Lecture 7: Two-Table Queries

Transcription

1 Lecture 7: Two-Table Queries CS1106/CS5021/CS6503 Introduction to Relational Databases Dr Kieran T. Herley Department of Computer Science University College Cork KH (10/10/17) Lecture 7: Two-Table Queries / 1

2 Summary Two-table database designs. Cartesian product. Simple (inner) joins and self joins. KH (10/10/17) Lecture 7: Two-Table Queries / 1

3 Favourite Foods DB Revisited persons person id first name last name gender date of birth street town county Aoife Ahern F Brown Street Cork Cork Barry Barry M Green Street Tralee Kerry Ciara Callaghan F White Avenue Limerick Limerick Declan Duffy M Black Crescent Cork Cork Eimear Early F Red Square Thurles Tipperary Fionn Fitzgerald M Yellow Lane Bandon Cork favourite foods person id food Crisps Beer Nutella Chips Chocolate Ice Cream Chocolate Pizza Beer Crisps Pizza Chocolate Brussels Sprouts Ice Cream Nutella persons table captures individuals details favourite foods captures who likes what e.g (Fionn Fitzgerald) is partial to ice cream and Nutella KH (10/10/17) Lecture 7: Two-Table Queries / 1

4 Why Not Just Mash the Two Tables Together? Redundancy details) Anomalies persons and foods person id first name last name food Aoife Ahern Crisps Aoife Ahern Beer Aoife Ahern Nutella Barry Barry Chips Barry Barry Chocolate Fionn Fitzgerald Nutella Some information is replicated several times (e.g. Aoife s What is we wish to update Aoife s details e.g. change address Need to modify multiple rows otherwise table contains inconsistent information DB designers go to great length to avoid this sort of situation KH (10/10/17) Lecture 7: Two-Table Queries / 1

5 Queries That Span Tables persons person id first name last name gender date of birth street town county Aoife Ahern F Brown Street Cork Cork Barry Barry M Green Street Tralee Kerry Ciara Callaghan F White Avenue Limerick Limerick Declan Duffy M Black Crescent Cork Cork Eimear Early F Red Square Thurles Tipperary Fionn Fitzgerald M Yellow Lane Bandon Cork favourite foods person id food Crisps Beer Nutella Chips Chocolate Ice Cream Chocolate Pizza Beer Crisps Pizza Chocolate Brussels Sprouts Ice Cream Nutella What about queries that require info. from both tables? List the favourite foods of the person(s) named Aoife Ahern List all the persons who like pizza and beer KH (10/10/17) Lecture 7: Two-Table Queries / 1

6 Cartesian Product Cartesian Product The Cartesian product of sets S and T is the set of pairs of the form (s, t), where s S and t T. Example S = {2, 3, 5} T = {a, e, i, o, u} S T = { (2, a), (2, e), (2, i), (2, o), (2, u), (3, a), (3, e), (3, i), (3, o), (3, u), (5, a), (5, e), (5, i), (5, o), (5, u) } Note size of S T is the size of S multiplied by the size of T KH (10/10/17) Lecture 7: Two-Table Queries / 1

7 Cartesian Product In SQL Can use CROSS JOIN to form Cartesian product SELECT FROM persons CROSS JOIN favourite foods; person id first name last name person id food Aoife Ahern Crisps Barry Barry Crisps Ciara Callaghan Crisps Declan Duffy Crisps Eimear Early Crisps Fionn Fitzgerald Crisps Aoife Ahern Beer Barry Barry Beer Fionn Fitzgerald Nutella # rows() = # rows(persons) # rows(favourite fo # cols() = # cols(persons)+ # cols(favourite foo KH (10/10/17) Lecture 7: Two-Table Queries / 1

8 Cartesian Product In SQL cont d Most rows in cross join are meaningless person id first name last name person id food Aoife Ahern Crisps Barry Barry Crisps Ciara Callaghan Crisps Declan Duffy Crisps Eimear Early Crisps Fionn Fitzgerald Crisps Aoife Ahern Beer Barry Barry Beer Fionn Fitzgerald Nutella KH (10/10/17) Lecture 7: Two-Table Queries / 1

9 Cartesian Product In SQL cont d Most rows in cross join are meaningless, but not all... person id first name last name person id food Aoife Ahern Crisps Barry Barry Crisps Ciara Callaghan Crisps Declan Duffy Crisps Eimear Early Crisps Fionn Fitzgerald Crisps Aoife Ahern Beer Barry Barry Beer Fionn Fitzgerald Nutella KH (10/10/17) Lecture 7: Two-Table Queries / 1

10 Cartesian Product In SQL cont d On closer inspection some rows in the product are potentially meaningful person id first name last name person id food Aoife Ahern Crisps Barry Barry Crisps Ciara Callaghan Crisps Declan Duffy Crisps Eimear Early Crisps Fionn Fitzgerald Crisps Aoife Ahern Beer Barry Barry Beer Fionn Fitzgerald Nutella The highlighted rows are formed by coupling s (i.e. Aoife s) row from the persons table s rows from the favourite foods table These rows collectively contain information on Aoife s weaknesses If only we could filter out the useless rows from the product and retain only the useful ones KH (10/10/17) Lecture 7: Two-Table Queries / 1..

11 Joins In SQL Consider the familiar SELECT-FROM template SELECT FROM X ; here X specifies a table to which query is to be applied SELECT FROM persons JOIN favourite foods ON persons.person id = favourite foods. person id ; applies the SELECT-FROM to a following table (our X for this query) subset of Cartesian product of persons and favourite foods tables but includes only those (combined) rows where the person id value from the persons row and that of the favourite foods match Note use of persons.person_id notation to distinguish between two person id columns in product KH (10/10/17) Lecture 7: Two-Table Queries / 1

12 Joins In SQL cont d for each row-pair (alpha, beta) from persons X favourite_foods do if alpha.person_id and beta.person_id are equal then Include (alpha, beta) in result KH (10/10/17) Lecture 7: Two-Table Queries / 1

13 Back to Our Query SELECT FROM persons JOIN favourite foods ON persons.person id = favourite foods. person id ; person id first name last name person id food Aoife Ahern Crisps Aoife Ahern Beer Aoife Ahern Nutella Barry Barry Chips Barry Barry Chocolate Ciara Callaghan Ice Cream Ciara Callaghan Chocolate Declan Duffy Pizza Declan Duffy Beer Declan Duffy Crisps Eimear Early Pizza Eimear Early Chocolate Eimear Early Brussels Sprouts Fionn Fitzgerald Ice Cream Fionn Fitzgerald Nutella Fifteen rows each formed by glueing together a row from favourite foods with the row from persons that corresponds to the individual whose food preference it is KH (10/10/17) Lecture 7: Two-Table Queries / 1

14 An Example List all the students who like pizza Can use WHERE clause to further filter results SELECT FROM persons JOIN favourite foods ON persons.person id = favourite foods. person id WHERE food = Pizza ; Can interpret as a standard SELECT-FROM-WHERE that is applied to the join of the two tables (on person id values) essentially result of previous query Yields following result person id first name last name person id food Declan Duffy Pizza Eimear Early Pizza KH (10/10/17) Lecture 7: Two-Table Queries / 1

15 A Tidier Version SELECT first name, last name, FROM persons AS p JOIN favourite foods AS f ON p.person id = f. person id WHERE food = pizza ; likes, food Clause persons AS p attaches shorter name (p) to table persons can use p instead of persons throughout query improves readability KH (10/10/17) Lecture 7: Two-Table Queries / 1

16 Reasoning About Join Queries Recall SELECT first name, last name, likes, food FROM persons AS p JOIN favourite foods AS f ON p. person id = f. person id WHERE food = Pizza ; Can be helpful to consider query execution in three stages 1 1 ( Create full Cartesian product C of tables persons and favourite foods) 2 ( Filter C to retain (in C ) only those rows in C that satisfy the ON condition) 3 Filter C to retain (in R) only those rows in C that satisfy the WHERE condition and return R as the result of the query 1 Conceptual model only; actual mechanics of query execurion may follow different approach KH (10/10/17) Lecture 7: Two-Table Queries / 1

17 Example 3 List all pairs of students who hail from the same county (not the same as all those from a specific county) Use just the persons table, but forms join of table with itself (two copies )! KH (10/10/17) Lecture 7: Two-Table Queries / 1

18 Example 3 List all pairs of students who hail from the same county (not the same as all those from a specific county) Use just the persons table, but forms join of table with itself (two copies )! First stab: SELECT p1. first name, p1.last name, and, p2. first name, p2.last name, both come from, p1.county FROM persons AS p1 JOIN persons AS p2 ON p1.county = p2.county; Any problems with this? KH (10/10/17) Lecture 7: Two-Table Queries / 1

19 A Closer Look Query completion Typical row from Cartestian product p id county p id county x Cork y Kerry from persons 1 from persons 2 Typical row from product filtered by ON condition p id county p id county x Cork y Cork from persons 1 from persons 2 KH (10/10/17) Lecture 7: Two-Table Queries / 1

20 A Closer Look Query completion Typical row from Cartestian product p id county p id county x Cork y Kerry from persons 1 from persons 2 Typical row from product filtered by ON condition p id county p id county x Cork y Cork from persons 1 from persons 2 Slight problem Result may contain duplicates (Aoife-Declan and Declan-Aoife) p id county p id county x Cork y Cork y Cork x Cork from persons 1 from persons 2 KH (10/10/17) Lecture 7: Two-Table Queries / 1

21 A Closer Look Query completion Typical row from Cartestian product p id county p id county x Cork y Kerry from persons 1 from persons 2 Typical row from product filtered by ON condition p id county p id county x Cork y Cork from persons 1 from persons 2 Slight problem Result may contain duplicates (Aoife-Declan and Declan-Aoife) p id county p id county x Cork y Cork y Cork x Cork from persons 1 from persons 2 May also contain self-pairs (Aoife-Aoife) p id county p id county x Cork x Cork from persons 1 from persons 2 KH (10/10/17) Lecture 7: Two-Table Queries / 1

22 Our Query First attempt lists each pair twice (Aoife-Declan and Declan-Aoife) and also self pairs like (Aoife-Aoife) SELECT p1. first name, p1.last name, and, p2. first name, p2.last name, both come from, p1.county FROM persons AS p1 JOIN persons AS p2 ON p1.county = p2.county AND p1.person id < p2.person id; The < clause filters out duplicates and self pairs KH (10/10/17) Lecture 7: Two-Table Queries / 1

23 Our Query Again Could rephrase as follows: SELECT p1. first name, p1.last name, and, p2. first name, p2.last name, both come from, p1.county FROM persons AS p1 JOIN persons AS p2 ON p1.county = p2.county WHERE p1.person id < p2.person id; Bothe the ON-condition and WHERE-condition filter rows; so the < criterion can migrate into the latter KH (10/10/17) Lecture 7: Two-Table Queries / 1

24 Who Likes Pizza and Beer? List all those individuals who like pizza and beer Idea: three-way join (persons favourite foods favourite foods) with ON condition that ensures: three-way match on person id favourite foods.food (first copy) should equal Pizza favourite foods.food (second copy) should equal Beer KH (10/10/17) Lecture 7: Two-Table Queries / 1

25 The Query SELECT first name, last name, f1.food, f2.food FROM persons AS p JOIN favourite foods AS f1 JOIN favourite foods AS f2 ON p. person id = f1. person id AND f1.person id = f2.person id AND f1.food = Pizza AND f2.food = Beer ; KH (10/10/17) Lecture 7: Two-Table Queries / 1

26 Our Query Typical row from Cartesian product: p id p id food p id food x y Nutella z Crisps from persons from fav foods 1 from fav foods 2 No meaningful relationship among the fragments from the joinee tables KH (10/10/17) Lecture 7: Two-Table Queries / 1

27 Our Query Typical row from Cartesian product: p id p id food p id food x y Nutella z Crisps from persons from fav foods 1 from fav foods 2 No meaningful relationship among the fragments from the joinee tables Typical row product filtered by ON-condition: p id p id food p id food x x Pizza x Beer from persons from fav foods 1 from fav foods 2 Note person id (here abbrev. to p id) must match up and food values must be Pizza and Beer respectively KH (10/10/17) Lecture 7: Two-Table Queries / 1

28 Note SQL supports several types of join We have been using the INNER JOIN (the default) There are also OUTER, LEFT, RIGHT joins etc. which we may see later KH (10/10/17) Lecture 7: Two-Table Queries / 1

29 Notes and Acknowledgements KH (10/10/17) Lecture 7: Two-Table Queries / 1