Uploaded on


More in: Technology , Business
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
No Downloads


Total Views
On Slideshare
From Embeds
Number of Embeds



Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

    No notes for slide
  • There existed a number of different ways to query a graph database. This one aims to make querying easy, and to produce queries that are readable. We looked at alternatives - SPARQL, SQL, Gremlin and other...


  • 1. Cypher Query Language Chicago Graph Database Meet-Up Max De Marzi
  • 2. What is Cypher?• Graph Query Language for Neo4j• Aims to make querying simple
  • 3. Why Cypher? • Existing Neo4j query mechanisms were not simple enough • Too verbose (Java API) • Too prescriptive (Gremlin)
  • 4. SQL? • Unable to express paths • these are crucial for graph-based reasoning • Neo4j is schema/table free
  • 5. SPARQL? • SPARQL designed for a different data model • namespaces • properties as nodes • high learning curve
  • 6. Design
  • 7. Design Decisions Declarative Most of the time, Neo4j knows better than you Imperative Declarative follow relationship specify starting pointbreadth-first vs depth-first specify desired outcome explicit algorithm algorithm adaptable based on query
  • 8. Design Decisions Pattern matching
  • 9. Design Decisions Pattern matching A B C
  • 10. Design Decisions Pattern matching
  • 11. Design Decisions Pattern matching
  • 12. Design Decisions Pattern matching
  • 13. Design Decisions Pattern matching
  • 14. Design Decisions ASCII-art patterns () --> ()
  • 15. Design Decisions Directed relationship A B (A) --> (B)
  • 16. Design Decisions Undirected relationship A B (A) -- (B)
  • 17. Design Decisions specific relationships LOVES A B A -[:LOVES]-> B
  • 18. Design Decisions Joined paths A B C A --> B --> C
  • 19. Design Decisions multiple paths A B C A --> B --> C, A --> C A --> B --> C <-- A
  • 20. Design Decisions Variable length paths A B A B A B ... A -[*]-> B
  • 21. Design Decisions Optional relationships A B A -[?]-> B
  • 22. Design Decisions Familiar for SQL users select start from match where where group by return order by
  • 23. STARTSELECT *FROM PersonWHERE firstName = “Max”START max=node:persons(firstName = “Max”)RETURN max
  • 24. MATCHSELECT skills.*FROM usersJOIN skills ON = skills.user_idWHERE = 101START user = node(101)MATCH user --> skillsRETURN skills
  • 25. Optional MATCHSELECT skills.*FROM usersLEFT JOIN skills ON = skills.user_idWHERE = 101START user = node(101)MATCH user –[?]-> skillsRETURN skills
  • 26. SELECT skills.*, user_skill.*FROM usersJOIN user_skill ON = user_skill.user_idJOIN skills ON user_skill.skill_id = = 1
  • 27. START user = node(1)MATCH user -[user_skill]-> skillRETURN skill, user_skill
  • 28. IndexesUsed as multiple starting points, not to speedup any traversalsSTART a = node:nodes_index(type=User) MATCHa-[r:knows]-bRETURN ID(a), ID(b), r.weight
  • 29.
  • 30. Complicated MatchSome UGLY recursive self join on the groupstableSTART max=node:person(name=“Max")MATCH group <-[:BELONGS_TO*]- maxRETURN group
  • 31. WhereSELECT person.*FROM personWHERE person.age >32 OR = "bald"START person = node:persons("name:*") WHEREperson.age >32 OR = "bald"RETURN person
  • 32. ReturnSELECT, count(*)FROM PersonGROUP BY person.nameORDER BY person.nameSTART person=node:persons("name:*"), count(*)ORDER BY
  • 33. Order By, ParametersSame as SQL{node_id} expected as part of requestSTART me = node({node_id})MATCH (me)-[?:follows]->(friends)-[?:follows]->(fof)-[?:follows]->(fofof)-[?:follows]->othersRETURN,,,, count(others)ORDER BY,,, count(others) DESC
  • 34.
  • 35. Graph FunctionsSome UGLY multiple recursive self and inner joins onthe user and all related tablesSTART lucy=node(1000), kevin=node(759) MATCH p= shortestPath( lucy-[*]-kevin ) RETURN p
  • 36. Aggregate FunctionsID: get the neo4j assigned identifierCount: add up the number of occurrencesMin: get the lowest valueMax: get the highest valueAvg: get the average of a numeric valueDistinct: remove duplicatesSTART me = node:nodes_index(type = user)MATCH (me)-[r?:wrote]-()RETURN ID(me),, count(r), min(, max(" ORDERBY ID(me)
  • 37. FunctionsCollect: put all values in a listSTART a = node:nodes_index(type=User)MATCH a-[:follows]->bRETURN, collect(
  • 38.
  • 39. Combine FunctionsCollect the ID of friendsSTART me = node:nodes_index(type = user)"MATCH (me)<-[r?:wrote]-(friends)RETURN ID(me),, collect(ID(friends)), collect( BY ID(me)
  • 40.
  • 41. UsesRecommend FriendsSTART me = node({node_id})MATCH (me)-[:friends]->(friend)-[:friends]->(foaf)RETURN
  • 42. UsesSix Degrees of Kevin BaconLength: counts the number of nodes along a pathExtract: gets the nodes/relationships from a pathSTART me=node({start_node_id}), them=node({destination_node_id})MATCH path = allShortestPaths( me-[?*]->them )RETURN length(path), extract(person in nodes(path) :
  • 43. UsesSimilar UsersUsers who rated same items within 2 points.Abs: gets absolute numeric valueSTART me = node(user1)MATCH (me)-[myRating:RATED]->(i)<-[otherRating:RATED]-(u)WHERE abs(myRating.rating-otherRating.rating)<=2RETURN u
  • 44. Boolean OperationsItems with a rating > 7 that similar users rated, but I have notAnd: this and that are trueOr: this or that is trueNot: this is falseSTART me=node(user1),        similarUsers=node(3) (result received in the first query)MATCH (similarUsers)-[r:RATED]->(item)WHERE r.rating > 7 AND NOT((me)-[:RATED]->(item)) RETURN item
  • 45. PredicatesALL: closure is true for all itemsANY: closure is true for any itemNONE: closure is true for no itemsSINGLE: closure is true for exactly 1 itemSTART london = node(1), moscow = node(2)MATCH path = london -[*]-> moscowWHERE all(city in nodes(path) = true)
  • 46. Design Decisions Parsed, not an internal DSL Execution Semantics Serialisation Type System Portability
  • 47. Design Decisions Database vs Application Design Goal: single user interaction expressible as single query Queries have enough logic to find required data, not enough to process it
  • 48. Implementation
  • 49. Implementation • Recursive matching with backtrackingSTART x=... MATCH x-->y, x-->z, y-->z, z-->a-->b, z-->b
  • 50. Implementation Execution Planstart n=node(0) Cypher is Pipesreturn n lazily evaluatedParameters() pulling from pipes underneathNodes(n)Extract([n])ColumnFilter([n])
  • 51. Implementation Execution Planstart n=node(0)match n-[*]-> breturn, n, count(*)order by n.ageParameters()Nodes(n)PatternMatch(n-[*]->b)Extract([, n])EagerAggregation( keys: [, n], aggregates: [count(*)])Extract([n.age])Sort(n.age ASC)ColumnFilter([,n,count(*)])
  • 52. Implementation Execution Planstart n=node(0)match n-[*]-> breturn, n, count(*)order by n.nameParameters()Nodes(n)PatternMatch(n-[*]->b)Extract([, n])Sort( ASC,n ASC)EagerAgregation( keys: [, n], aggregates: [count(*)])ColumnFilter([,n,count(*)])
  • 53. Thanks for Listening! Questions?