Load from CSV to create a binary tree - csv

Question

I need to create a binary tree in Neo4j. I've started with creating two CSV, one for vertices and one for edges and then I launched two query to create the entire tree.
I thought that I could create the entire tree with only one query.
The CSV from where I start is this:
"parent","child_1","child_1_attr1","child_1_attr2","edge_1_attr1","edge_1_attr2","child_2","child_2_attr1","child_2_attr2","edge_2_attr1","edge_2_attr2"
"vertex_1","vertex_2","2","5","4","1","vertex_3","5","3","2","2"
"vertex_2","vertex_4","3","5","2","3","vertex_5","4","4","4","3"
"vertex_3","vertex_6","2","1","2","4","vertex_7","2","2","5","5"
"vertex_4","vertex_8","4","4","4","5","vertex_9","2","3","2","5"
"vertex_5","vertex_10","1","1","3","3","vertex_11","1","3","2","3"
"vertex_6","vertex_12","3","1","1","1","vertex_13","1","2","5","1"
"vertex_7","vertex_14","4","2","2","1","vertex_15","2","5","4","3"
Then I tried this query:
LOAD CSV WITH HEADERS FROM 'file:///Prova1.csv' AS line
Match (p:Vertex {name: line.parent})
Create (c1:Vertex {name: line.child_1, attr1: line.child_1_attr1, attr2: line.child_1_attr2})
Create (c2:Vertex {name: line.child_2, attr1: line.child_2_attr1, attr2: line.child_2_attr2})
Create (p)<-[:EDGE {attr1: line.edge_1_attr1, attr2: line.edge_1_attr2}]-(c1)
Create (p)<-[:EDGE {attr1: line.edge_2_attr1, attr2: line.edge_2_attr2}]-(c2)
Before this query I create manually the first Vertex, and the I run this query, but the only result that I get is the creation of Vertices 1,2 and 3.
It should match the parent ( that is always already created ), then create the two childs and then it should connect these two children to his father.
Who can help me?

Answer 1

Likely your view of execution is, for each line/row, execute all Cypher code, then repeat this for the next line/row until finished, and this is incorrect.
Instead, each individual Cypher operation will execute for all rows, then the next Cypher operation will execute for all rows, etc.
This means your MATCH operation:
Match (p:Vertex {name: line.parent})
is performed across all lines in your CSV at the same time, and only then will it proceed to the next operation (your CREATE, acting on all lines), and so on.
Since you stated that you manually created the first vertex, that vertex is the only one that will be matched upon, the MATCH will fail for all other lines in your CSV because the CREATE statements haven't executed yet so those nodes don't exist. This means only two vertexes will be created, the children of that solely matched node.
It's usually good practice when importing CSV data to create all nodes first, and then have a separate CSV processing for matching upon the already created nodes and creating the relevant relationships.
However, if you do want to create everything in one goal, you'll likely want to use MERGE in various places, but this is also tricky if you don't fully understand the behavior of MERGE (it's like an attempt to MATCH, and if no match is found, a CREATE) or don't fully understand how Cypher is executed (as in this case).
You'll also want to MERGE according to unique node values instead of all properties, and SET the remaining properties. It's also especially helpful to have either unique constraints or indexes (whichever is appropriate) on the relevant label/properties for faster execution, especially as the size of your graph grows.
This query may work.
LOAD CSV WITH HEADERS FROM 'file:///Prova1.csv' AS line
MERGE (p:Vertex {name: line.parent})
MERGE (c1:Vertex {name: line.child_1})
SET c1.attr1 = line.child_1_attr1, c1.attr2 = line.child_1_attr2
MERGE (c2:Vertex {name: line.child_2})
SET c2.attr1 = line.child_2_attr1, c2.attr2 = line.child_2_attr2
Create (p)<-[:EDGE {attr1: line.edge_1_attr1, attr2: line.edge_1_attr2}]-(c1)
Create (p)<-[:EDGE {attr1: line.edge_2_attr1, attr2: line.edge_2_attr2}]-(c2)
The reason this one works is that by the time your very first MERGE is completed for all parent nodes, that will have created ALL parent nodes (or rather, nodes that will be parents) in your graph.
So when we reach the MERGE for your child nodes, that will MATCH most of those already created nodes in your graph...the only new nodes that will be created at that point will be leaf nodes, which would not have been created by your first MERGE, since they won't be parents to any other nodes, and do not show up in your parent column in your CSV.

Answer 2

For some reason the import query is not working because you match the parent first and then create your nodes and relationships. I modified the query this way and it is working now:
LOAD CSV WITH HEADERS FROM 'file:///test.csv' AS line
CREATE (c1:Vertex {name: line.child_1, attr1: line.child_1_attr1, attr2: line.child_1_attr2}),
(c2:Vertex {name: line.child_2, attr1: line.child_2_attr1, attr2: line.child_2_attr2}) WITH c1,c2, line
MATCH (p:Vertex {name:line.parent}) CREATE (p)<-[:EDGE {attr1: line.edge_1_attr1, attr2: line.edge_1_attr2}]-(c1),
(p)<-[:EDGE {attr1: line.edge_2_attr1, attr2: line.edge_2_attr2}]-(c2)
So if you first create the nodes and then match the parent and create the relationships the query is working. The result looks like this:
I will investigate on your query to find a reason why it is not working because I do not really understand why it is not working.

Answer 3

foreach (num in range(1,15) |
merge (parent:Node {number: num})
merge (left:Node {number: num + num})
merge (right:Node {number: num + num + 1})
merge (left)<-[:LEFT]-(parent)-[:RIGHT]->(right)
)
Explanation:
This creates a perfect binary tree structure with 31 nodes. Then you can include the same numbers in the CSV to find and add properties to each correspondingly numbered node.
In a binary tree if you include a number property on the first (root or top most node) with value 1, then increment each subsequent node's number value by 1 (left to right; top to bottom) then you get a convenient mathematical relationship where each node's left child has a number value of the parent's number + number and the right child is number + number + 1.