Data warehouse design from XML sourcesMatte0 Golfarelli Stef.docx

Data warehouse design from XML sources
Matte0 Golfarelli Stefano Rizzi
DEIS - University of Bologna DEIS - University of Bologna
Viale Risorgimento, 2 Viale Risorgimento, 2
40136 Bologna - Italy 40136 Bologna - Italy
+39-051-642862 +39-051-2093542
[email protected][email protected]
Boris Vrdoljak
FER - University of Zagreb
Unska 3
10000 Zagreb - Croatia
+385-(O)l -6129756
[email protected]
ABSTRACT
A large amount of data needed in decision-making processes is
stored in the XML data format, which is widely used for e-
commerce and Internet-based information exchange. Thus, as
more organizations view the web as an integral part of their
communication and business, the importance of integrating
XML
data in data warehousing environments is becoming increasingly
high. In this paper we show how the design of a data mart can
be
carried out starting directly from an XML source. Two main
issues arise: on the one hand, since XML models semi-
structured

data, not all the information needed for design can be safely
derived; on the other, different approaches for representing
relationships in XML DTDs and Schemas are possible, each
with
different expressive power. After discussing these issues, we
propose a semi-automatic approach for building the conceptual
schema for a data mart starting from the XML sources.
Keywords
Data warehouse design, Data warehousing and the web, XML
1. INTRODUCTION
A large amount of data needed in decision-making processes is
stored in the XML (Extensible Markup Language) data format.
The structure of XML, composed of nested custom-defined tags
that can describe the meaning of the content itself, makes it
usable
as a semantic-preserving data exchange format on the web. As
the
Internet has evolved into a global platform for e-commerce and
information exchange, the interest in XML has been growing
and
large volumes of XML data already exist.
XML can be considered as a particular standard syntax for the
exchange of semi-structured data [l]. One common feature of
semi-structured data models is the lack of schema, so that the
data
is self-describing. However, XML documents can be associated
with and validated against either a Document Type Definition
(DTD) or an XML Schema, both of which allow the structure of
XML documents to be described and their contents to be
constrained. DTDs are defined as a part of the XML 1.0
Permission to make digital or hard copies of all or part of this
work for

personal or classroom use is granted without fee provided that
copies
are not made or distributed for profit or commercial advantage
and that
copies bear this notice and the full citation on the first page. To
copy
otherwise, or republish, to post on servers or to redistribute to
lists.
requires prior specific permission and/or a fee.
DOLAP ‘01, November 9,2001, Atlanta, Georgia, USA.
Copyright 2001 ACM l-581 l3-437-l/Ol/OOl I . ..$5.00.
Specification [15], while XML Schemas have recently become
a
W3C Recommendation [16]. XML Schemas considerably
extend
the capabilities of DTDs, especially from the point of view of
data
typing and constraining. With DTDs or Schemas, the
applications
exchanging data can agree about the meaning of the tags and, in
that case, XML reaches its full potential.
During the recent years, the enterprises have been asking for
support in the process of extracting useful, concise and handy
information for decision-making out of the tons of data stored
in
their expensive and complex information systems.
Consequently,
a substantial effort has been made in the academic world to
devise
methodologies for designing data warehousing systems capable
of
seamlessly integrating information from different sources, in
particular from heterogeneous databases. Now, as more
organizations view the web as an integral part of their

communication and business, the importance of integrating
XML
data in data warehousing environments is becoming increasingly
high. Some commercial tools now support data extraction from
XML sources to feed the warehouse, but both the warehouse
schema and the logical mapping between the source and the
target
schemas must be defined by the designer.
In this paper we show how multidimensional design for data
warehouses can be carried out starting directly from an XML
source. Two main issues arise: on the one hand, different
approaches for representing relationships in XML DTDs and
Schemas are possible, each achieving a different expressive
power; on the other, since XML models semi-structured data,
not
all the information needed for design can be safely derived.
Thus,
our contribution in this work is twofold: first, we propose a
warehouse-oriented review and comparison of the approaches
for
structuring XML documents; then, we outline an algorithm in
which the problem of correctly inferring the needed information
is
solved by querying the source XML documents and, if
necessary,
by asking the designer’s help.
One alternative approach to design from XML sources consists
in
first translating them into an equivalent relational schema, then
starting from the latter to design the warehouse. Some
approaches
for translating XML documents into a relational database can be
found in the literature, both leaning on the DTD [10][13] or not
[5], but insufficient emphasis is given to the problem of

determining the cardinality of relationships, which instead has a
primary role in multidimensional design.
The paper is structured as follows. In Section 2 the basics of
multidimensional modeling and design are given, while in
Section
3 the design alternatives for modeling relationships in both
DTDs
4 0
and XML Schemas are reviewed and discussed. In Section 4 our
approach to multidimensional design is presented with reference
to a case study, and in Section 5 the conclusions are drawn.
2. MULTIDIMENSIONAL MODELING AND
DESIGN
It is now widely recognized that an accurate conceptual design
is
the necessary foundation for building a data warehouse which is
both well-documented and fully satisfies requirements. In order
to
be independent of the specific issues involved in logical and
physical modeling, the approach proposed here is referred to the
conceptual level, from which the logical schemas of the data
marts
can be easily derived.
2.1 Conceptual Modeling
Several conceptual models for data warehouses were devised in
the literature [2][3][6][7][8][14]; they mainly differ for the
graphical representation of concepts, with small differences in
expressive power. In this paper we will adopt the Dimensional
Fact Model (DFM) [7], which represents the data mart by

means
of a set offact schemas.
In the following we will briefly discuss the DFM representation
of
the main concepts of the multidimensional model with reference
to the fact schema CLICK, which describes the analysis of the
web site traffic. The reason for choosing this example is that,
due
to the significant role now played by the web in attracting new
clients and supporting sales, analyzing the web server traffic
may
be crucial for improving the enterprise business. In this context,
multidimensional modeling allows many unpredictable complex
queries to be answered, such as:
. What is the trend for the most and the least accessed pages?
. Is there a relationship between business events (for instance,
sale promotions in an e-commerce site) and the number of
accesses?
For such an analysis, we need information about the hostname
or
IP address of the computer requesting the file, the date and time
of
the request, and the URL of the file being requested.
m o n t h 0
day of week
date (
h o l i d a y
d o m a i n

(category I
n a t i o n ) CLICK URL f i l e t y p e
0 n n o . o f c l i c k s A 0
host
( h o s t n a m e /
IP address)
hour ()
Figure 1. Fact schema for click-stream analysis
In the fact schema shown in Figure 1, the fact CLICK, focus of
interest for the decision-making process, is associated to the
measures which describe it, i.e. no. of clicks, and to the
dimensions determining its minimum granularity, namely host,
date, hour, and URL. Each dimension is the root of a hierarchy
which determines how the fact may be aggregated and selected
significantly for the decision-making process; each hierarchy
includes a set of attributes linked by functional dependencies.
For
instance, the URL of the file being requested determines the file
type, and the hostname or IP address of the computer requesting
the file determines its domain. The values for the domain
attribute
can be extracted from the hostname suffix that is indicating
either
the category (for instance, “.com” for commercial companies)
or
the nation.
Within the DFM, as in all the other conceptual models, a strong
relevance is given to functional dependencies, since they
represent
many-to-one relationships between attributes which enable

flexible aggregation of data in OLAP queries [9]. Thus, the
main
problem in building a conceptual data mart schema is to identify
those relationships in the business domain.
2.2 Conceptual Design
In most approaches to design of data marts, the conceptual
schema
is built starting from the (logical or conceptual) schema of the
operational sources [2][7][8]. The common core of these
approaches consists in navigating the functional dependencies
in
the source schema in order to determine the hierarchies for the
fact. In particular, the methodology to build a fact schema from
an
E/R schema proposed in [7] consists of the following steps:
1. Choosing facts.
2. For each fact:
2.1 Building the attribute tree.
2.2 Rearranging the attribute tree.
2.3 Defining dimensions and measures
For briefly illustrating the methodology, we will use the E/R
diagram describing the web site traffic shown in Figure 2.
Facts typically correspond to events occurring dynamically in
the
enterprise world. On the E/R schema a fact may be represented
either by an entity F or by an n-ary relationship R. In our
example,
the fact of primary interest is represented by entity CLICK.

Given a portion of interest of a source schema and an entity F
belonging to it, we call attribute tree the tree such that:
. each vertex corresponds to an attribute - simple or compound
- of the schema;
. the root corresponds to the identifier of F,
. for each vertex V, the corresponding attribute functionally
determines all the attributes corresponding to the descendants
of v.
The attribute tree for F may be constructed automatically by
navigating, starting from F, the functional dependencies
expressed
by many-to-one relationships between entities in the source
schema. Each entity E analyzed is represented in the attribute
tree
by: (1) a node corresponding to the identifier of E, (2) a child
node for each of the non-identifier attributes of E; (3) a child
node
for each entity G connected to E by a many-to-one relationship.
41
time d a t e
(IN
r-5U R LC A T E G O R Y
Figure 2. E/R schema for web site traffk
The resulting attribute tree for the web site traffic analysis

elements (i.e. sub-elements) and attributes. Both elements and
example is shown in Figure 3. Starting from the CLICK entity,
attributes are allowed to have values. Document structures can
be
new vertices are added by following many-to-one relationships.
nested to any level of complexity; between the opening and
The URL category attribute could not be included in the
attribute closing tags for an element, any number of additional
elements or
tree because there is a many-to-many relationship between URL
textual data may be present. Attributes are included, with their
and URL category. The nation attribute is included twice in the
respective values, within the element’s opening declaration. An
attribute tree because it can be added to the attribute tree by
XML document that contains data about the web site traffic is
navigating from both the host and the site entities. shown in
Figure 4.
d a t e time <webTraffic>
< c l i c k >
< h o s t hostId=“ares.csr.unibo.it”>
<nation>italy</nation>
</host>
<date>23-MAY-ZOOl</date>
<time>16:43:25</time>
curl urlID="BL0023">
<site siteID="www.hr">
<nation>croatia</nation>
</site>
Figure 3. Attribute tree for fact CLZCK

Once the attribute tree has been built, some uninteresting nodes
may be dropped from it. The time attribute, that represents the
exact time of a click, is replaced with a coarser hour attribute.
<fileType>shtml</fileType>
<urlCategory>catalog</urlCategory>
</url>
</click>
<click>
Finally, dimensions and measures must be selected among the
children of the root. In our example, the attributes chosen as
dimensions are host, date, hour and URL; number of clicks,
determined by counting the clicks from the same host to the
same
URL on a given date and hour, is chosen as a measure.
. .
</click>
. . .
</webTraffic>
F i g u r e 4 . A n X M L d o c u m e n t d e s c r i b i n g t h
e w e b s i t e t r a f f i c
Some further minor arrangements must be made in order to
obtain
the fact schema in Figure 1; in particular, the date dimension is
enriched by building a hierarchy which includes attributes
month,
day of week, and holiday. Besides, the host category and nation
optional attributes are replaced by attribute domain, which
indicates either the category or the nation of the host.

An XML document is valid if it has an associated schema, such
as
a DTD or an XML Schema, and if it conforms to the constraints
expressed in that schema. Since our methodology for conceptual
design is based on detecting many-to-one relationships, in the
following we will focus on the way those relationships can be
e x p r e s s e d i n t h e D T D a n d t h e X M L S c h e m
a .
3. MODELING RELATIONSHIPS IN XML
An XML document consists of nested element structures,
starting
with a root element. Each element may contain component
3.1 Relationships in DTDs
A DTD defines elements and attributes allowed in an XML
document, and the nesting and occurrences of each element. The
42
structure of an XML document is constrained using element-
type
and attribute-list declarations. Element-type declarations
specify
which sub-elements can appear as children of the element;
attribute-list declarations specify the name, type, and possibly
default value of each attribute associated with a given element
type. Among the different attribute types, types ID, IDREF and
IDREFS have particular relevance for our approach: the ID type
defines a unique identifier for the element; the IDREF type
means
that the attribute’s value is some other element’s identifier;
IDREFS means that its value is a list of identifiers. IDREF(S)
must match the value of some ID attribute in the document [ 11.

3. I .2 Modeling relationships by ID and IDREF(S)
Another way to specify relationships between elements in DTDs
is by means of ID and IDREF(S) attributes. The way these
attributes operate resembles the key and foreign key mechanism
used in relational databases, with some important differences
and
limitations. For instance, if we take the last part of the DTD
from
Figure 5 and let the relationships of the url element be defined
by
ID and IDREF(S) attributes rather than by sub-elements, we
obtain the DTD in Figure 6.
3.1. I Modeling relationships by sub-elements
<!ELEMENT webTraffic ( c l i c k * , f i l e T y p e + ,
urlCategory+) >
Relationships can be specified in DTDs by sub-elements that
may
have different cafdinalities. The optional character following a
child element name or list in the element-type declarations
determines whether the element or the content particles in the
list
may appear one or more (+), zero or more (*), or zero or one
times (?); the default cardinality is exactly one.
Figure 5 presents a DTD according to which the XML document
presented in Figure 4 is valid. Element webTraffic is defined as
a
document element, thus becomes the root of XML documents. A
webTraffic element may have many click elements, while in an
url
element the site sub-element must occur exactly once, followed
by
onefiJeType and many urlCategory sub-elements. A host

element
may have either a category or a nation element.
<!DOCTYPE webTraffic [
<!ELEMEIW webTraffic ( c l i c k * ) >
<!ELEMENT c l i c k ( h o s t , d a t e , t i m e , url) >
<!ELEMENT h o s t ( c a t e g o r y 1 n a t i o n ) >
<!ATTLIST h o s t
hostId ID #REQUIRED>
<!ELEMENT c a t e g o r y (#PCDATA)>
<!ELEMENT d a t e (#PCDATA)>
<!ELEMENT t i m e (#PCDATA)>
<!ELEMENT url ( s i t e , fileType,
urlCategory+) >
<!ATTLIST u r l
urlId ID #REQUIRED>
< ! ELEMENT site (nation) >
<!ATTLIST s i t e
siteId ID #REQUIRED>
< ! ELEMENT nation (#PCDATA) >
<!ELEMENT fileType (#PCDATA)>
<!ELEMENT urlcategory (#PCDATA)>
I>
Figure 5. A DTD where relationships are specified by sub-
e l e m e n t s
If a one-to-one or one-to-many relationship must be represented
in
XML, sub-elements with the above mentioned cardinalities can
be

used without loss of information. However, given a DTD, we
can
follow only one direction of a relationship. For instance,
according to the DTD in Figure 5, an url element may have
many
urlCutegory sub-elements, but it is not possible to find out,
from
the DTD, whether an URL category can refer to many URLs.
Only by having some knowledge about the domain described by
the DTD, we can conclude that the latter is the case.
. . .
<!ELEMENT u r l ( s i t e )
<!ATTLIST url
urlId ID #REQUIRED
fileTypeRef I D R E F # R E Q U I R E D
urlCategoryRef IDREFS # R E Q U I R E D >
<!ELEMENT fileType EMPTY>
<!ATTLIST fileType
typeID ID #REQUIRED
typeDescription CDATA #IMPLIED>
<!ELEMENT urlcategory EMPTY>
<!ATTLIST urlcategory
urlCategoryId ID #REQUIRED
urlCateqorvDesc CDATA #IMPLIED>
Figure 6. A DTD where relationships are specified by
IDREF(S)
In this example we assume that the$ZeTypeRefattribute of the
url

element references the fireType element. On the other hand, an
instance of urlCategoryRef references many instances of
urZCategory. Therefore, for one URL there is exactly one file
type
while there may be several categories. The problem is that,
using
IDREF(S), the participating elements cannot be constrained to
be
of a certain element type. For instance, jZeTypeRef could also
contain a reference to an urZCutegory element, while obviously
we would like to constrain such references tofiZeType elements
only. Further, though the value of an ID attribute is unique
within
the whole document, element types are not required to have an
ID,
and even if an element type has an ID, its usage may be
optional.
For these reasons, there is no means to actually constrain the
allowed relationships by the ID/IDREF(S) mechanism.
3.2 Relationships in XML Schemas
XML Schemas give more accurate representation of the XML
structure constraints than DTDs; in particular, the cardinal&y
can
be specified in more detail by using the minoccurs and
maxoccurs attributes. An XML Schema consists of type
definitions, which can be derived from each other, and element
declarations. The possibility of separating an element
declaration
from the definition of its type enables the sharing and reusing
of
simple and composite types. Further, besides ID and IDREF(S)
attributes, XML Schemas support the use of key and keyRef
elements for defining keys and their references.
The two different ways of specifying relationships in XML

Schemas, i.e. sub-elements and key/keyRef mechanisms, will be
described in the following.
43
3.2.1 Modeling relationships by sub-elements
Figure 7 presents a part of an XML Schema that defines the
same
structure and constraints as the DTD in Figure 5. Since, for
defining elements, attributes minOccurs and maxOccurs default
to
1, and only in the case of the urlCategory element the
maxOccurs
attribute is set to “unbounded”, an w-1 element will consist of
one
site, one$leType and many urlcategory elements. We will not
further discuss this solution since, from the point of view of
cardinality modeling, it is essentially equivalent to the solution
in
Section3.1.1.
< e l e m e n t name=“click”>
<complexType>
<sequence>
<element name="host"
type="hostType"/>
<element name="date" type="date"/>
<element name="time" type="time"/>
<element name="url" type="urlType"/>
</sequence>
</complexType>

</element>
. .
<complexType name="urlType">
<sequence>
<element name="site" type="siteType"/>
<element name="fileType"
type="string"/>
<element name="urlCategory"
type="string"
maxOccurs="unbounded"/>
</sequence>
<attribute name="urlID" type="string"/>
</complexType>
. . .
F i g u r e 7 . A n X M L S c h e m a w h e r e r e l a t i o n
s h i p s a r e s p e c i f i e d
b y s u b - e l e m e n t s
3 . 2 . 2 Modeling relationships by key and keyRef
elements
In addition to ID and IDREF(S) attributes, XML Schemas
introduce more powerful and flexible mechanisms, similar to
the
relational concept of foreign key. The key element is used to
indicate that every attribute or element value must be unique
within a certain scope and not null. By using keyRef elements,
keys can be referenced.
For the purpose of describing the declaration of keys and their

references, an element namedfileTypes (represented in Figure
8)
is added as a sub-element to the click element from the XML
document in Figure 4. The fileTypes element consists of type
elements with z’ypeZD attributes. Figure 9 presents an
evolution of
the XML Schema in Figure 7, according to which every click
element also contains a fileTypes element. This element has a
composite type named fileTypesType (its definition is not
presented in Figure 9), according to which the XML document
in
Figure 8 is valid.
XML Schemas allow to specify the scope for each key by means
of an XPath expression [17]. In our example, the key element
is
namedfileTypeKey. The typeID attribute from Figure 8 is
specified as the key by means of the selector and thefield sub-
elements of the key attribute in Figure 9. The xpath attribute of
the
selector element contains an XPath expression,$leTypes/type,
that
selects the type elements that are sub-elements of the fileTypes
element. The xpath attribute of the field element contains the
@typeZD expression, that specifies the typeZD attribute of the
type
element as the key. Further, thefileType element, that is a sub-
element of w-1, is declared as a keyReJ this means that, for
every
value ofjZeType, afileTypeKey with the same value must
exist.
<fileTypes>
<type typeID="html">

hypertext markup language</type>
<type typeID="gif">
graphic interchange format</type>
. . .
</fileTypes>
Figure 8. ThefiZeTypes element as a sub-element of click
<element name="click">
<complexType>
. . .
<element name="url" type="urlType"/>
<element name="fileTypes"
type="fileTypesType"/>
. . .
</complexType>
<key name="fileTypeKey">
<selector xpath="fileTypes/type"/>
<field xpath="@typeID"/>
</key>
ckeyref name="fileTypeRef"
refer="fileTypeKey">
<selector xpath=“url”/>
<field xpath="fileType"/>
i/keyref>
</element>

Figure 9. An XML Schema where relationships are specified
by key/keyRef
In conclusion, the key/keyRef mechanism may be applied to any
element and attribute content, and the scope of the constraint
can
be specified, while an ID is a type of attribute whose scope is
fixed to be the whole document. Furthermore, combinations of
element and attribute content can also serve as keys and
references in XML Schemas.
4. CONCEPTUAL DESIGN FROM XML
SOURCES
In the previous section, we showed different approaches for
representing relationships in DTDs and XML Schemas. Three
of
them are suitable for specifying relationships: sub-elements in
DTDs, sub-elements and key/keyRef in Schemas; though their
expressive power is different, in the context of this paper they
may be considered to be equivalent since with reference to
multidimensional design they allow the same knowledge to be
captured. We do not consider the fourth approach, ID/IDREF(S)
4 4
in DTDs, since it is not precise and useful enough in
constraining
relationships.
In this section we propose a semi-automatic approach for
building
the conceptual schema of a data mart starting from the XML
sources. Of the three above-mentioned approaches, we have
chosen sub-elements in DTDs for the presentation of our

methodology, since Schemas are still not as widely used as
DTDs;
however, the methodology is essentially the same when dealing
w i t h S c h e m a s .
Starting with the assumption that the XML document has a DTD
and conforms to it, the methodology consists of the following
steps:
1. Simplifying the DTD.
2 . Creating a DTD graph.
3. Choosing facts.
4. For each fact:
4.1 Building the attribute tree from the DTD graph.
4.2 Rearranging the attribute tree.
4.3 Defining dimensions and measures.
In the following paragraphs we will describe steps from (1) to
(4.1) referring to the web site traffic example; once the attribute
tree is built, steps 4.2 and 4.3 are identical, respectively, to
steps
2.2 and 2.3 described in Section 2.3.
Simvlifvina t h e D T D
The sub-elements in DTDs may have been declared in a
complicated and redundant way. However, those details of a
DTD
can be simplified [ 131. The transformations for simplifying a
DTD
include converting a nested definition into a flat representation:

f o r i n s t a n c e , i n t h e w e b s i t e t r a f f i c e x a m
p l e ,
host(category /nation) i s t r a n s f o r m e d i n t o
host(category?,nation?). Further, the sub-elements having the
same name are grouped, and many unary operators are reduced
to
a single unary operator. Finally, all “+” operators are
transformed
intO “*” operators.
Creating a DTD waph
After simplifying the DTD, a DTD graph representing its
structure
can be created as described in [lo] and [13]; its vertices
correspond to elements, attributes and operators in the DTD.
Attributes and sub-elements are not distinguished in the graph
since, in our methodology, they are considered as equivalent
nesting mechanisms. The DTD graph for the DTD in Figure 5 is
given in Figure 10.
Definina facts
The designer chooses one or more vertices of the DTD graph as
facts; each of them becomes the root of a fact schema. In our
example, we choose the click vertex as the only interesting fact.
Building the attribute tree
The vertices of the attribute tree are a subset of the element and
attribute vertices of the DTD graph. The algorithm to build the
attribute tree is sketched in Figure Il.
webTraffic
1
click

host d a t e t i m e url
Figure 10. DTD graph for web site traffic analysis
root=newVertex (F) ;
// newVertex(<vertex>) r e t u r n s a n e w v e r t e x
// o f t h e a t t r i b u t e t r e e
// c o r r e s p o n d i n g t o < v e r t e x >
// o f t h e D T D g r a p h
expand(F,root) ;
expand(E,V) :
// E i s t h e c u r r e n t DTD vertex,
// V i s t h e c u r r e n t a t t r i b u t e t r e e v e r t e x
{ f o r e a c h c h i l d W o f E d o
if W i s e l e m e n t o r a t t r i b u t e d o
{ next=newVertex(W);
addchild (V, next) ;
// a d d s c h i l d W t o V
expand (W, next ) ;
else
if W="?" do
expand (W, V) ;
for each parent Z of E such that
Z i s n o t a d o c u m e n t e l e m e n t d o
if Z=“3” or Z=“*” do
expand(Z,V) ;
e l s e
if n o t toMany(E,Z) do

if askDesignerToOne (E, Z) do
{ next=newVertex(Z);
addChild(V,next) ;
expand (Z, next) ;
1
Figure 11. Algorithm for building the attribute tree
The attribute tree is initialized with the fact vertex F; then, it is
enlarged by recursively navigating the functional dependencies
between the vertices of the DTD graph. Each vertex V inserted
in
the attribute tree is expanded as follows (procedure expand):
1. For each vertex W that is a child of V in the DTD graph:
When examining relationships in the same direction as in the
DTD graph, the cardinality information is expressed either
explicitly by “?” and “*” vertices or implicitly by their
4 5
2 . For each vertex Z that is a parent of V in the DTD graph:
absence. If W corresponds to an element or attribute in the since
there is no need for the existence of both host and hostld
DTD, it is added to the attribute tree as a child of V; if W is a
vertices, only host should be left, the same logic should be
applied
“ ? ” operator, its child is added to the attribute tree as a child
of for urlld and siteld. Finally, the time attribute is replaced
with the

V; if W is a “*” operator, no vertex is added. hour attribute.
When examining relationships in this direction, vertices
corresponding to “*” and “?,, operators are skipped since they
only express the cardinality in the opposite direction. Since
the DTD yields no further information about the …

Data warehouse design from XML sourcesMatte0 Golfarelli Stef.docx

More Related Content

Similar to Data warehouse design from XML sourcesMatte0 Golfarelli Stef.docx (20)

More from randyburney60861 (20)

Recently uploaded (20)

Data warehouse design from XML sourcesMatte0 Golfarelli Stef.docx