SQL Saturday Cork Welcome to Cork. SQL Server 2017 Graph feature

Transcription

1 SQL Saturday Cork Welcome to Cork SQL Server 2017 Graph feature 1

2 About me Rudi Bruchez 2

3 Agenda What are graphs anyway? Graph databases and SQL Server What you cannot do 3

4 What are graphs anyway? 4

5 Graph theory Discrete mathematics Discrete vs continous Finite mathematics E.g. Integers Graphs Statements in logic svg 5

6 Graphs a set of objects in which some pairs of the objects are in some sense "related A lot of problems and algorithms can be modeled as graphs 6

7 History of graph theory Seven Bridges of Königsberg, 1736 The city of Königsberg in Prussia (now Kaliningrad, Russia) was set on both sides of the Pregel River, and included two large islands which were connected to each other, or to the two mainland portions of the city, by seven bridges. The problem was to devise a walk through the city that would cross each of those bridges once and only once. By way of specifying the logical task unambiguously, solutions involving either - reaching an island or mainland bank other than via one of the bridges, or - accessing any bridge without crossing to its other end are explicitly unacceptable. Euler proved that the problem has no solution. The difficulty he faced was the development of a suitable technique of analysis, and of subsequent tests that established this assertion with mathematical rigor. 7

8 History of graph theory Seven Bridges of Königsberg, Euler pointed out that the choice of route inside each land mass is irrelevant. The only important feature of a route is the sequence of bridges crossed. This allowed him to reformulate the problem in abstract terms (laying the foundations of graph theory), eliminating all features except the list of land masses and the bridges connecting them. In modern terms, one replaces each land mass with an abstract "vertex" or node, and each bridge with an abstract connection, an "edge", which only serves to record which pair of vertices (land masses) is connected by that bridge. The resulting mathematical structure is called a graph. 8

9 node / vertex (vertices) 9

10 Edge / arc 10

11 Undirected Directed an undirected graph consists of a set of vertices and a set of edges (unordered pairs of vertices) a directed graph consists of a set of vertices and a set of arcs (ordered pairs of vertices). 11

12 Property graph A graph database is a directed property graph 1 2 Type: enemy Name: Doctor Who Age: 953 Name: Dalek A property graph contains connected entities (the nodes) which can hold any number of attributes (key-value-pairs) 12

13 Degree (valency) Number of edges reaching the vertex the degree (or valency) of a vertex of a graph is the number of edges incident to the vertex, with loops counted twice. 13

14 Adjacent vs incident 1 2 Adjacent to 2 Edge A Incident to 1 and 2 Adjacent to 1 Two vertices are called adjacent if they are connected by an edge. Two edges are called incident, if they share a vertex. Also, a vertex and an edge are called incident, if the vertex is one of the two vertices the edge connects. 14

15 What kind of graphs? 15

16 An example: a soft song W D F K 16

17 An example: a soft song FWDK * FWD * K FWK * D FDK * W FW * DK Direction DK * FW K * FWD W * FKD D * FWK * FWDK 17

18 An example: a soft song FWDK * FWD * K FWK * D FDK * W FW * DK Direction DK * FW K * FWD W * FKD D * FWK * FWDK The top line shows all possibilities with the ferryman being on the first border of the river. The bottom line shows all possibilities with the ferryman being on the other border of the river. 18

19 An example: a soft song FWDK * FWD * K FWK * D FDK * W FW * DK DK * FW K * FWD W * FKD D * FWK * FWDK 19

20 Graph databases and SQL Server 20

21 A native graph database Nordenskiöldsgatan Malmö SWEDEN

22 Domain Specific Languages Cypher Neo4J MATCH (actor:person)-[:acted_in]->(movie:movie) WHERE movie.title STARTS WITH "T" RETURN movie.title AS title, collect(actor.name) AS cast ORDER BY title ASC LIMIT 10; Apache Tinkerpop g.v().as("a").out("knows").as("b"). select("a","b"). by("name"). by("age") 22

23 Graph support in SQL Server 2017 Native nodes and edge table support Query language extension Join-free pattern matching Integrated syntax 23

24 Graph support in SQL Server 2017 (a bit of) Interoperability Columnstore Optimization through join algorithms (nested loop, hash) 24

25 Graph support in SQL Server 2017 One graph per database, containing multiple node and edge tables 25

26 Graph support in SQL Server 2017 One graph per database Can contain multiple node and edge tables Tables have «hidden» metadata to support the graph feature 26

27 Node table CREATE TABLE dbo.person ( PersonId int IDENTITY(1,1) NOT NULL PRIMARY KEY, FirstName varchar(50) NOT NULL, LastName varchar(50) NOT NULL ) AS NODE; 27

28 Edge table CREATE TABLE dbo.isfriendof AS EDGE; CREATE TABLE dbo.isfriendof ( Since date not null ) AS EDGE; No index, no unique constraint, no referential integrity 28

29 Demo create a movie database 29

30 Pseudo columns Nodes Edges implicit $node_id pseudo column object_id + graph_id maps to an internal name with hex string JSON representation when selected implicit $edge_id pseudo column $from_id = $node_id of origin $to_id = $node_id of termination 30

31 Demo metadata 31

32 Match syntax MATCH (<graph_search_pattern>) <graph_search_pattern>::= {<node_alias> { { <-( <edge_alias> )- } { -( <edge_alias> )-> } <node_alias> } } [ { AND } { ( <graph_search_pattern> ) } 32

33 Match syntax SELECT actor.personname, Movie.Title FROM dbo.person AS actor, dbo.actedin, dbo.movie WHERE MATCH(actor-(ActedIn)->Movie); $from_id $to_id 33

34 Draw it! SELECT actor.personname, Movie.Title FROM dbo.person AS actor, dbo.actedin, dbo.movie WHERE MATCH(actor-(ActedIn)->Movie); Person ActedIn Movie 34

35 Graph traversal who acted in the same movie as another actor SELECT actor1.personname, Movie.Title, actor2.personname FROM dbo.person AS actor1, dbo.actedin AS ActedIn1, dbo.movie, dbo.actedin AS ActedIn2, dbo.person AS actor2 WHERE MATCH(actor1-(ActedIn1)->Movie<-(ActedIn2)-actor2) AND actor1.personid <> actor2.personid; 35

36 Many joins Actor1 ActedIn1 Movie Actor2 ActedIn2 36

37 Multigraphs? SELECT p.personname, m.title, 'acted in' as role FROM dbo.person p, dbo.actedin a, dbo.movie m WHERE MATCH(p-(a)->m) AND p.personname = 'Tom Hanks' UNION SELECT p.personname, m.title, 'directed' as role FROM dbo.person p, dbo.directed d, dbo.movie m WHERE MATCH(p-(d)->m) AND p.personname = 'Tom Hanks'; 37

38 Demo queries and optimization 38

39 What you cannot do 39

40 Support Not supported on Memory-optimized tables Temporary tables Temporal tables Supported ColumnStore index Index on pseudo columns You can t ALTER to convert a table to Graph Create referential integrity Use MERGE 40

41 Are we in a RDBMS? Old-style JOIN why not replace FROM by MATCH? no OUTER JOIN No referential integrity an edge is not bound to a type of node multigraph 41

42 Demo no way 42

43 Restrictions in MATCH Directed edges, directed match Basic match syntax An edge name cannot be repeated inside MATCH No distance path (shortest path) OR and NOT not supported Subqueries using (NOT) EXISTS 43

44 Not supported Transitive closure Connection trough any number of hops Polymorphism Find any type of node connected to a given node UNION 44

45 MATCH advanced syntax in Neo4j Cypher Length paths MATCH (martin { name: 'Charlie Sheen' })- [:ACTED_IN*1..3]-(movie:Movie) Unnamed paths, here even zero length MATCH (wallstreet:movie { title: 'Wall Street' })-[*0..1]-(x) 45

46 MATCH advanced syntax in Neo4j Cypher Named paths MATCH p =(michael { name: 'Michael Douglas' })-->() Shortest path MATCH (martin:person { name: 'Martin Sheen' }),(oliver:person { name: 'Oliver Stone' }), p = shortestpath((martin)-[*]-(oliver)) RETURN p 46

47 That s all folks! 47