3D_XML A three-Dimensional XML-based Model

1. 3D_XML A three-Dimensional XML-based Model Khadija Ali, Jaroslav Pokorný Czech Technical University Prague - Czech Republic

2. Content of Presentation • Current approaches • Motivation • 3D_XML Model • Time dimensions • Time model • Data modeling • Temporal constructs • Temporal Queries with XQuery • Conclusions and Future investigations.

3. Current Approaches Much research work has recently focused on adding temporal features to XML. • XBIT- an XML-based bitemporal data model. • XML-based model for Versioned Documents. • XML-based model for Versioned normative texts. • Valid time XPath data model. • Key-based model for archiving data. • Multidimensional XML Model (MXML).

4. Motivation Non-Temporal XML databases • The contents of XML documents change over time. • Past data are overwritten; they are replaced by the new data. • The latest information is recorded as a result. But,in the real life: there is a real need to keep past and current data.

5. Motivation For instance, at a bank, querying the current state is important, e.g., “How much is in my account right now“ But it is also often useful and required to know how the data has changed over time, e.g., “When has my account been below $4000“

6. 3D_XML Model The main aim of 3D_XML is to propose a temporal extension of XML by: • representing XML changes; we retain a single representation of all versions of the XML document. • using XQuery to express complex temporal queries on the evolution of the document contents via user-defined functions.

7. 3D_XML Model Time dimensions • Valid time • Transaction time • Efficacy time (abrogated data continues to be applicable) Our Extension Valid data stops to be applicable

8. 3D_XML Model Time Model • The time domain Tis linear and discrete. A time constant t = [a, b], is either a time instant or a time interval. • We limit our event measures to dates (granularity = one day). • The valid time constant vt = [vtStart, vtEnd] • The transaction time constant tt = [ttStart, ttEnd] • The efficacy time constant et = [etStart, etEnd]

9. 3D_XML Model Valid/Efficacy Times Relationship 1.(vtStart < etStart), or (vtEnd > etEnd) Valid data stops to be applicable

10. 3D_XML Model 2. (vtStart > etStart),or(vtEnd < etEnd) Data is applicable although it is not valid

11. 3D_XML Model 3. (vtStart = etStart), and (vtEnd = etEnd) The normal case

12. 3D_XML Model Example 1.Consider a company database. Suppose the manager mgr of Design department is Ali from "2002-01-01" till "2006-01-01". Due to some unexpected circumstances, another person started managing Design department from "2002-09-02". 2002-01-01 2006-01-01 Valid time 2002-09-02 2006-01-01 2002-01-01 2002-09-01 Efficacy time

13. Data Modeling A temporal element can be specified in DTD notation as: (1) <!ELEMENT ele (sub-element+)> <!ELEMENT sub-element Any> <!ATTLIST sub-elementvtStart CDATA #REQUIRED vtEnd CDATA #REQUIRED ttStart CDATA #REQUIRED ttEnd CDATA #REQUIRED etStart CDATA #REQUIRED etEnd CDATA #REQUIRED>

14. Data Modeling (2) <!ELEMENT ele (sub-element+)> <!ELEMENT sub-element Any> <!ATTLIST sub-element inherits CDATA #REQUIRED> Example 2 <employee vtStart="2000-01-01" vtEnd="now" ttStart="2000-01-01" ttEnd="now" etStart="2000-01-01" etEnd="now"> <name inherits = ”1”>Anas</name> <job inherits = ”1”>Sr Engineer</job> </employee>

15. Data Modeling A temporal attributeisrepresented by: • a special empty elementtemporal_Attribute. • the name and value of the temporal attribute are represented by special attributes name and value. • representing time dimensions is similar to time dimensions in temporal elements. <temporal_Attribute name="currency" value="crown" vtStart="2000-08-31" vtEnd="2014-12-31" ttStart="2000-09-01" ttEnd="now" etStart="2000-08-31" etEnd="2014-12-31“ />

16. Data Modeling Time dimensions Constraints • Conjunction Constraints • Disjunction Constraints

17. Data Modeling Conjunction Constraints <employee vtStart="2000-01-01" vtEnd="now" ttStart="2000-01-01" ttEnd="now" etStart="2000-01-01" etEnd="now"> <salary vtStart="2000-08-31" vtEnd="2005-09-01" ttStart="2000-08-31“ ttEnd="2005-09-30" etStart="2000-08-31" etEnd="2005-09-01"> 60000 <temporal_Attributename="currency" value="crown" vtStart="2000-08-31" vtEnd="2014-12-31" ttStart="2000-09-01" ttEnd="now" etStart="2000-08-31" etEnd="2014-12-31"/> <temporal_Attributename="currency" value="euro" vtStart= "2015-01-01" vtEnd="now" ttStart="2000-09-01" ttEnd="now" etStart="2015-01-01" etEnd="now"/> </salary> … </employee> Cover

18. Data Modeling Disjunction Constraints <employee vtStart="2000-01-01" vtEnd="now" ttStart="2000-01-01" ttEnd="now" etStart="2000-01-01" etEnd="now"> <salary vtStart="2000-08-31" vtEnd="2005-09-01" ttStart="2000-08-31“ ttEnd="2005-09-30" etStart="2000-08-31" etEnd="2005-09-01"> 60000 <temporal_Attributename="currency" value="crown" vtStart="2000-08-31" vtEnd="2014-12-31" ttStart="2000-09-01" ttEnd="now" etStart="2000-08-31" etEnd="2014-12-31"/> <temporal_Attributename="currency" value="euro" vtStart= "2015-01-01" vtEnd="now" ttStart="2000-09-01" ttEnd="now" etStart="2015-01-01" etEnd="now"/> </salary> … </employee> disjoint

19. Temporal Constructs XQuery is natively extensible and Turing-complete: • any extensions needed for temporal queries can be defined in the language . • user-defined functions which are not typical features of query languages can be used to define the temporal constructs.

20. Temporal Constructs Get time dimensions declare function get_vtStart($s) {let $g := string($s/@inherits) return if ($g ) then xs:date($s/ancestor::node()[$g]/@vtStart) else xs:date($s/@vtStart)};

21. Temporal Constructs Valid/efficacy times relationship constructs declare function valid-notEfficient($a) {if (get_etStart($a) > get_vtStart($a) or get_etEnd($a)< get_vtEnd($a) ) then true() else false()};

22. Temporal Constructs Snapshot data construct is used to construct snapshots of 3D-XML documents. declare function dataShot ($e, $v) {if (get_vtStart($e)<= xs:date($v) and get_vtEnd($e) >= xs:date($v) ) then element {name($e)} {$e/text(),$e/@* except $e/@* [string(name(.))="vtStart" or string(name(.))="vtEnd" or string(name(.))="etStart" or string(name(.))="etEnd" or string(name(.))="ttStart" or string(name(.))="ttEnd"], for $c in $e/* return dataShot ($c, $v)} else ()

23. Temporal Constructs Interval comparison operators: a small library of interval comparison operators is defined to help users with interval-based queries. Breakconstruct: it checks if valid/efficacy times of a temporal element having breaks or not. declare function Tbreak($g) {let $c := count($g) - 1 let $o := for $i in (1 to $c) let $j := $i +1 return if (TmeeTs(get_vtEnd($g[$i]), get_vtStart($g[$j]))) then() else ”break” return count ($o)};

24. Temporal Queries with XQuery • We use XQuery to express complex temporal queries. • The queries can be greatly simplified by a suitable library of built-in temporal functions (temporal constructs). Query 1 Return employee’s names who have one break in their employment histories (fired and rehired) and their salaries have been changed for the first time at any time when they are assigned in “Design“ department as “Sr Engineer“.

25. Temporal Queries with XQuery for $t in collection ("C1")//employee/job[.= Engineer"] for $dep in collection("C1")//employee/dept[.="Design"] where $t/../emp_no = $dep/../emp_no return if (Tcontains ($dep, $t) and Toverlaps ($dep/../salary[1],$t) and Tbreak ($dep/../job)= 1) then $dep/../name else ()

26. Conclusions • XML provides a flexible mechanism to represent temporal data. • The most usual temporal representation does not require changes of current XML standards. • An effective temporal information system should provide: • expressive temporal data model. • powerful language for temporal queries and snapshot queries as well.

27. Future Work • Add more temporal constructs. • Updates will be a real area of future investigation. • Improve the performance of our system.