Introduction to Databases Exam

Transcription

1 Introduction to Databases Exam Werner Nutt 03/09/09 The exam comprises 5 questions some of which consist of several subquestions. There is a total of 60 points that can be achieved in this exam. You will have 2 hours time to answer the questions. Please, write down the answers to your questions in the exam booklet handed out to you. For drafts use the blank paper provided by the university. If the space in the booklet turns out to be insufficient, please use the university paper for additional answers and return them with the booklet. For students that submitted assignments for the labs the final mark for the course will be a weighted average of 70% exam mark + 30% lab mark or the exam mark, whichever is higher. For students that did not submit assignments the final mark will consist solely of the mark for the exam. 1

2 1 Relational Database Design A database needs to be developed that supports a supermarket. (i) Design an entity-relationship diagram that captures, as far as possible, the requirements stated below. (Statements in brackets are additional explanations and not part of the requirements proper.) If you make any assumptions in your design, please write them down. Assumptions, however, must not contradict the requirements. 1. The supermarket sells products. Each product is uniquely identified by its product number. Moreover, a product has price. The supermarket also records how many items of each product are currently on stock. 2. A product can be on special offer. A special offer is valid for a period, which has a start date and an end date. A special offer refers to exactly one product. Two special offers for the same product cannot have overlapping periods. A special offer for a product specifies that a certain number of items of that product are sold at a certain price (e.g., 2 items for 1,99 EUR). 3. There is a card scheme for which customers can register. A registered customer holds a customer card. Each registered customer has a customer number, which is unique. For each registered customer the supermarket records the name and the address of the customer and the day when the customer joined the card scheme. 4. When a customer buys one or more items and pays for them, a purchase is being completed. A purchase is uniquely identified by the number of the checkout where the customer pays, together with the date and the time when the purchase is made. 5. Some purchases are made by registered customers. 6. A purchase consists of one or more line items. Within a given purchase, each line item is uniquely identified by its line number. Each line item refers to a unique product. (In fact, the line items correspond to the items of the purchase). Some line items have an additional comment, which is a string. (The comment may say that a special price applies to the product item because it is part of a special offer.) (10 Points) (ii) Based on the ER-diagram from above, develop a corresponding relational database schema. 2

3 List tables with their attributes. For each table, underline the attributes that make up the primary key. Identify referential integrity constraints (i.e., foreign keys). (5 Points) 2 Relational Algebra Queries The database of a research center contains the following three tables about employees, projects, and the time spent by the employees on the projects. Employee(ssn: int, name: string, jobtitle: string) Project(pid: int, name: string, sponsor: string, startyear: int, endyear: int) WorkedOn(ssn: int, pid: int, year: int, month: int, nohours: int). The table Employee lists all the employees of the centre. The table Project lists all the projects of the centre with their sponsor and the start and end year of the project. The table WorkedOn records how many hours the employees have spent on which project in which month. For each table, the attributes that make up the primary key are underlined. Express each of the following queries in relational algebra. (i) Return the names of the projects that were active in (2 Points) (ii) Return the names of those programmers who in some month spent more than 60 hours on a project sponsored by the EU. (4 Points) (iii) Return the names of those programmers who never worked on a project sponsored by the EU. 3 SQL Queries Consider the relational schema of Question 2. Write SQL queries over this schema that answer the following questions. 3

4 (i) How many projects that were active in 2008 were sponsored by the EU? (2 Points) (ii) For each project, year, and month, how many hours of work have been spent? (Return only data for a project, year and month if some time has been spent.) (iii) How many programmers are there who have experience in working on a project sponsored by the EU? (Note that a programmer who worked on two or more projects should be counted only once.) (4 Points) (iv) Return the names of the programmers who worked on no more than two projects sponsored by the EU. (Note that this includes the programmers who never worked on any project sponsored by the EU.) (v) Which programmer(s) spent the maximal total number of hours on EU projects among all programmers working on EU projects? 4 Query Execution (5 Points) We consider a database with the employee and project schema of Question 2: Employee(ssn: int, name: string, jobtitle: string) Project(pid: int, name: string, sponsor: string, startyear: int, endyear: int) WorkedOn(ssn: int, pid: int, year: int, month: int, nohours: int). We assume that the database management system created B-tree indexes for the primary key of each relation. Suppose that queries of the following type are posed frequently SELECT s.name FROM Employee e, WorkedOn wo, Project p WHERE e.ssn = wo.ssn AND p.pid = wo.pid AND p.sponsor = EU AND 4

5 p.startyear <= 2008 AND p.endyear >= 2009 AND e.jobtitle = Programmer ; where the values of sponsor, startyear, endyear, and jobtitle are varying. (i) If you could create as many indexes as you liked to speed up the execution of this query, which ones would you choose? (Note that it is sufficient to create indexes for a single execution plan.) (ii) Explain how those indexes would support the execution of the query. Draw a relational algebra tree to represent the optimal execution plan. Annotate each node with the algorithm by which the operation of the node is executed (for example, index lookup or sort merge join ). (5 Points) (iii) If you could only create a single index, which one would you choose to obtain the greatest speed-up? State the assumptions under which that index would be the best one and explain why that choice would be better than other possible choices. 5 Normalization Consider the following relation that keeps track of the bookings in a hotel: Booking(guestID, guestname, creditcard, roomno, roomcat, from, to). Suppose the following functional dependencies hold on the relation: guestid guestname, creditcard roomno roomcat roomno, from guestid, to roomno, to guestid, from. (i) Decompose the relation in smaller relations such that each of the smaller relations is in BNCF with respect to the projection of the original dependencies; 5

6 the decomposition is a lossless join decomposition. (6 Points) (ii) Is your decomposition dependency preserving? If your answer is yes, argue why. If your answer is no, show which dependencies have been lost. (2 Points) 6