Temporal Databases (1)

8/17/2019 Temporal Databases (1)

1/23

From Handbook of TemporalReasoning in Articial Intelligence

By Jan Chomicki & David Toman

Temporal Databases

Presented by Leila Jalali

CS224 presentation


2/23

Temporal databases

Some data may be inherently historicale.g., medical or jdicial records

Temporal databases pro!ide a niform and systematic "ay ofdealing "ith historical data

#onsiderable e$ort has been e%pended on the de!elopment

of temporal databases and &ery langages T'el (Snodgrass)*+, TS'- (Snodgrass/+, S'0Temporal

(Snodgrass1+

─ 2t none of them has been adopted as the standard lan!ae o" temporaldatabases in practice

─ 3o established the theoretical "o!ndations for management of time4

dependent data ─ 3o ni!ersal consenss on ho# temporal features sho!ld be added to

the standard relational model


3/23

5tline

The fndamental notions of temporaldatabasesA formal fondation for temporal data models

Ho" to introdce time into the relational model

'ery langages for temporal databases Temporal e%tensions of S'

imitations of simple linearly4ordered, rst4

order temporal data models6ore comple% models of time


4/23

Strctre of time

They sed a !ery simple notion of time inthis chapter7a linear ordering of time instants

In addition to linear ordering, "e may

consider7Discrete or dense2onded or nbonded

Single dimensional or mlti4dimensional

inear or non4linear


5/23

The time stamp modelAll the tples in a relation ha!e an additional

temporal attribte8%ample7 Bookin $meetin% room% timeA tple 9m,r,t: denotes the fact that7

meetin m is in room r at time t


6/23

Temporal attribteSinle'dimensional( temporal relations "ere

allo"ed only a single temporal attribte)!ltiple dimensional( "ith each tple in a

relation there can be more than one temporalattribte8%ample7 t"o kinds of time are stored7 the valid

time 9"hen a particlar tple is tre: and thetransaction time 9"hen the particlar tple "asinserted0deleted in the database:

*on'+*F( can be ;attened to obtain the


7/23

The snapshot model

Di$erent !ie" from the time stamp model 9of thesame data:


8/23

Relational database histories

A history o!er a database schema p and adata domain D is a se&ence H : (Do,..., Dn)of database instances sch that7

1. all the states Do , . . . , Dn share the same

schema p and the same data domain D2. Do is the initial instance of the database

3. Di reslts from applying an pdate to Di-1,for i > <

Do D1 D2

time

…

…


9/23

Temporal database design

Reconstrction of ?ensen@s formal frame"ork (?ensen1+2ased on the notion of temporal "!nctional dependency7

8%ample7 the temporal =D

means e!ery meeting is held in a single room at any gi!en

timeSe!eral ad!antages7 can se the classical notions of =D

inference, dependency closre, normal forms, mi% temporaland non4temporal =Ds

A temporal =D holds in a snapshot temporal relation

D2 if the 9classical: =D holds in e!ery snapshot ofD2


10/23

6ltiple dimensions

Ho" to e%press t"o temporal dimensions sing

temporal =D7valid time (VT)transaction time (TT)

kind of temporal =Ds7Transaction time:Valid time:Bitemporal:

8%ample7 means the record atany time of the room booked for a meeting at any time is ni&elydetermined

Disad!antage7#an no longer talk abot, e.g., temporal keys, bt only abot !alid4time,

transaction4time or bitemporal keys The frame"ork becomes so complicated that it is nlikely to be of any

se


11/23

Temporal &eries

Databases are inherently rst4order strctres Temporal e%tensions rst4order logic

'ery7 sing a natral rst4order &erylangage

The ans"er7 the set of tple that make the&ery tre in the gi!en relational database

8%amples7

fnd all meetings that always meet in the sameroom

fnd all rooms in which the last meeting was D!gro"p


12/23

Temporal logic

Historically, many di$erent !ariants of temporallogic based on di$erent sets of connecti!esha!e been de!eloped (Babbay#C+

Some connecti!es are "ell4kno"n and ha!ebeen ni!ersally accepted7 sometime in the $"t"re

always in the $"t"re

In general any appropriate rst4order formla inthe langage of the temporal domain can besed to dene a temporal connecti!e


13/23

=irst order temporal logic

=irst they dene the rst order langage of Tp e%tended "ith propositional !ariables %i 7

Then se it to dene a 9&-ary : temporal

connective(an 54formla "ith e%actly one free !ariable t' and &

propositional !ariables %1,.., %& They assme ti is the only temporal !ariable in the

formla to be sbstitted for %i8%ample7 common binary temporal connecti!es7


14/23

Temporal connecti!es

5ther temporal connecti!es7Sometime in the ftre7

Sometime in the past7Al"ays in the ftre7Al"ays in the past73e%t7

re!ios7


15/23

=irst order temporal logic

7 A set of temporal connecti!es , e.g.Esince, ntilF

7 =irst order temporal logic 9=5T: o!er aschema

7


16/23

8%amples

Ho" to se temporal connecti!es to formlate&eries7

=ind all rooms in "hich the last meeting "as GD2grop7

=ind all meetings "ith a schedled break7

Sometime in the " !t!re(

Sometime in the past(

,l#ays in the "!t!re(

,l#ays in the past(


17/23

Temporal e%tensions of S'

A point based e%tension of S'7 S'0T(Toman*+

The synta% and semantics of S'0T aredened as a natral e%tension of S'An additional data type based on the point4based

temporal domain Tp 9i.e., a linearly ordered set oftime instants:


18/23

S'0T 8%ample

ist all meetings "ith a schedled break 7


19/23

8%tensions of S' based on inter!albased langage

TS'- or S'0Temporal (Snodgrass/+

Time attribtes range o!er inter!als and thebefore relationship denotes the before

relationship bet"een t"o inter!als


20/23

pdating temporal databases

Insertion7 a ne" booking for a room for ameeting

nit is an a%iliary table that contains a singletple

The inner &ery prodces7

Deletion7 #reating -J minte break in themiddle of meeting


21/23

#omple% strctre of time

#omple% strctre of time7 more comple% thanlinearly ordered sets of time instants3atral nmbers, integers, reals

Additional strctres7 drations, temporal

distances, periodic setsImpact on integrity constraints 7 more comple%

constraint dependencies

Impact on &ery langages 9se ne" predicatesymbols in the same "ay the linear order Ksymbol has been sed so far:


22/23

2eyondL

2eyond the rst order logic6ore e%pressi!e po"er than rst order logic

2eyond the #losed Morld Assmption#losed Morld Assmption7 temporal databases

hold complete information abot trthAn alternati!e7 to treat the relational strctres

representing temporal databases as incompletespecications and se the 5pen Morld

Assmption to ans"er &eries


23/23

#onclsion

-ss!es not Covered in the chapter(#onceptal 6odeling of Temporal Data

hysical Design for Temporal Databases Time Series and Temporal Data 6ining

.esearch directions that are closelyrelated to temporal data manaement(Spatio4Temporal Databases

Streaming Data 6anagement Time in Docment 6anagement and N6

Temporal Databases (1)

Documents