph
1

Hi , i have this query for recommendation nodes. {
find as var(func: eq(first_name,"mehrdad")) {
sc as math(1)
fr as follow {
follow {
fscore as math(sc)
}
}
}
recommendationProfile(func: uid(fscore), orderdesc: val(fscore), first: 10) @filter(not uid(find, fr)) {
id
first_name
}
} i want set a filter for this query .
filter is :
when find users for recommend to a user , filter users that have a experience common with the user (example: “mehrdad”)
experience is a edge that user(mehrdad) has a connect to this.this picture is a sample data for this recommend.
the green edge is experience edge and blue edge is follow edge.
in this data all 2 experience edge is same node.i don’t know how to write this .
thank you for help

2

Screenshot from 2020-01-25 16-01-00
Screenshot from 2020-01-25 16-01-001920×1080 154 KB

3 
{
	find as var(func: eq(first_name,"مهرداد")) {
		sc as math(1)
		fr as follow {
			follow {
				fscore as math(sc)
			}
		}
	}
	recommendationProfile(func: uid(fscore), orderdesc: val(fscore), first: 10) @filter(not uid(find, fr)) {
		id
		first_name
	}
}
4

Just posted a blog post about this kind of use cases :slight_smile:
Normally you should use the reverse edge of follow relationship in the nested block.
follow predicate must have the @reverse index.

{
	find as var(func: eq(first_name,"مهرداد")) {
		sc as math(1)
		fr as follow {
			~follow @filter(not uid(find)) {
				fscore as math(sc)
			}
		}
	}
	recommendationProfile(func: uid(fscore), orderdesc: val(fscore), first: 10)  {
		id
		first_name
	}
}

Unless there is something I did not get in your use case.

Now you want to add a constraint on the person. You can do that with @cascade directive to find only the nodes matching a patterns. In your case the person must have at least one relationships with an Experience nodes from ‘find’.
So something like that should work for you , assuming that experience is the relationship between a user and an Experience node.

{
	find as var(func: eq(first_name,"مهرداد")) {
		sc as math(1)
        find_experience as experience  # keep the set of experiences for user A
		fr as follow {
			~follow @filter(not uid(find))  @cascade { 
            # with @cascade all nested levels must have at least one data
                experience @filter( uid(find_experience))  { }
				fscore as math(sc)
			}
		}
	}
	recommendationProfile(func: uid(fscore), orderdesc: val(fscore), first: 10)  {
		id
		first_name
	}
}

with @cascade, you only get the nodes with a non empty sub-graph

experience @filter( uid(find_experience))  { }

which means at least one experience in common. I don’t think you need the {} at all.

Let me know if it is the result you wanted.