Convert string into multidimensional array in Neo4J - neo4j

Question

I loaded data from csv into Neo4J. One column in the file is an array of arrays that Neo4J now considers one large string. How can I convert it back into an array?
My file looks like this:
Id, name, reviews
1, item1, "[[date1, User1, Rating1],
[date2, User2, Rating2],
[date3, User3, Rating3]] "
Into Neo4J:
USING PERIODIC COMMIT
LOAD CSV WITH HEADERS FROM "file:///data/file.csv" AS line
CREATE(:Product{
Id:toInteger(line.Id),
name:toString(line.name),
reviews:line.reviews})
RETURN line
Now the review column is loaded, but considered one large string.
"[date1, User1, Rating1], [date2, User2, Rating2], [date3, User3, Rating3]"
Is there any way to split it into two arrays like this:
First Array:
[date1, User1, Rating1], //0
[date2, User2, Rating2], //1
[date3, User3, Rating3] //2
Second Array example:
// 0 , 1 , 2
[date1, User1, Rating1]
I'd like to be able to acces my data like this:
MATCH (n) RETURN n.reviews[2] (output: date3, User3, Rating3)
MATCH (n) RETURN n.reviews[2][0] (output: date3)
MATCH (n) RETURN n.reviews[1][1] (output: User2)
Is there any way to do this?

Answer 1

Using APOC Procedures, you can use the apoc.convert.fromJsonList() function to convert the list, though you'll need to make sure each of the subitems in the arrays is quoted so they'll be interpreted as strings.
WITH "[['date1', 'User1', 'Rating1'], ['date2', 'User2', 'Rating2'], ['date3', 'User3', 'Rating3']]" as inputString
WITH apoc.convert.fromJsonList(input) as input
RETURN input[2][0] //returns 'date3'
Just a note, the conversion functions are currently missing in the APOC docs, but you can reference them and their signature by entering this in the Neo4j browser:
CALL apoc.help('fromjson')
And now for the bad news.
Although you can do this with literal input and parameters and convert from properties which are JSON strings, you cannot use a nested list as a property of a node or relationship, that's just a current limitation of the properties implementation.
That said, this feels like your modeling may need some improvement. We'd recommend using separate nodes for your reviews rather than this nested structure, so something like:
(:Product)-[:HAS_REVIEW]->(:Review)
Where the :Review node has the date and rating, and either has the user ID, or has a relationship to the user node who rated the product.
Usage would look something like:
MATCH (p:Product {id:12345})-[:HAS_REVIEW]->(r)
WITH p, collect(r) as reviews
...
At that point you have an (unordered) array of the review nodes, and can do index access to get a review at a particular index, then use dot notation to access the property or properties you want. If you want ordering, you'll need to do an explicit ORDER BY before the collect(), and you'll need something to order it by, probably the date of the review.