SS 2017 (MA-INF 3302) Prof. Dr. Rainer Manthey Institute of Computer Science University of Bonn. Temporal Information Systems 1

Size: px

Start display at page:

Download "SS 2017 (MA-INF 3302) Prof. Dr. Rainer Manthey Institute of Computer Science University of Bonn. Temporal Information Systems 1"

Nathaniel Copeland
5 years ago
Views:

1 Temporal Information Systems (MA-INF 3302) SS 2017 Prof. Dr. Rainer Manthey Institute of Computer Science University of Bonn Temporal Information Systems 1

2 TIS: TIS: Topic Topic in in a a Nutshell Temporal : time-related (from lat. tempus : time ) In just one sentence, the topic of this lecture could be summarized as follows: How How to to properly manage manage and and?? Data concerned with Time Time concerned with Data Whereas the 1 st phrase ( Data concerned with Time ) may appear reasonable, the 2 nd phrase ( Time concerned with Data ) may be irritating however, it is meant exactly as stated! Temporal Information Systems 2

3 Organisation Temporal Information Systems SS 2017 Organisation Temporal Information Systems 3

4 Modules Offered Offered by by the the IDB IDB Group Group Intelligent Information Systems (MA-INF (MA-INF 3203) 3203) (9 CP) WS 2017/18 and WS 2018/19 IDB (Intelligent Databases) Group: Prof. Dr. Manthey, PD Dr. Behrend, MSc Vahdati next 4 semesters Seminar Seminar Selected Topics Topics in in Intelligent IS IS (MA-INF 3210) (MA-INF 3210) (4 CP) Temporal Information Systems (MA-INF (MA-INF 3302) 3302) (6 CP) Scientific Communication and Scientific Writing (4 CP) (MA-INF (MA-INF 3107) 3107) SS 2018 and SS 2019 Temporal Information Systems 4

5 TIS TIS Weekly Weekly Schedule 10:00 11:00 (nearly) every Wednesday during this semester 10:30 12:00 Lecture Temporal Information Systems (MA-INF (MA-INF 3302) 3302) 12:00 Long break (45 mins) 13:00 14:00 12:45 14:15 Exercises (both in A 207) Temporal Information Systems 5

6 Calendar of of this this Semester Relevant for exam Dies Dies academicus Whitsun week week Exam (written): August 3, 2017 (most likely) 2 nd exam: September 20, 2018 (most likely) L L+E L+E L+E L+E L+E L+E L+E L+E L+E L+E L (+E) 12 L+11 E Temporal Information Systems 6

Lecture Lecture Homepage https://pages.iai.uni-bonn.

7 Lecture Lecture Homepage Slides Slides for for download download Temporal Information Systems 7

We don t control attendance it is up to you if you make use of this offer or not.

8 Exercises and and Admission to to Exams Exams Exercises will be offered as a regular opportunity to improve your understanding of the topics covered and to train for the final exam online during the semester. We don t control attendance it is up to you if you make use of this offer or not. There is no further study achievement required for being admitted to the exam! (Thus, no homework to be submitted.) Temporal Information Systems 88

of Arizona) Freely downloadable as PDF file from: http://www.cs.arizona.edu/people/rts/tdbbook.

9 Main Main Source Source for for Additional Reading: The The Snodgrass Book Book Probably the most active and prominent researcher in temporal databases: Richard T. Snodgrass (Univ. of Arizona) Freely downloadable as PDF file from: originally published by: Morgan Morgan Kaufman, Kaufman, 1999, 1999, pp., pp., ISBN ISBN Temporal Information Systems 9

A Different Approach Written Written by by a a Relational DB DB Guru Guru Chris Chris J. J. Date, Date, Hugh Hugh Darwen, Nikos Nikos A. A. Lorentzos: Time Time and and Relational Theory Morgan Morgan Kaufman Publ.

10 A Different Approach Written Written by by a a Relational DB DB Guru Guru Chris Chris J. J. Date, Date, Hugh Hugh Darwen, Nikos Nikos A. A. Lorentzos: Time Time and and Relational Theory Morgan Morgan Kaufman Publ., Publ., 2 nd nd edition edition (2014), (2014), pp., ISBN ISBN (~ (~ ) ) Temporal Information Systems 10

11 Prerequisites Attending TIS doesn t require much of a specific background; thus, it is even well-suited for beginners in our master programme. The only prerequisite expected is that you have a good basic understanding of relational databases and some solid basic skills in SQL programming. In SQL, you should master the most common DDL commands (CREATE TABLE, CREATE VIEW) including various forms of integrity constraints (e.g., KEY, CHECK, REFERENCES) the main concepts of the query sublanguage of SQL (SELECT-FROM-WHERE incl. AS, JOIN/UNION/MINUS, AND/OR/NOT, EXISTS, GROUP BY/HAVING, aggregate functions), the basic data modification commands incl. transactions and integrity checking (INSERT, DELETE, UDPATE; COMMIT). Some of the more specific aspects of SQL will be briefly revisited on the first exercise date on May 3 (constraints incl. assertions, expressing forall conditions in terms of NOT EXISTS etc.), but it is up to you to make sure that you are sufficiently fit in SQL by the end of May. Temporal Information Systems 11

12 Motivation Temporal Information Systems SS 2017 Motivation Temporal Information Systems 12

13 Information system: system: The The DB-centered View View Information System + External media of communication + Application- specific methods Database System This This is is the the most most commonly commonly agreed agreed view view on on the the concept concept of of an an IS IS in ininformatics informatics provided provided people people agree agree on on the the meaning meaning of of DBS!! DBS!! Temporal Information Systems 13

14 Database System: System: A Computer Science Science Notion Notion Database System DBMS DB.... Databases Users and application programmes DBMS: Data Base Management System (Many powerful application-independent independent services: schema mgt, query optimization, storage mgt, transaction mgt, etc.) Temporal Information Systems 14

15 Temporal IS IS vs. vs. Temporal DB DB The The modifier temporal is is used used to to indicate indicate that that the the modified concept concept concerns some some aspect aspect of of time. time. (A Consensus Glossary of Temporal Database Concepts, 1993) The lecture is called Temporal Information Systems, because it thus refers to the more general concept (IS) and therefore can be used in a wider sense, and this is in line with other namings in our MSc programme, e.g., Intelligent IS. The term nearly exclusively used in scientific literature (in CS) is the related notion Temporal Databases even though literally speaking this is too limited! Certainly managing temporal data requires involving the DBMS as well, therefore the term Temporal Database Systems would be more appropriate than just TDB. In In this this lecture, lecture, we we will will focus focus on on temporal database research, which which actually actually means means research on on temporal database systems! Temporal Information Systems 15

16 Temporal Database: Attempts of of Definition (1) (1) Databases which which store store all all the the past past history history as as is is best best known known at at every eevery state state of of the the database are are called called temporal databases. (Sripada, 1988) A A A database that that maintains past, past, present, present, and and future future data data is is called called a a temporal database. (Tansel, 1993) A A temporal database is is a a repository of of temporal information. (Chomicki, 1995) A A temporal database is is one one that that supports some some aspect aspect of of time. time. (Snodgrass, 1995) A A temporal database can can be be thought thought of, of, very very loosely, loosely, as as a a database that that contains historical data data instead instead of of or or in in addition addition to to current current data. data. (Date/Darwen/Lorentzos, 2002) Temporal Information Systems 16

17 Temporal Database: Attempts of of Definition (2) (2) Even now, ~20 years later, there is no agreement on an exact definition of the term among scientists the spectrum ranges from very liberal to rather strict claims: wide A A temporal database is is one one that that supports some some aspect aspect of of time. time. (Snodgrass, 1995) narrow Databases which which store store all all the the past past history history as as is is best best known known at at every eevery state state of of the the database are are called called temporal databases. (Sripada, 1988) In our lecture, we will start from the most liberal definition and first try to understand what the more restrictive approaches are really talking about. Temporal Information Systems 17

18 Why Why Temporal Databases? (1) (1) Most Most applications of of database technology are are temporal in in nature. nature. Examples include include financial applications such such as as portfolio management, accounting, and and banking; record-keeping applications such such as as personnel, medical medical record, record, and and inventory management; scheduling applications such such as as airline, airline, train, train, and and hotel hotel reservations and and project project management; nt; and and scientific applications such such as as weather weather monitoring. Applications such such as as these these rely rely on on temporal databases,, which which record record time- time- referenced data. data. (Christian Jensen in Introduction to Temporal Database Research ) Temporal Information Systems 18

19 Why Why Temporal Databases? (2) (2) Two Two decades decades of of research into into temporal databases have have unequivocally ly shown shown that that a a time-varying table, table,, containing certain certain kinds kinds of of DATE DATE columns, is is a a completely different animal animalthan its its cousin, cousin, the the table table without without such such columns. (R. Snodgrass in his book, in 1999) Despite the the near near universality of of time time and and the the time-varying nature nature of of the the enterprise being being modeled a a static static and and unmalleable configuration is is rare rare and and uninteresting SQL SQL quite quite frankly frankly does does a a lousy lousy job jobin in capturing those those aspects aspects that that are are changing in in time, time, or or in in providing constructs to to effectively model, model, query, query, or or modify modify such such information. (R. Snodgrass again) Temporal Information Systems 19

20 Time Time about about Data Data vs. vs. Data Data about about Time: Time: An An Intuitive Introduction Data concerned with Time Time concerned with Data Temporal Information Systems 20

21 1 st 1 st st Example Database: History History of of American Presidents Presidents of the USA Presidency President Barack Obama Barack Obama Birthday From To Term 1 George Washington Grover Cleveland Benjamin Harrison Grover Cleveland Donald Trump What do these time-valued attributes mean? What do they refer to? Do all three time columns play the same role? Temporal Information Systems 21

22 1 st 1 st st Example Database: Meaning of of Temporal Values Values (1) (1) Presidents of the USA Presidency President Barack Obama Barack Obama Donald Trump Birthday From To Term 1 George Washington Grover Cleveland Barack Obama 23 was Benjamin born! Harrison Grover Cleveland st president of the US was inaugurated! Each of the dates appearing in a field of this table refers a particular (past) event that has happened in reality. Obama leaves office! An event that did not yet happen, but certainly will do one day! Temporal Information Systems 22

23 1 st 1 st st Example Database: Meaning of of Temporal Values Values (2) (2) Presidents of the USA Presidency 44 President Barack Obama Birthday From To Term 1 George Washington Cleveland serves as 22 nd president of the US! 22 Grover Cleveland Benjamin Harrison Grover Cleveland Barack Obama Donald Trump Entries in colums From and To together in addition represent time periods during which a particular fact was true in reality! (How does the Birthday column fit with this view?) Temporal Information Systems 23

24 1 st 1 st st Example DB: DB: How How to to Represent Now Now and and Future? Future? Presidency 44 President Barack Obama Birthday From To Term 1 George Washington Grover Cleveland Benjamin Harrison Grover Cleveland Barack Obama Donald Trump ? 57 What does the empty To-field for Trump mean? Should there better be a concrete entry instead? If so, which one: The current date (changing every midnight), or the scheduled end of term ( )? Temporal Information Systems 24

25 1 st 1 st st Example DB: DB: An An Alternative Design Design Any Any Better? Better? Presidency 44 President Barack Obama Birthday From To Term 1 George Washington Grover Cleveland Benjamin Harrison Grover Cleveland Since 44 Barack Obama Donald Trump If we mean until now in the Trump case, dropping To altogether and renaming From into Since might be the better choice if we can rely on gap-free presidencies (which we can, up till now)! Temporal Information Systems 25

26 1 st 1 st st Example DB: DB: Queries Queries Referring to to Past, Past, Present Present and and Future Future Presidency 44 President Barack Obama Birthday From To Term 1 George Washington Grover Cleveland Benjamin Harrison Grover Cleveland Barack Obama Donald Trump How to express these queries in SQL? What answers to expect? Who Who is isthe current current president? Who Who was waspresident on on July July 4, 4, 1890? 1890? Who Who will will be bepresident on on July July 4, 4, 2017? 2017? Temporal Information Systems 26

27 1 st 1 st st Example DB: DB: Same Same Queries Queries Different Design Design Presidency President Barack Obama Barack Obama Donald Trump Birthday Since Term 1 George Washington Grover Cleveland Benjamin Harrison Grover Cleveland Is a different color still appropriate? How to express these queries for the other design? Who Who is isthe current current president? Who Who was waspresident on on July July 4, 4, 1890? 1890? Who Who will will be bepresident on on July July 4, 4, 2015? 2015? Temporal Information Systems 27

28 1 st 1 st st Example DB: DB: Two Two Tables Tables Lifetime of of Presidents Included Presidency President From To Term 1 George Washington Abraham Lincoln John F. Kennedy Richard Nixon President Born Died 44 Barack Obama George Washington Which Which presidents died died in in office? office? Which Which presidents resigned before before the the regular regular end end of of their their term term (alive)? (alive)? Who Who was was the the oldest oldest president at at the the begin begin of of his his presidency? How How many many former former presidents are are still still alive alive today? today? Abraham Lincoln John F. Kennedy Richard Nixon Presideny Barack Obama Lifetime Temporal Information Systems 28

29 1 st 1 st st Example Database: More More Non-Obvious Queries Queries Presidency President Birthday From To Term 1 George Washington Grover Cleveland Benjamin Harrison Grover Cleveland George W. Bush George W. Bush Which Which period periodis is covered covered by by the the rd rd presidency? How How long longdid did the the rd rd presidency last? last? For For how how many many years years did did Cleveland serve serve as as president? Which Which presidents served served during during different presidencies? Temporal Information Systems 29

30 1 st 1 st st Example Database: A Purely Purely Historical Table Table fixed forever Presidents of the USA Presidency 44 President Barack Obama Birthday From To Term 1 George Washington Grover Cleveland Benjamin Harrison Grover Cleveland George W. Bush George W. Bush Barack Obama The facts recorded in this table are (presumably) all settled, known to be historically correct, never to be reconsidered. The only modifications to be expected are appending new rows and filling in missing To-dates. 56 Temporal Information Systems 30

31 The The First First Example: Managing Data Data about about Time Time (History) How How to to properly manage manage and and?? Data concerned with Time Time concerned with Data Temporal Information Systems 31

32 2 nd 2 nd nd Example Database : : Classes Classes and and Exams Exams in in a a University Exams Student Class Signed_up Dropped Grade Exam Date John , Jack Tim , Pete , John , Jack Tim , Assume, this is how the Exams table looked like on Sunday, April 1, At first glance, this looks like another settled history table, but Temporal Information Systems 32

33 2 nd 2 nd nd Example DB: DB: Potential Mistakes in in Exam Exam DB DB Student Class Signed_up Dropped Grade Exam Date John , Jack Tim , Pete , John , Jack Tim , Sign-up date missing? Failed or dropped? Mistyped? There There are are several several suspicious entries entries in in this this table! table! Temporal Information Systems 33

34 2 nd 2 nd nd Example DB: DB: Typo Typo Corrected History History Cleaned Cleaned Up Up On April 4th, 2011, the typo of Tim s exam date is discovered and corrected: Student Class Signed_up Dropped Grade Exam Date John , Jack Tim , Pete , John , Jack Tim , So this is the version of history as known on April 4th, Temporal Information Systems 34

35 2 nd 2 nd nd Example DB: DB: Failed Failed Exam Exam Discovered as as a a Mistake Mistake On April 6th, 2011, it turns out that Pete didn t appear for the exams in class 1203 because he had indeed dropped the course but the examiner didn t know: Student Class Signed_up Dropped Grade Exam Date John , Jack Tim , Pete , John , Jack Tim , This is the next version of history as known on April 6th, Temporal Information Systems 35

36 2 nd 2 nd nd Example DB: DB: Missing Missing Sign-up Date Date Discovered and and Added Added On April 7th, 2011, Jack was asked when he actually signed-up for It turned out that Jack never signed up for this class, but erroneously entered a drop date for that course (and couldn t get rid of it anymore): Student Class Signed_up Dropped Grade Exam Date John , Jack Tim , Pete John , Jack Tim , After deleting this error, we have a version as known on April 7th, Temporal Information Systems 36

37 2 nd 2 nd nd Example DB: DB: Missing Missing Information About About Improved Grade Grade Finally, on April 8th, 2011, Tim complains about the grade he found in the DB for It had been improved to 2,7 by the examiner after reconsidering the evaluation of one answer: Student Class Signed_up Dropped Grade Exam Date John , Jack Tim , Pete John , Tim , Since April 8th, 2011, this is the version stored in the DB it seems to be correctly reflecting what happened in reality. Temporal Information Systems 37

38 2 nd 2 nd nd Example DB: DB: A Solution Keeping Track Track of of All All Changes Student Class Signed_up Dropped Grade Exam Date From To John John , Jack Jack Tim Tim , Tim , Pete Pete Pete , Pete John John , Jack Tim , Tim , Temporal Information Systems 38

39 Keeping Track Track of of All All Changes? There is an increasing number of real-life life applications where keeping track of all changes ever performed to the real database is useful, sometimes even required by law (auditing, liability, etc.). Transaction processing, in particular recovery, already requires to track changes made during a transaction at least temporarily: transaction log. Keeping track of changes forever is a different affair, but it is a form of logging, too. However, it is not clear whether this kind of history log should be part of the table to be logged itself: Should the history log be changeable by users, or by the system only? Should users be able to query the log? Which tables should be logged at all? Recently, big vendors of relational DBMS started to include system-controlled history keeping into their systems (e.g., Total Recall of Oracle 11g). Temporal Information Systems 39

40 On On Logs Logs Log: : originally part of the stem of a tree (The log lady from the TV series Twin Peaks ) Later: Tally stick,, used for recording consumption Even later: Ship s logbook, used for recording changes of direction and rele- vant events of a trip Temporal Information Systems 40

41 Second Second Example: Managing Time Time About About (the (the History History of of Changes of) of) Data Data How How to to properly manage manage and and?? Data concerned with Time Time concerned with Data Temporal Information Systems 41

42 Some Some of of the the Problems to to be be Addressed in in this this Lecture Lecture What What is is the the correct correct meaning of of temporal data data wrt wrt real real world? world? How How to to model modeltemporal aspects aspects of of change change appropriately? How How to to express expresstemporal queries queriesadequately? How How to to design design good good query query languages supporting time? time? How How to to evaluate evaluatetemporal queries queriesefficiently? How How to to update updatetemporal data data correctly and and consistently? How How to to reason reasonabout about temporal data? data? Temporal Information Systems 42

43 Outline Outline of of the the Lecture Lecture Last time (in 2015), TIS covered the following chapters it is likely to be structured like this again (though some evolution is still possible): 0 Organisation and and Motivation 1 Some Some General General Reflections About About Time Time 2 Representing Time Time in in SQL SQL 3 Recording Time Time About About Data Data Keeping a a History History of of Change Change 4 Managing Data Data About About Time Time In In Combination with with History History Keeping 5 Temporal Perspectives for for SQL SQL Temporal Information Systems 43

44 Next week! Some General Reflections About Time Chapter 1 Temporal Information Systems 44

Database in Depth: Relational Theory for Practitioners

Database in Depth: Relational Theory for Practitioners a technical seminar for DBAs, data architects, DBMS implementers, database application programmers, and other database professionals ABOUT THIS SEMINAR