1. The Structured Query Language Zachary G. Ives University of Pennsylvania CIS 550 – Database & Information Systems September 27, 2005 Some slide content courtesy of Susan Davidson & Raghu Ramakrishnan

2. 2 Administrivia  Homework 2 handed out today  Due in 1 week, 10/4  Please sign up for Oracle accounts ASAP  http://www.seas.upenn.edu/ora/ http://www.seas.upenn.edu/ora/  You can do the homework without this, but:  Will be useful in testing your HW2’s!  You’ll need to get familiar with Oracle anyway

3. 3 Recall Basic SQL SELECT [DISTINCT] {T 1.attrib, …, T 2.attrib} FROM {relation} T 1, {relation} T 2, … WHERE {predicates}  SELECT *  All STUDENTs  AS  As a “range variable” (tuple variable): optional  As an attribute rename operator select-list from-list qualification

4. 4 Expressions in SQL  Can do computation over scalars (int, real or string) in the select-list or the qualification  Show all student IDs decremented by 1  Strings:  Fixed (CHAR(x)) or variable length (VARCHAR(x))  Use single quotes: ’A string’  Special comparison operator: LIKE  Not equal: <>  Typecasting:  CAST(S.sid AS VARCHAR(255))

5. 5 Set Operations  Set operations default to set semantics, not bag semantics: (SELECT … FROM … WHERE …) {op} (SELECT … FROM … WHERE …)  Where op is one of:  UNION  INTERSECT, MINUS/EXCEPT (many DBs don’t support these last ones!)  Bag semantics: ALL

6. 6 Exercise  Find all students who have taken DB but not AI  Hint: use EXCEPT

7. 7 Revised Example Data Instance sidname 1Jill 2Qun 3Nitin 4Marty fidname 1Ives 2Saul 8Martin sidexp-gradecid 1A550-0105 1A700-1005 3A 3C501-0105 4C cidsubjsem 550-0105DBF05 700-1005AIS05 501-0105ArchF05 555-1006SysS06 fidcid 1550-0105 2700-1005 8501-0105 STUDENT Takes COURSE PROFESSOR Teaches

8. 8 Nested Queries in SQL  Simplest: IN/NOT IN  Example: Students who have taken subjects that have (at any point) been taught by Martin

9. 9 Correlated Subqueries  Most common: EXISTS/NOT EXISTS  Find all students who have taken DB but not AI

10. 10 Universal and Existential Quantification  Generally used with subqueries:  {op} ANY, {op} ALL  Find the students with the best expected grades

11. 11 Table Expressions  Can substitute a subquery for any relation in the FROM clause: SELECT S.sid FROM (SELECT sid FROM STUDENT WHERE sid = 5) S WHERE S.sid = 4 Notice that we can actually simplify this query! What is this equivalent to?

12. 12 Aggregation  GROUP BY SELECT {group-attribs}, {aggregate-operator}(attrib) FROM {relation} T 1, {relation} T 2, … WHERE {predicates} GROUP BY {group-list}  Aggregate operators  AVG, COUNT, SUM, MAX, MIN  DISTINCT keyword for AVG, COUNT, SUM

13. 13 Some Examples  Number of students in each course offering  Number of different grades expected for each course offering  Number of (distinct) students taking AI courses

14. 14 Data Instance, Again sidname 1Jill 2Qun 3Nitin 4Marty fidname 1Ives 2Saul 8Martin sidexp-gradecid 1A550-0105 1A700-1005 3A 3C501-0105 4C cidsubjsem 550-0105DBF05 700-1005AIS05 501-0105ArchF05 555-1006SysS06 fidcid 1550-0105 2700-1005 8501-0105 STUDENT Takes COURSE PROFESSOR Teaches

15. 15 What If You Want to Only Show Some Groups?  The HAVING clause lets you do a selection based on an aggregate (there must be 1 value per group): SELECT C.subj, COUNT(S.sid) FROM STUDENT S, Takes T, COURSE C WHERE S.sid = T.sid AND T.cid = C.cid GROUP BY subj HAVING COUNT(S.sid) > 5  Exercise: For each subject taught by at least two professors, list the minimum expected grade

16. 16 Aggregation and Table Expressions (aka Derived Relations)  Sometimes need to compute results over the results of a previous aggregation: SELECT subj, AVG(size) FROM ( SELECT C.cid AS id, C.subj AS subj, COUNT(S.sid) AS size FROM STUDENT S, Takes T, COURSE C WHERE S.sid = T.sid AND T.cid = C.cid GROUP BY cid, subj) GROUP BY subj

17. 17 Thought Exercise…  Tables are great, but…  Not everyone is uniform – I may have a cell phone but not a fax  We may simply be missing certain information  We may be unsure about values  How do we handle these things?

18. 18 One Answer: Null Values  We designate a special “null” value to represent “unknown” or “N/A”  But a question: what does: do? NameHomeFax Sam123-4567NULL Li234-8972234-8766 Maria789-2312789-2121 SELECT * FROM CONTACT WHERE Fax < “789-1111”

19. 19 Three-State Logic  Need ways to evaluate boolean expressions and have the result be “unknown” (or T/F)  Need ways of composing these three-state expressions using AND, OR, NOT:  Can also test for null-ness: attr IS NULL, attr IS NOT NULL  Finally: need rules for arithmetic, aggregation T AND U = U F AND U = F U AND U = U T OR U = T F OR U = U U OR U = U NOT U = U

20. 20 Nulls and Joins  Sometimes need special variations of joins:  I want to see all courses and their students  … But what if there’s a course with no students?  Outer join:  Most common is left outer join: SELECT C.subj, C.cid, T.sid FROM COURSE C LEFT OUTER JOIN Takes T ON C.cid = T.cid WHERE …

21. 21 Data Instance, Again (!) sidname 1Jill 2Qun 3Nitin 4Marty fidname 1Ives 2Saul 8Martin sidexp-gradecid 1A550-0105 1A700-1005 3A 3C501-0105 4C cidsubjsem 550-0105DBF05 700-1005AIS05 501-0105ArchF05 555-1006SysS06 fidcid 1550-0105 2700-1005 8501-0105 STUDENT Takes COURSE PROFESSOR Teaches

22. 22 Warning on Outer Join  Oracle doesn’t support standard SQL syntax here: SELECT C.subj, C.cid, T.sid FROM COURSE C, Takes T WHERE C.cid =(+) T.cid

23. 23 Beyond Null  Can have much more complex ideas of incomplete or approximate information  Probabilistic models (tuple 80% likely to be an answer)  Naïve tables (can have variables instead of NULLs)  Conditional tables (tuple IF some condition holds)  … And what if you want “0 or more”?  In relational databases, create a new table and foreign key  But can have semistructured data (like XML)

24. 24 Modifying the Database: Inserting Data  Inserting a new literal tuple is easy, if wordy: INSERT INTO PROFESSOR (fid, name) VALUES (4, ‘Simpson’)  But we can also insert the results of a query! INSERT INTO PROFESSOR (fid, name) SELECT sid AS fid, name FROM STUDENT WHERE sid < 20

25. 25 Deleting Tuples  Deletion is a fairly simple operation: DELETE FROM STUDENT S WHERE S.sid < 25

26. 26 Updating Tuples  What kinds of updates might you want to do? UPDATE STUDENT S SET S.sid = 1 + S.sid, S.name = ‘Janet’ WHERE S.name = ‘Jane’

27. 27 Now, How Do I Talk to the DB?  Generally, apps are in a different (“host”) language with embedded SQL statements  Static: SQLJ, embedded SQL in C  Runtime: ODBC, JDBC, ADO, OLE DB, …  Typically, predefined mappings between host language types and SQL types (e.g., VARCHAR  string or char[])

28. 28 Embedded SQL in C EXEC SQL BEGIN DECLARE SECTION int sid; char name[20]; EXEC SQL END DECLARE SECTION … EXEC SQL INSERT INTO STUDENT VALUES (:sid, :name); EXEC SQL SELECT name, age INTO :sid, :name FROM STUDENT WHERE sid < 20

29. 29 The Impedance Mismatch and Cursors  SQL is set-oriented – it returns relations  There’s no relation type in most languages!  Solution: cursor that’s opened, read DECLARE sinfo CURSOR FOR SELECT sid, name FROM STUDENT … OPEN sinfo; while (…) { FETCH sinfo INTO :sid, :name … } CLOSE sinfo;

30. 30 JDBC: Dynamic SQL  Roughly speaking, a Java version of ODBC  See Chapter 6 of the text for more info import java.sql.*; Connection conn = DriverManager.getConnection(…); PreparedStatement stmt = conn.prepareStatement(“SELECT * FROM STUDENT”); … ResultSet rs = stmt.executeQuery(); while (rs.next()) { sid = rs.getInteger(1); … }

31. 31 Database-Backed Web Sites  We all know traditional static HTML web sites: Web-Browser HTTP-Request GET... Web-Server File-System Load File HTML-File

32. 32 Common Gateway Interface (CGI) Can have the web server invoke code (with parameters) to generate HTML Web Server HTTP-Request HTML-File Web Server File-System Load File File HTML? HTML Execute Program Program?Output I/O, Network, DB

33. 33 CGI: Discussion  Advantages:  Standardized: works for every web-server, browser  Flexible: Any language (C++, Perl, Java, …) can be used  Disadvantages:  Statelessness: query-by-query approach  Inefficient: new process forked for every request  Security: CGI programmer is responsible for security  Updates: To update layout, one has to be a programmer

34. 34 Java-Server-Process DB Access in Java Sybase Java Applet TCP/UDP IP Oracle... JDBC- Driver JDBC Driver manager Browser JVM

35. 35 Java Applets: Discussion  Advantages:  Can take advantage of client processing  Platform independent – assuming standard java  Disadvantages:  Requires JVM on client; self-contained  Inefficient: loading can take a long time...  Resource intensive: Client needs to be state of the art  Restrictive: can only connect to server where applet was loaded from (for security … can be configured)

36. 36 *SP Server Pages and Servlets (IIS, Tomcat, …) File-System Web Server HTTP Request HTML File Web Server Load File File HTML? HTML I/O, Network, DB Script/ Servlet? Output Server Extension May have a built- in VM (JVM, CLR)

37. 37 DB-Driven Web Server One Step Beyond: DB-Driven Web Sites (Strudel, Cocoon, …) Local Database HTTP Request HTML File Web Server Cache Data HTML Other data sources Script? Dynamic HTML Generation Styles

38. 38 Wrapping Up  We’ve seen how to query in SQL (DML)  Basic foundation is TRC-based  Subqueries and aggregation add extra power beyond *RC  Nulls and outer joins add flexibility of representation  We can update tables  We’ve also seen that SQL doesn’t precisely match standard host language semantics  Embedded SQL  Dynamic SQL  We’ve seen a hint of data-driven web site architectures