Neo4j: multiple counts from multiple matches - neo4j

Given a neo4j schema similar to
(:Person)-[:OWNS]-(:Book)-[:CATEGORIZED_AS]-(:Category)
I'm trying to write a query to get the count of books owned by each person as well as the count of books in each category so that I can calculate the percentage of books in each category for each person.
I've tried queries along the lines of
match (p:Person)-[:OWNS]-(b:Book)-[:CATEGORIZED_AS]-(c:Category)
where person.name in []
with p, b, c
match (p)-[:OWNS]-(b2:Book)-[:CATEGORIZED_AS]-(c2:Category)
with p, b, c, b2
return p.name, b.name, c.name,
count(distinct b) as count_books_in_category,
count(distinct b2) as count_books_total
But the query plan is absolutely horrible when trying to do the second match. I've tried to figure out different ways to write the query so that I can do the two different counts, but haven't figured out anything other than doing two matches. My schema isn't really about people and books. The :CATEGORIZED_AS relationship in my example is actually a few different relationship options, specified as [:option1|option2|option3]. So in my 2nd match I repeat the relationship options so that my total count is constrained by them.
Ideas? This feels similar to Neo4j - apply match to each result of previous match but there didn't seem to be a good answer for that one.

UNWIND is your friend here. First, calculate the total books per person, collecting them as you go.
Then unwind them so you can match which categories they belong to.
Aggregate by category and person, and you should get the number of books in each category, for a person
match (p:Person)-[:OWNS]->(b:Book)
with p,collect(b) as books, count(b) as total
with p,total,books
unwind books as book
match (book)-[:CATEGORIZED_AS]->(c)
return p,c, count(book) as subtotal, total

Related

Is there a simpler version of this cypher query?

I have constructed a query to find the people who follow each other and who have read books in the same genre. Here it is:
MATCH (u1:User)-[:READ]->(b1:Book)
WITH collect(DISTINCT b1.genre) AS genres,u1 AS user1
MATCH (u2:User)-[:READ]->(b2:Book)
WHERE (user1)<-[:FOLLOWS]->(u2) AND b2.genre IN genres
RETURN DISTINCT user1.username AS user1,u2.username AS user2
The idea is that we collect all the book genres for one of them, and if a book read by the other is in that list of genres (and they follow each other), then we return those users. This seems to work: we get a list of distinct pairs of individuals. I wonder, though, if there a quicker way to do this? My solution seems somewhat clumsy, but I found it surprisingly finicky trying to specify that they have read a book in the same genre without getting back all the pairs of books and duplicating individuals. For example, I
first wrote the following:
MATCH (b1:Book)<-[:READ]-(u1:User)-[:FOLLOWS]-(u2:User)-[:READ]->(b2:Book)
WHERE b1.genre = b2.genre
RETURN DISTINCT u1.username AS user1, u2.username AS user2
Which seems simpler, but in fact it returned repeated names for all the books that were read in the same genre. Is my solution the simplest, or is there a simpler one?
This is one way of rewriting the query
MATCH (n1:User)-[:FOLLOWS]-(n2:User)
MATCH (n1)-[:READ]->(book), (n2)-[:READ]->(book2)
WHERE book.genre = book2.genre
RETURN n1.username, n2.username, count(*)
Here is another collecting genres for each user
MATCH (n1:User)-[:FOLLOWS]-(n2:User)
WITH n1, n2,
[(n1)-[:READ]->(book) | book.genre] AS g1,
[(n2)-[:READ]->(book) | book.genre] AS g2
WHERE ANY(x IN g1 WHERE x IN g2)
RETURN n1, n2, count(*)
Note that sometimes longer queries are not especially better in the sense that the ways the data are retrieved need to make sense to yourself.
Your model however clearly shows that you would benefit from a bit of graph refactoring, extracting the genre into its own node, for eg
MATCH (n:Book)
MERGE (g:Genre {name: n.genre})
MERGE (n)-[:HAS_GENRE]->(g)
And this would be the new query which leverages a graph model
PROFILE
MATCH (n1:User)-[:FOLLOWS]-(n2:User)
WHERE (n1)-[:READ]->()-[:HAS_GENRE]->()<-[:HAS_GENRE]-()<-[:READ]-(n2)
RETURN n1.username, n2.username, count(*)

Nodes with relationship to multiple nodes

I want to get the Persons that know everyone in a group of persons which know some specific places.
This:
MATCH (:Place {name:'Breiter Weg'})<-[:knows]-(b:Person)-[:knows]->(:Place {name:'Buchhandel'})
WITH collect(DISTINCT b) as persons
Match (a:Person)
WHERE ALL(b in persons WHERE (a)-[:knows]->(b))
RETURN a
works, but for the second part does a full nodelabelscan, before applying the where clause, which is extremely slow - in a bigger db it takes 8~9 seconds. I also tried this:
MATCH (:Place {name:'Breiter Weg'})<-[:knows]-(b:Person)-[:knows]->(:Place {name:'Buchhandel'})
Match (a:Person)-[:knows]->(b)
RETURN a
This only needs 2ms, however it returns all persons that know any person of group b, instead of those that know everyone.
So my question is: Is there a effective/fast query to get what i want?
We have a knowledge base article for this kind of query that show a few approaches.
One of these is to match to :Persons known by the group, and then count the number of times each of those persons shows up in the results. Provided there aren't multiple :knows relationships between the same two people, if the count is equal to the collection of people from your first match, then that person must know all of the people in the collection.
MATCH (:Place {name:'Breiter Weg'})<-[:knows]-(b:Person)-[:knows]->(:Place {name:'Buchhandel'})
WITH collect(b) as persons
UNWIND persons as b // so we have the entire list of persons along with each person
WITH size(persons) as total, b
MATCH (a:Person)-[:knows]->(b)
WITH total, a, count(a) as knownCount
WHERE total = knownCount
RETURN a
Here is a simpler Cypher query that also compares counts -- the same basic idea used by #InverseFalcon.
MATCH (:Place {name:'Breiter Weg'})<-[:knows]-(b:Person)-[:knows]->(:Place {name:'Buchhandel'}), (a:Person)-[:knows]->(b)
WITH COLLECT({a:a, b:b}) as data, COUNT(DISTINCT b) AS total
UNWIND data AS d
WITH total, d.a AS a, COUNT(d.b) AS bCount
WHERE total = bCount
RETURN a

Neo4j: Query to find the nodes with most relationships, and their connected nodes

I am using Neo4j CE 3.1.1 and I have a relationship WRITES between authors and books. I want to find the N (say N=10 for example) books with the largest number of authors. Following some examples I found, I came up with the query:
MATCH (a)-[r:WRITES]->(b)
RETURN r,
COUNT(r) ORDER BY COUNT(r) DESC LIMIT 10
When I execute this query in the Neo4j browser I get 10 books, but these do not look like the ones written by most authors, as they show only a few WRITES relationships to authors. If I change the query to
MATCH (a)-[r:WRITES]->(b)
RETURN b,
COUNT(r) ORDER BY COUNT(r) DESC LIMIT 10
Then I get the 10 books with the most authors, but I don't see their relationship to authors. To do so, I have to write additional queries explicitly stating the name of a book I found in the previous query:
MATCH ()-[r:WRITES]->(b)
WHERE b.title="Title of a book with many authors"
RETURN r
What am I doing wrong? Why isn't the first query working as expected?
Aggregations only have context based on the non-aggregation columns, and with your match, a unique relationship will only occur once in your results.
So your first query is asking for each relationship on a row, and the count of that particular relationship, which is 1.
You might rewrite this in a couple different ways.
One is to collect the authors and order on the size of the author list:
MATCH (a)-[:WRITES]->(b)
RETURN b, COLLECT(a) as authors
ORDER BY SIZE(authors) DESC LIMIT 10
You can always collect the author and its relationship, if the relationship itself is interesting to you.
EDIT
If you happen to have labels on your nodes (you absolutely SHOULD have labels on your nodes), you can try a different approach by matching to all books, getting the size of the incoming :WRITES relationships to each book, ordering and limiting on that, and then performing the match to the authors:
MATCH (b:Book)
WITH b, SIZE(()-[:WRITES]->(b)) as authorCnt
ORDER BY authorCnt DESC LIMIT 10
MATCH (a)-[:WRITES]->(b)
RETURN b, a
You can collect on the authors and/or return the relationship as well, depending on what you need from the output.
You are very close: after sorting, it is necessary to rediscover the authors. For example:
MATCH (a:Author)-[r:WRITES]->(b:Book)
WITH b,
COUNT(r) AS authorsCount
ORDER BY authorsCount DESC LIMIT 10
MATCH (b)<-[:WRITES]-(a:Author)
RETURN b,
COLLECT(a) AS authors
ORDER BY size(authors) DESC

Neo4j/Cypher matching first n nodes in the traversal branch

I have graph: (:Sector)<-[:BELONGS_TO]-(:Company)-[:PRODUCE]->(:Product).
I'm looking for the query below.
Start with (:Sector). Then match first 50 companies in that sector and for each company match first 10 products.
First limit is simple. But what about limiting products.
Is it possible with cypher?
UPDATE
As #cybersam suggested below query will return valid results
MATCH (s:Sector)<-[:BELONGS_TO]-(c:Company)
WITH c
LIMIT 50
MATCH (c)-[:PRODUCE]->(p:Product)
WITH c, (COLLECT(p))[0..10] AS products
RETURN c, products
However this solution doesn't scale as it still traverses all products per company. Slice applied after each company products collected. As number of products grows query performance will degrade.
Each returned row of this query will contain: a sector, one of its companies (at most 50 per sector), and a collection of up to 10 products for that company:
MATCH (s:Sector)<-[:BELONGS_TO]-(c:Company)
WITH s, (COLLECT(c))[0..50] AS companies
UNWIND companies AS company
MATCH (company)-[:PRODUCE]->(p:Product)
WITH s, company, (COLLECT(p))[0..10] AS products;
Updating with some solutions using APOC Procedures.
This Neo4j knowledge base article on limiting results per row describes a few different ways to do this.
One way is to use apoc.cypher.run() to execute a limited subquery per row. Applied to the query in question, this would work:
MATCH (s:Sector)<-[:BELONGS_TO]-(c:Company)
WITH c
LIMIT 50
CALL apoc.cypher.run('MATCH (c)-[:PRODUCE]->(p:Product) WITH p LIMIT 10 RETURN collect(p) as products', {c:c}) YIELD value
RETURN c, value.products AS products
The other alternative mentioned is using APOC path expander procedures, providing the label on a termination filter and a limit:
MATCH (s:Sector)<-[:BELONGS_TO]-(c:Company)
WITH c
LIMIT 50
CALL apoc.path.subgraphNodes(c, {maxLevel:1, relationshipFilter:'PRODUCE>', labelFilter:'/Product', limit:10}) YIELD node
RETURN c, collect(node) AS products

Select nodes that has all relationships in Neo4j

Suppose I have two kinds of nodes, Person and Competency. They are related by a KNOWS relationship. For example:
(:Person {id: 'thiago'})-[:KNOWS]->(:Competency {id: 'neo4j'})
How do I query this schema to find out all Person that knows all nodes of a set of Competency?
Suppose that I need to find every Person that knows "java" and "haskell" and I'm only interested in the nodes that knows all of the listed Competency nodes.
I've tried this query:
match (p:Person)-[:KNOWS]->(c:Competency) where c.id in ['java','haskell'] return p.id;
But I get back a list of all Person that knows either "java" or "haskell" and duplicated entries for those who knows both.
Adding a count(c) at the end of the query eliminates the duplicates:
match (p:Person)-[:KNOWS]->(c:Competency) where c.id in ['java','haskell'] return p.id, count(c);
Then, in this particular case, I can iterate the result and filter out results that the count is less than two to get the nodes I want.
I've found out that I could do it appending consecutive match clauses to keep filtering the nodes to get the result I want, in this case:
match (p:Person)-[:KNOWS]->(:Competency {id:'haskell'})
match (p)-[:KNOWS]->(:Competency {id:'java'})
return p.id;
Is this the only way to express this query? I mean, I need to create a query by concatenating strings? I'm looking for a solution to a fixed query with parameters.
with ['java','haskell'] as skills
match (p:Person)-[:KNOWS]->(c:Competency)
where c.id in skills
with p.id, count(*) as c1 ,size(skills) as c2
where c1 = c2
return p.id
One thing you can do, is to count the number of all skills, then find the users that have the number of skill relationships equals to the skills count :
MATCH (n:Skill) WITH count(n) as skillMax
MATCH (u:Person)-[:HAS]->(s:Skill)
WITH u, count(s) as skillsCount, skillMax
WHERE skillsCount = skillMax
RETURN u, skillsCount
Chris
Untested, but this might do the trick:
match (p:Person)-[:KNOWS]->(c:Competency)
with p, collect(c.id) as cs
where all(x in ['java', 'haskell'] where x in cs)
return p.id;
How about this...
WITH ['java','haskell'] AS comp_col
MATCH (p:Person)-[:KNOWS]->(c:Competency)
WHERE c.name in comp_col
WITH comp_col
, p
, count(*) AS total
WHERE total = length(comp_col)
RETURN p.name, total
Put the competencies you want in a collection.
Match all the people that have either of those competencies
Get the count of compentencies by person where they have the same number as in the competency collection from the start
I think this will work for what you need, but if you are building these queries programatically the best performance you get might be with successive match clauses. Especially if you knew which competencies were most/least common when building your queries, you could order the matches such that the least common were first and the most common were last. I think that would chunk down to your desired persons the fastest.
It would be interesting to see what the plan analyzer in the sheel says about the different approaches.

Resources