e-Science e-Business
e-Government and their
Technologies
Core XML
Bryan Carpenter, Geoffrey Fox, Marlon Pierce
Pervasive Technology Laboratories
Indiana University Bloomington IN 47404
January 12 2004
[email protected]
[email protected]
[email protected]
http://www.grid2004.org/spring2004
1
What are we doing








This is a semester-long course on Grids (viewed as technologies
and infrastructure) and the application – mainly to science but
also to business and government
We will assume a basic knowledge of the Java language and then
interweave 6 topic areas – first four cover technologies that will
be used by students
1) Advanced Java: including networking, Java Server Pages and
perhaps servlets
2) XML: Specification, Tools, Linkage to Java
3) Web Services: Basic Ideas, WSDL, Axis and Tomcat
4) Grid Systems: GT3/Cogkit, Gateway, XSOAP, Portlet
5) Advanced Technology Discussions: CORBA as history, OGSADAI, security, Semantic Grid, Workflow
6) Applications: Bioinformatics, Particle Physics, Engineering,
Crises, Computing-on-demand Grid, Earth Science
2
Contents of this Lecture Set


Intro: HTML and XML and Unicode
Core XML:
•
•
•
•

XML syntax and well-formedness, DTDs and validity.
XML namespaces.
The XML DOM with linkage to Java.
XPath basics.
XML Schema.
• Validation for data-centric applications.

Later lectures may include additional information on:
•
•
•
•
XHTML, SVG, RDF.
XML style languages: XSLT and CSS.
XML Databases (Xindice, Sleepycat).
Search: advanced XPath, XQuery.
3
Motivations for XML: a Better HTML?

Limitations of HTML:
• Extensibility: HTML does not allow users to specify
their own tags or attributes in order to parameterize or
otherwise semantically qualify their data.
• Structure: HTML does not support the specification of
deep structures needed to represent database schema or
object-oriented hierarchies.
• Validation: HTML does not support the kind of
language specification that allows applications to check
data for structural validity when it is imported.
4
XML in the HTML world

XML = eXtensible Markup Language.
• XML is a subset of SGML—Standard Generalized Markup
Language, but XML is specifically designed for the web.


Specification by W3C: http://www.w3.org/XML and lots of
links like http://www.xml.org
XML 1.0 in February 98.
• XML 1.1 became a W3C recommendation 4 Feb, 2004!

How XML fits into the new HTML world:
• XML describes the logical structure of the document.
• CSS (Cascading Style Sheets) and/or XSL describes the visual
presentation of the document.
• DOM (Document Object Model) allows scripting languages like
JavaScript to access and dynamically change document objects.
5
Logical vs. Visual Design

Logical design of a document (content) should be
separate from its visual design (presentation).
•
•
•
•

Promotes sound typography.
Encourages better writing.
Is more flexible.
Allows the same “knowledge/information” (defined
in XML) to be displayed on PC’s, PDA’s, Braille
devices etc.
XML used to define the logical design, with XSL
(Extensible Style Language) or other mechanism
used to define the visual layout (e.g. by mapping
XML into HTML).
6
XML Design Goals
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
XML shall be usable over the Internet.
XML shall support a variety of applications.
XML shall be compatible with SGML.
It shall be easy to write programs that process XML
documents.
Optional features in XML shall be kept to the absolute
minimum, ideally zero.
XML documents should be human-legible and
reasonably clear.
Design of XML should be prepared quickly.
Design of XML shall be formal and concise.
XML documents shall be easy to create.
Terseness in XML markup is of minimal importance.
7
Document-Centric or Data-Centric?


Roots of XML in document markup (HTML-like).
In practice use of XML as a data format has become at
least as pervasive. Examples:
• Use of XML format in configuration and deployment files of
EJB, Tomcat, …
• Uses of XML as a format for message exchange (e.g. SOAP,
BEEP).

There is also an important intermediate case—XML as
program text for machine interpretation. E.g.:
• XSLT declarative transformation language.
• WSDL interface definition language for Web services.
• BPEL Web services workflow language.
8
Features of XML


Documents are stored in plain text and thus can
be transferred and processed anywhere.
Unifying principles make it easily acceptable:
• “Everything is a tree” (DOM).
• Unicode for different languages.
9
XML and Unicode


All XML documents must be written using the Unicode
character set.
Unicode is also the character set used by Java, C#,
ECMAScript, …, so we should know something about it.
10
Special Topic: Unicode
11
Unicode

Unicode (http://www.unicode.org) is an international
standard character set that covers alphabets of all the
World’s common written languages.
• Eventually it should cover all languages, living and dead.
• Unicode helps make the Web truly “worldwide”?!


Unlike, say, ASCII, which allows for 128 characters,
Unicode has space for over 1,000,000, of which around
96,000 are currently allocated.
Unicode itself assigns a unique sequence number (code
point) to any character, regardless its alphabet.
• Three Unicode encoding forms map these code points to
sequences of fixed size units—UTF-8, UTF-16, UTF-32.
12
Unicode Code Points


A Unicode code point is a numeric value between 0 and
10FFFF16, commonly denoted in one of the formats:
U+XXXX
U+XXXXX
U+10XXXX
where X is a hexadecimal digit.
There are a total of 1,114,112 (= 17 · 164) code points, but most of
the World’s common characters are encoded in the first 65,536
points—the Basic Multilingual Plane (BMP).
• 2048 code points in BMP are disallowed because their values have a
special role in UTF-16 encoding.

For each assigned character code, the Unicode standard defines
a name, and “semantic” properties like case, directionality, ...
13
Planes


The space of 17 · 216 Unicode code points is
conventionally divided into 17 planes of 216 points each.
Currently used planes include:
Plane #
Plane Name
Code Range
0
Basic Multilingual Plane
000016..FFFF16
1
Supplementary Multilingual Plane
1000016..1FFFF16
2
Supplementary Ideographic Plane
2000016..2FFFF16
• Note early versions of Unicode used a strict 16-bit encoding,
and essentially contained just BMP
14
Unicode
Allocation

Layout of
planes:
15
Blocks



Planes are subdivided into blocks.
Blocks have variable size. Each block contains the
characters of one alphabet or a group of related
alphabets.
The following slides are a random sampling of the
blocks in BMP.
• I have put 128 code points on each slide, but this is just what
would fit… no general significance to pages of size 128.

For all blocks in the current Unicode standard see:
http://www.unicode.org/charts/
16
“Basic Latin” (a.k.a. ASCII)
17
“Latin 1” (supplement)
18
“Greek and Coptic”
19
“Arabic” (1 of 2)
20
“Devanagari”
21
“Hangul Jamo” (1 of 2)
22
“CJK Unified Ideographs” (1 of 164)
23
Unicode
Allocation

Layout of
Basic
Multilingual
Plane:
24
Unicode
Allocation

Layout of
Plane 1:
25
Encoding Forms





In electronic documents or computer programs the
space of Unicode code points is normally broken down
into a sequence of units, each unit having a convenient,
fixed number of bits.
The Unicode standard defines 3 encoding forms.
The most straightforward is UTF-32, in which the units
have size 32 bits.
This unit is easily large enough to hold the integer value
of a single code point, so UTF-32 encoding is “obvious”.
But for nearly all documents, UTF-32 wastes at least
half the available storage space.
• Also, most programming languages work with 8 bit or 16 bit
character units.
26
UTF-16

The UTF-16 encoding form breaks Unicode characters
into 16 bit units.
• Java, for example, uses UTF-16 for chars and Strings.


One 16 bit unit is not large enough to represent all
possible Unicode code points.
Code points higher than 216-1 are split over two
consecutive units.
• These are called surrogate pairs. The leading unit is a highsurrogate unit; trailing is a low-surrogate unit.
• There are 1024 code points reserved in the BMP for high
surrogates, and 1024 more reserved for low surrogates.

This allows for 1024 · 1024 surrogate pairs representing code points
higher than 216-1, while ensuring a legal BMP code point can always
be represented in a single unit, and such a unit can never be confused
with a surrogate unit.
27
UTF-16 Bit Distribution
28
UTF-8


The UTF-8 encoding form breaks Unicode characters
into 8 bit units (i.e., individual bytes).
UTF-8 is a variable-width encoding with the following
properties:
• Any Unicode code point maps to 1, 2, 3, or 4 bytes.
• Byte sub-sequences for individual characters can always be
recognized by local search in the encoded string.
• The Basic Latin block coding points (U+0000..U+007F) map
to one byte, identical to their ASCII value.
• All code points in the BMP map to at most 3 bytes.
• For European texts UTF-8 will normally use 8 or 16 bits per
character (vs 16 bits for UTF-16).
• For East Asian texts UTF-8 will normally use 24 bits per
character (vs 16 bits for UTF-16).
29
UTF-8 Bit Distribution
30
Encoding Schemes


The 3 encoding forms don’t quite complete the encoding
schemes of Unicode, because they don’t address the
endianness with which the UTF-32, UTF-16 numeric
unit values are rendered to bytes (byte-serialization).
To allow applications to distinguish the endianness of a
given document instance, Unicode allows a Byte Order
Mark (BOM) as the first character of a document.
• BOM is a code point (U+FEFF) for which the byte-reversed
unit value doesn’t correspond to a legal code point, so serves
to determine the actual byte order.
31
The Seven Unicode Encoding Schemes
32
Unicode Summary


Unicode is a large and important standard that is a
foundation for XML, HTML, etc.
Although you are unlikely to manipulate the encodings
yourself, you should be aware of the pros and cons of
UTF-16, UTF-8.
• UTF-8 is backwards compatible with ASCII—Basic Latin
texts can be read by legacy applications.
• UTF-16 is better-suited for internationalization. It is the
internal representation used by Java, C#, ECMAScript, …
33
Core XML I:
The XML Specification
34
Introduction




In this section we will describe core XML, as
defined by the XML specification document from
W3C.
XML is a format for documents—originally
documents for the Web—but its scope is wider than
that.
XML is a subset of SGML—Standard Generalized
Markup Language. Some features of XML exist
simply for compatibility with SGML.
XML can also be viewed as a kind of generalization
of HTML—presumably familiar from the Web.
35
XML Parsers and Applications




For purposes of this section an application is any
program that reads data from an XML document.
Applications normally do not (and probably should
not) read the text of XML documents directly.
The XML specification assumes that this text is initially
processed by a piece of software called an XML
processor. We will also refer to this as an XML parser.
The parser exhaustively checks that the text is in a legal
XML form, then extracts the essential data from the
document, and hands that data to the application.
36
Reading XML Data
<?xml version="1.0"?>
<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.0//EN"
"http://www.w3.org/TR/2001/REC-SVG-20010904/DTD/svg10.dtd">
<svg width="500" height="500">
XML Parser
<g transform='rotate(45)'>
<circle cx='150' cy='50' r='25'/>
<text x='125' y='100'>A Circle</text>
</g>
</svg>
XML Source
svg
g
Application
width
500
height
500
Parsed XML Data
transform
rotate(45)
circle
cx
150
cy
50
r
25
text
x
125
y
100
A Circle
37
Well-formed Documents

An XML document follows a strict syntax. For
example:
• An XML document contains regions of text called elements,
delimited by matching start-tags and end-tags. Elements must
be correctly nested.
• Start-tags may include attribute specifications, where attribute
values are strings delimited by matching quote marks.


A document that obeys the full set of these rules is
called well-formed.
Every legal XML document must be well-formed,
otherwise it cannot be parsed.
38
Examples

Well-formed:
<html>
<body style="font-style: italic">
This is a well-formed document.
</body>
</html>

Not well-formed:
<html>
<body style=font-style: italic>
This is not a well-formed document.
</html>
</body>
• The style attribute value is not in quote marks, and the html
and body tags don’t nest correctly.
39
Install Xerces



The Xerces parser is a product of the Apache XML
project, http://xml.apache.org.
Follow the “Xerces Java 2” project link and go to the
download area, then to the master distribution
directory or a mirror directory.
Download Xerces-J-bin.2.6.2.zip, and extract it to a
suitable place, e.g. C:\
• When extracting, remember to select “Use folder names”!!
• This should create a folder called, e.g., C:\xerces-2_6_2\.
40
Put Xerces on your Class Path

Using the menu at
Control Panel→System→Advanced→Environment Variables
add the jar files xercesSamples.jar, xercesImpl.jar,
and xml-apis.jar, to you class path.
• E.g. append
…;C:\xerces2_6_2\xercesSamples.jar;C:\xerces2_6_2\xer
cesImpl.jar;C:\xerces-2_6_2\xml-apis.jar
to the value of your CLASSPATH variable.
41
Example Using Xerces

Copy the two HTML examples given above to files
called, say, wellformed.html and illformed.html. Then,
in a new Command Prompt window, try running the
commands:
> java dom.Writer wellformed.html
…
> java dom.Writer illformed.html
…
• The first command should just echo the document. The
second should print a syntax error message.
• dom.Writer is one of the sample applications in the Xerces
release. It simply uses the Xerces parser to convert the source
file to a tree data structure (DOM), then converts the tree
back to nicely formatted XML, which it prints.
42
“Rolling Your Own” Parser?

People approaching XML sometimes decide they can
write their own “lightweight” parser that handles just
the bit of XML their application needs.
• In general this is a bad idea!
• We will see later that even basic XML is a moderately
complex specification; unless you are going to invest a lot of
effort it is unlikely you can parse the full specification more
efficiently than existing parsers.
• If you subset the specification you may be compromising the
most crucial advantage that XML brings to your
application—interoperability.
• Later in these lectures we will see how to use the Xerces
parser from your own Java programs, to read XML input.
43
Valid Documents

An XML document may optionally include a Document
Type Definition (DTD).
• This declares the names of all elements and attributes
appearing in the document, and how they may nest.
• The DTD also declares and defines entities that may be
referenced from within document content.

A well-formed XML document that includes a DTD—
and accurately follows the declarations in that DTD—is
called valid.
44
Invalid Documents



It is quite possible to parse invalid (but well-formed)
documents, by using a non-validating parser.
Many applications accept XML files without DTDs,
which are therefore technically invalid.
Applications may exploit “validation” mechanisms
other than DTDs. An important one is XML Schema
which we will discuss later.
• A document validated against an XML Schema usually does
not have a DTD, so technically is invalid as far as the base
XML specification is concerned.
• But of course it is valid relative to the XML Schema
specification!
45
Validation Side Effects


The use of a validating parser certainly affects what
documents are treated as legal.
In some cases “switching on” validation may also alter
the exact data passed from the parser to application.
These effects will be considered when we discuss DTDs.
46
Physical Entities


An XML document is represented by one or more
“storage units” (typically files), called “entities”.
We can enumerate five kinds:
• Document entities—root XML documents.
• Parsed external entities, which contain fragmentary XML
content.
• External DTD subsets, which contain some or all of the DTD
declarations needed by a document.
• External parameter entities, which also contain fragmentary
DTD content.
• Unparsed external entities, which are usually complete
“binary” files in some native format (not XML).
47
Physical Structure



The structure of a non-trivial XML document is
illustrated in the following figure.
Every XML document must have exactly one document
entity.
It may also involve zero or more external entities:
• The document entity may reference any number of external
general entities. These can be parsed external entities or
unparsed external entities. A parsed external entity may in
turn reference other external general entities.
• The document may have at most one external DTD subset.
• A DTD subset in the document entity, or an external DTD
subset, may reference any number of external parameter
entities (which may in turn reference other external
parameter entities).
48
A Complex XML Document
Document
Entity
External
Parameter Entity
External DTD
Subset
DTD
External
Parameter Entity
Content
Parsed External
Entity
Unparsed
External Entity
Parsed External
Entity
Parsed External
Entity
49
Syntactic Features

The following two tables summarize the “top-level”
syntax of all the constructs in XML. Full details will be
given in later slides, as needed.
• The first columns give an abbreviated example of the syntax,
the second columns (“what?”) describe the construct, and the
third columns (“where?”) specify the places in an XML
document where the construct may appear.
• In a “where?” column, Document means at the top-level of the
document entity, and Content means in the kind of content
allowed in an element—also called Parsed Character Data.
• A Literal is character data in quotes—exactly what can
appear in a literal depends strongly on its context.
• XML Names will be discussed shortly.
50
Syntax I: Logical Structures
Example Syntax
What?
Where?
<Name …>Content</Name> Element
Document, Content
Name = Literal
Attribute specification
Element start tag
<?xml …>
XML declaration/
Text declaration
Document/
External entity
<?Name …>
Processing instruction
Document, DTD, Content
<!-- … -->
Comment
Document, DTD, Content
<!DOCTYPE …>
DTD
Document
<!ELEMENT …>
Element declaration
DTD
<!ATTLIST …>
Attributes declaration
DTD
<!ENTITY …>
Entity declaration
DTD
<!NOTATION …>
Notation declaration
DTD
51
Syntax II: References, Sections
Example Syntax What?
Where?
&#Code-point;
Character reference
Content, Literal
&Name;
Entity reference
Content, Literal
%Name;
Parameter entity reference
DTD
<![[ … ]]>
CDATA section
Content
<![IGNORE[ … ]]>
Conditional section
DTD
<![INCLUDE[ … ]]>
Conditional section
DTD
52
Character Set

Every XML document must be composed using the
Unicode character set.
• The specification does not stipulate any particular encoding,
though defaults are UTF-8 or UTF-16.
• ASCII is a subset of Unicode, so you can create XML
documents using your favorite, pre-Unicode, text editor.
53
Allowed Character Ranges

The allowed characters are
• white space:
U+0020 (space), U+0009 (tab), U+000A (line feed),
U+000D (carriage return),
plus all Unicode characters higher than U+0020, excluding:
 The surrogate blocks U+D800..DFFF.
 U+FFFE and U+FFFF (noncharacters in Unicode. Note
FFFE16 is the BOM after byte-reversal in UTF-16)

Because some codes are forbidden, can’t consider
including raw binary data in parsed XML (without
additional encoding).
54
Names and Name Tokens

In XML, names are used in many places:
• An element has a name, an attribute has a name, an entity is
referenced by a name, etc.


As in programming languages, there are rules about
what constitutes a valid name (next slide).
In XML there is also a concept of name tokens, which
are strings similar to names that can be specified as
values of certain types of attribute.
• They are less restricted. For example a number can be a valid
name token.
55
What is a Name?

Well-formed XML name tokens include any sequence of
the following characters:
• a letter
• a digit
• a period (“.”), hyphen (“-”), underscore (“_”), or colon (“:”)

Well-formed XML names include any name token that
starts with a letter, “_” or “:”.
• Names that begin with “XML” (in upper, lower or mixed case)
are reserved.

Case is significant in XML names: names are only
identified if they consist of identical character
sequences.
56
Examples

Some well-formed names:
bryan
Big-Foot
Century_21
rated:PG
_.--._

Some illegal or reserved names:
2004
.com
#xFFFF
xml-1.0
57
What is a Letter?



In defining names and name tokens, XML 1.0 relies on
a definition of letter and digit. These are “Westerncentric” notions—see Chapter 4 of the Unicode 4
standard for relevant discussion.
XML 1.0 defines a “letter” as an alphabetic or syllabic
base character or an ideograph, and gives a (probably
dated) recipe for extracting these from Unicode
character databases.
XML 1.1 gives intentionally liberal Unicode ranges for
NameStartChar and NameChar, then has a nonnormative appendix suggesting what kind of characters
should be preferred.
58
Elements


The most characteristic markup feature of XML is the
element.
The basic syntax for an element is either
<Name>Content</Name>
or
<Name/>


Here Name is an XML name—the type of element.
In the first case, Content stands for further text, which
may include nested elements.
The second case is called an empty element. It is
formally equivalent to
<Name></Name>
59
Examples from XHTML


XHTML is an XML-compatible dialect of HTML.
Examples:
• <h1>A header</h1>
This is an element of type h1 with content text “A header”.
• <body><h1>My Page</h1>Welcome.</body>
This is an element of type body. The content is a nested h1
element plus the text “Welcome.”.
• <hr/>
This is an empty element of type hr.
• Examples here don’t illustrate, but it is allowed to include
white space in the tags, on either side of the element name.
60
Attributes


An element start tag (or an empty element tag) may
include one or more attribute specifications.
An attribute specification has the general syntax
Name = "Value"
or
Name = 'Value'
where Name is the name of the attribute and Value is
some text.
• This value text must not include the literal character “<”
(directly or through an entity replacement, see later).
• It must not include the character (" or ') used to delimit it.
• The value text may include line breaks and other white space.
• Attribute specifications can have white space around the “=”.
61
Examples from SVG


SVG (Scalable Vector Graphics) is an XML notation for
representing graphical objects.
Examples:
• <rect x="50" y="50" width="100" height="75"/>
An empty element representing a rectangle at position
(50,50), with shape 100 × 75 pixels.
• <g transform='rotate(45)'>
<circle cx='150' cy='50' r='25'/>
<text x='125' y='100'>A Circle</text>
</g>
A group containing a circle and some text. The group as a
whole is rotated 45° about the origin, as specified by the
attribute with name transform and value rotate(45).
62
Possible Display of Examples
63
The Document

It is recommended to start a document entity with an XML
declaration, although technically this is optional.
• An XML declaration must be strictly the first thing in the file—no white
space can go before an XML declaration (a single BOM is allowed).

Every XML document entity must contain exactly one top-level
element called the root element.
• Of course there can be any number of elements nested inside the root
element.


If there is a Document Type Declaration, this must appear before
the root element.
Miscellaneous other things are allowed at the document level
(anywhere between XML declaration and end of file):
• White space
• Comments
• Processing instructions

Anything before the root element is collectively called the prolog.
64
Document Layout
XML declaration (optional)
…optional comments and processing instructions …
Prolog
Document Type Declaration (optional)
…optional comments and processing instructions …
root element (required)
…optional comments and processing instructions …
65
XML Declaration


It is strongly recommended to start any document
entity with an XML Declaration.
The XML declaration has syntax:
<?xml version=Literal …optional declarations… ?>

The version specification and optional declarations have
a syntax similar to attribute specifications on elements
(declared values are quoted the same way).
• The value assigned to version must be 1.0 or 1.1, according
to which version of the XML specification you are following
(1.0 may be prudent, for now).
• The optional declarations are the encoding declaration and
the standalone declaration.
66
Example XML Declaration

Here is a complete (and completely pointless) XML
document with an XML declaration—including
optional parts—and an empty root element:
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<my-root/>
• In general the encoding declaration, if included, specifies the
character encoding scheme.
 This must be an encoding of Unicode, but it may be some
encoding not defined in the Unicode standard.
• Informally a standalone declaration says whether this
document stands alone. See later for details. Meanwhile, “if
in doubt, leave it out”—the default is always safe.
67
On Encoding Declarations

If you are awake, you may wander what is the point of
declaring the character encoding inside a document
encoded according to that scheme—apparently we
can’t read the declaration unless we already know the
scheme!
• The XML specification gives an argument about autodetection of the encoding scheme allowing us to read far
enough into the document to read the contents of the
declaration.
• Debatable how much point there is giving an explicit encoding
declaration—seems more logical to rely on auto-detection of
standard encodings alone, or external metadata?
68
Comments

Comments can appear at the top level of a document, in
the DTD, or in the document content.
• Specification says the parser may or may not pass the text of
the comment to the application.
• XML comments are not white space—placement is more
restricted.

Comments have the syntax
<!-- Text -->
where Text is any text, except it must not contain two
adjacent hyphens, “--”, and it must not end with a
hyphen:
<!-- This--is--not--legal -->
<!-- Neither is this--->
69
Processing Instructions

Processing instructions can appear anywhere that
comments can—they effectively are comments so far as
the XML parser is concerned.
• But the specification requires the parser to pass the text of a
processing instruction to the application.

The syntax is
<?Target Text ?>

where Target is an XML name, and Text is any text,
except that it must not contain the string “?>”.
Processing instructions are a convenience for the
application. They allow application-specific annotation
of the document (example <?php … ?>). The target
name may be declared as a notation (see later).
70
The Document Type
Definition
71
Document Types




In the syntax for the document entity, we saw that the
document type declaration was an optional feature.
This declaration, if present, contains the document type
definition, or DTD.
A validating parser will read the DTD, which should
contain (among other things) declarations of all the
elements and attributes appearing in the body of the
document.
The DTD is required if the parser is validating, but
optional for a non-validating parser
• Even a non-validating parser may read the DTD if it is
present, to look for entity declarations. These will be
discussed later.
72
Document Type Declaration


The document type declaration, if present, appears
before the root element of the document.
The most general form of this declaration contains
three things:
1. The type (i.e. name) of the following root element
2. An identifier for an External DTD Subset
3. An Internal DTD Subset.

Items 2. and 3. are optional.
The DTD itself is either given in an external file
(“external subset”), or “in line” in the document
(“internal subset”), or divided between the two.
73
General Syntax

General syntax is one of:
<!DOCTYPE Name [ Declarations ] >
<!DOCTYPE Name External-ID >
<!DOCTYPE Name External-ID [ Declarations ] >
where Name is the type of the root element, External-ID
is an identifier for an external entity—a separate file
containing the external DTD subset—and Declarations
is the internal DTD subset.
• Syntactically, the form:
<!DOCTYPE Name >
is allowed, but can never yield a valid document. (Why?)
74
External Entity Identifiers



External entity identifiers can occur in a couple of
places: they will be discussed in the next section when
we discuss entity declarations.
Meanwhile the simplest form is “SYSTEM Literal”,
where Literal contains a URI—a file name or URL or
(in principle) a URN.
So a valid document with an external DTD might be:
<!DOCTYPE my-root SYSTEM "my-type.dtd">
<my-root>
…
</my-root>
where my-type.dtd is a local file name.
75
Example DTD

A DTD (subset) contains a series of DTD Declarations.
• Comments and processing instructions can be interleaved
amongst the declarations.

A possible pair of declarations in an internal DTD:
<!DOCTYPE my-root [
<!ELEMENT my-root (my-leaf)>
<!ELEMENT my-leaf EMPTY>
]>
• These declare two types of element: my-root and my-leaf.
• They also specify a very restricted kind of nesting. The only
valid document content would be something equivalent to:
<my-root> <my-leaf/> </my-root>
76
Validation Using dom.Writer

If I save the document type declaration and root
element in a file called “internaldtd.xml”, I can validate
the file with the Xerces parser by using the dom.Writer
sample application as follows:
> java dom.Writer –v internaldtd.xml
• If validation is successful, this simply prints a formatted
version of the input file. If validation fails, you will see error
messages early in the output.
• The –v flag is needed here: validation is “off”, by default.
 “v” must be lower case: “-V” forces no validation.
77
Element Type Declarations

The general form of an element type declaration is:
<!ELEMENT Name Content-Specification >
where
• Name is an XML name, the element type, and
• Content-Specification describes the content allowed in
elements of type Name (i.e. it describes the structure of any
text surrounded by Name tags).
• No element type may be declared more than once.
78
Empty Content Specification

The simplest content specification that can appear in a document
type definition is “EMPTY”. This means elements of this type
never have any content, not even white space!

E.g. the declaration
<!ELEMENT b EMPTY>
allows elements:
<b/>, <b></b>
but not:
<b>Hello</b>, <b><!-- Hello --></b>, <b>
<b>
</b>
</b>,
• The form <b/> is preferred when the element type is declared EMPTY;
<b></b> is preferred when empty content matches another declaration.
79
Parsed Character Data

The content specification “(#PCDATA)” means elements include
only “flat” text—no nested elements are allowed. PCDATA
stands for “Parsed Character Data”.
• Comments and processing instructions are allowed in parsed character
data.

E.g. the declaration
<!ELEMENT a (#PCDATA)>
allows elements:
<a>Hello, hello </a>,
<a>Hello <!-- pause --> hello</a>,
but not:
<a>Hello <b/> hello</a>,
<a> <a>Hello</a> hello </a>.
80
Content Models and Mixed Content

Two sorts of content specification allow for nested
elements:
• If an element may contain only nested elements, with no
interspersed character data, one gives a content model.
 E.g. typical content like this:
<c> <a>Hello</a> <b/> </c>
c contains nested a and b elements, but no directly nested
character data.
• Otherwise, if character data and elements can both appear,
the content is specified (less precisely) as mixed content.
 E.g. typical content like this :
<c> Hello <b/> </c>
c can’t have a content model because character data
“Hello” directly nested in c.
81
Structure of Content Models


A content model is a kind of regular expression, in
which elements as whole are treated as the tokens.
The content model:
(a, b, c)

means the content is a sequence of an a element, a b
element, and a c element, in exactly that order.
The content model:
(a | b | c)
means the content is a choice of an a element, a b
element, or a c element.
82
Composing Content Models

You can compose sequences and choices with appropriate
parentheses, e.g.
((a | b), c)
means a sequence of a c or a sequence b c, and
((a, b) | c)

means a sequence of a b or a single c.
In these expressions you can follow an individual element name
or a parenthesized expression by one of the modifiers ?, *, or +.
• ? Means the term is optional (occurs zero or one times),
• * Means the term can be repeated zero or more times,
• + Means the term can be repeated one or more times, e.g.
(a | b | c)+
means any combination of one or more as, bs and cs
83
Content Model Example


Suppose the format of a simple “report” is: a title,
followed by any mix of paragraphs and figures, followed
by an optional bibliography section.
A suitable document type might start:
<!DOCTYPE report [
<!ELEMENT report
(title, (paragraph | figure)*, bibliography?) >
<!ELEMENT title (#PCDATA)>
<!ELEMENT paragraph (#PCDATA)>
<!ELEMENT figure EMPTY>
<!ELEMENT bibliography (reference)* >
…
]>
84
Example report Data
<report>
<title>Early Use of XML</title>
<paragraph>Recently uncovered documents (see figure)
prove XML was first used by the Incas of
ancient Peru.</paragraph>
<figure source="notafake.jpg"/>
<paragraph>The author is grateful to W3C for making this
research possible.</paragraph>
<bibliography>
…
</bibliography>
</report>
85
Content Model Miscellany



As stated earlier, character data cannot appear in a
(valid) element described by a content model. But
white space, comments, and processing instructions can
be interleaved in an allowed element sequence.
The first character in the top level of the content model
expression must be a left parenthesis: (a*) and (a)* are
allowed as content models, but a* is not.
The specification says a content model must be
deterministic. This fairly technical requirement (needed
for SGML compatibility only) is outlined in the
following few slides.
86
Positions in Content Models

For purposes of this discussion, we label every
appearance of an element type in a content model with
some unique identifier. We call a labeled element type a
position.
• E.g. if the content model is:
(a, (b, c)*, (b | a))
then a labeled version could be:
(a1, (b2, c3)*, (b4 | a5))
which includes positions a1, b2, c3, b4, and a5.
87
“Deterministic”


Imagine reading some input string of elements from the
document, one element at a time, and assume that at the
time each individual element is read we have no
knowledge of what follows it in the document.
A content model is deterministic if, as elements are read
from the document and matched against the content
model, there is never more than one position in the
content model that could match each inputted element.
• This must be true for all possible input strings.

To give a full formal definition, need to define a follows
relation between positions. Won’t do that here.
88
Determinism Examples

The content model
((b1, c2) | (b3, d4))
is not deterministic. Suppose the first element read is a
b. At the time it is read, the b may match against b1 or
against b3.
• The content model (b, (c | d)) is equivalent and deterministic.

The content model
(a1, (b2, c3)*, (b4 | a5))
is not deterministic. Suppose one a element has already
been read and the next element encountered is a b. At
the time b is read, it may match against b2 or b4.
• In this case DTDs provide no equivalent deterministic model.
89
Mixing Content

Many XML formats allow character data and text to be
intermingled in the content of a single element.
• For example XHTML allows a mixture of text and element
markup in the content of its body element (or, say, its p
element).


Neither the parsed character content specification
“(#PCDATA)”, nor a content model specification, will
allow this mixture.
Instead one must use a mixed content specification.
90
Mixed Content Specifications

A mixed content specification looks like:
(#PCDATA | Name1 | Name2 | … | Namen)*
where Name1, Name2, …, Namen are different element
types.
• This may look like a content model, but it is not! Mixed
content specification must appear in the ELEMENT
declaration in exactly the form above.
• You can’t replace any of the Name fields with more general
expressions; the #PCDATA token must appear first in the
parentheses; the right parenthesis must be followed by *.

Nevertheless the matching rules are what the syntax
suggests: a valid element can contain an unrestricted
mix of character data and the listed types of element.
91
Mixed Content Example

Earlier we said that an element c like :
<c> Hello <b/> </c>

cannot be described by a content model.
It can be described by the declaration:
<!ELEMENT c (#PCDATA | b)* >

Note however this allows any sequence of text and b
elements.
• There is no way to specify there must be exactly one piece of
text preceding exactly one b element.
92
ANY Content



The last kind of content specification allowed in an
element declaration is “ANY”.
This is equivalent to a mixed specification naming all
elements declared anywhere in this DTD.
It does not allow appearance of elements that are
undeclared; and all nested elements must be valid
according to their own declarations!
93
Attribute-List Declarations



For a valid document, besides declaring the elements
themselves, you must also declare all attributes that
appear on all elements.
An ATTLIST declaration declares a list of attributes
for a named element type.
For each attribute in the list, the declaration defines
three things:
1. the name of the attribute,
2. the type of values it can be assigned, and
3. whether the attribute has a default value, and if so what it is.
94
Syntax of ATTLIST

Is fairly unstructured—it contains the name of the
element the attributes apply to, then simply a list of
triples:
<!ATTLIST Element-Name
Name1 Type1 Default1
Name2 Type2 Default2
…
>
where Name1, Type1, Default1, etc are the attribute
properties mentioned on the previous slide.
95
Attribute Types

There is a longish list of allowed types for attributes
(not nearly as long as in XML Schema). The
specification subdivides them into:
• string type,
• tokenized types, and
• enumerated types.

The simplest and most general is string type, indicated
by the keyword CDATA in the attribute declaration.
• The value of an attribute is declared with this type can be any
literal string.

Other types will be described shortly.
96
Attribute Defaults

The attribute type is followed by a default rule, one of:
Literal
#FIXED Literal
#REQUIRED
#IMPLIED
• Literal is a default value. The attribute is always logically
present (passed to application), but optionally specified.
• #FIXED modifier means the attribute can only take its default
value (trying to specify something else is invalid).
• #REQUIRED means the attribute must be specified (so no
default is necessary).
• #IMPLIED means the attribute is optional (if unspecified it is
absent, so no default is necessary).
97
Attribute Default Examples

Attribute list declaration:
<!ATTLIST a
val
fix
req
opt

CDATA "nothing"
CDATA #FIXED "constant"
CDATA #REQUIRED
CDATA #IMPLIED>
Instances of element a:
<a val="something" fix="constant“
req="reading" opt="extra"/>
<a req="no experience"/>
<!-- OK: val = “nothing”, fix = “constant”, opt absent. -->
<a fix="variable"/>
<!-- Invalid! fix not “constant” and req unspecified. -->
98
“Tokenized” Attribute Types

In place of “CDATA” we may have:
• NMTOKEN: syntax of attribute value is an XML name token
(defined earlier), e.g. “2004”.
• NMTOKENS: …a list of name tokens.
• ID: attribute is an identifier for this element.
• IDREF: attribute is a reference to another element in this
document.
• IDREFS: …a list of references to other elements.
• ENTITY: attribute is a reference to an unparsed external.
Entity (see next section).
• ENTITIES: …a list of references to unparsed external entities.

Items in a “list” are separated by white space.
99
Element Identifiers

The value given to an attribute with type ID should
follow the syntax of an XML name. This name acts as
an identifier for the element instance on which it
appears. For validity:
• For each element type in the DTD, there should be at most one
attribute declared with type ID.
• All identifiers on all element instances in the document must
be different (regardless element type).

For validity, the value given to an attribute of type
IDREF should be an identifier for an element appearing
somewhere in the document.
100
Example

Assume the attribute name on element type agent is
declared to have type ID, and the attribute boss is
declared to have type IDREF.
<agent name="Alice" boss="Alice"/>
<agent name="Bob" boss="Alice"/>
<agent name="Carole" boss="Alice"/>
Alice
<agent name="Dave" boss="Bob"/>

This document captures
the hierarchy illustrated
on the right. Can use this
technique to represent
general graphs.
Bob
Carole
Dave
101
Enumerated Attribute Types

The type field in an ATTLIST declaration may have the
form:
(Token1 | Token2 | … | Tokenn)

where each Token is an XML name token.
This says that the specified value of the attribute must
be one of these token values.
• Example: attribute list declaration:
<!ATTLIST a
color (red | green | blue | white) "white">
• Instances of element a:
<a color="red"/>
<!-- OK. -->
<a color="black"/>
<!-- Invalid! -->
102
Notation Attribute Types

The type field in an ATTLIST declaration may have the
form:
Notation (Name1 | Name2 | … | Namen)

where each Name is declared elsewhere in the DTD as a
notation (see next section).
The handling of this type is similar to enumeration
types.
103
Attribute Declaration Miscellany



Attributes for a single element type may all appear in a
single ATTLIST declaration, or they may be divided
over several ATTLIST declarations.
It is allowed to declare the same attribute more than
once, but any declarations after the first are ignored.
It is allowed (but pointless) to declare an attribute for
an undeclared element type.
104
Attribute Order


Attribute specifications can appear in element tags in
any order.
The same attribute cannot be specified more than once
on a single element.
105
Entities, References, and
other Processing Issues
106
Collecting Things Together



So far we have described an ideal subset of XML in
which a document contains DDT, elements, attributes,
and character data, all laid out linearly.
In reality fragments of content and DTD may be
defined in places other than where they ultimately
appear (perhaps in other files), and portions of text
may need special processing before they are made
available to the application.
To complete the discussion we must cover:
•
•
•
•
Character references and entity references.
Internal and external entity declarations, and notations.
CDATA sections.
Conditional sections.
107
Character References



Character references can be viewed as an “escape”
mechanism that allows us to include specially-treated
or hard-to-type characters in the XML document.
They include the Unicode code point for a character,
taking either of the forms “&#dd…;” or “&#xXX…;”,
where ds are decimal digits and Xs are hexadecimal
digits.
For example:
• &#38; or &#x26; represents “&” (ampersand).
• &#60; or &#x3C; represents “<” (left angle bracket).
• &#963; or &#x3A3; represents “Σ” (large Greek sigma).

One application is for including the literal characters
“<” or “&” in parsed character data.
108
CDATA Sections

CDATA sections provide a way of including a section of
character data in an XML document.
• The section can include “<” and “&” characters without
escaping—in a CDATA section these characters have no
special significance (so “markup” syntax is ignored).

The syntax is
<![CDATA[ Text ]]>
where Text is any text, except that it must not contain
the string “]]>”.
• One application is for including scripting in XML—e.g.
JavaScript uses < and & for operators.
• You cannot include any characters generally forbidden in
XML: can’t put raw “binary” data in a CDATA section!
109
Entity References


A character reference includes a single Unicode
character in the document. An entity reference includes
the content of some “entity”, which may be the contents
of an external file.
The simple syntax is:
&Name;

where Name is an XML name.
The name of the entity, Name, must have been declared
in the document DTD. There are just five exceptions to
this rule.
110
Predefined Entities

As a convenience the entities amp, lt, gt, apos, and quot
are considered predefined.
• You may declare them in a DTD, but it isn’t necessary.

They must contain the single-character values:
•
•
•
•
•


&amp; expands to “&” (ampersand).
&lt;
expands to “<” (left angle bracket, or less than).
&gt; expands to “>” (right angle bracket, or greater than).
&apos; expands to “'” (single quote, or apostrophe).
&quot; expands to “"” (double quote).
Provide a more convenient way of including reserved
characters.
Note these are entity references, not character
references (affects processing in some contexts).
111
A Hofstadteresque XHTML Example
<html>
<body>
The source of this document is:
<pre>
&lt;html&gt;
&lt;body&gt;
The source of this document is:
&lt;pre&gt;
&amp;lt;html&amp;gt;
…
&amp;lt;html&amp;gt;
&lt;/pre&gt;
&lt/body&gt;
&lt/html&gt;
</pre>
</body>
</html>
112
Declaring Entities

Entities are defined in the DTD by an ENTITY
declaration with the syntax:
<!ENTITY Name Definition >

Here Name is of course the name of the declared
entity.
This general form can declare:
1. Internal entities
2. Parsed external entities
3. Unparsed external entities
Only 1. and 2. can be included by an entity reference.
•
Later we will see another form of the ENTITY declaration
that defines parameter entities.
113
Internal Entities


The simplest kind of entity is an internal entity. In this
case the Definition is just a literal string, and the entity
behaves like a kind of macro.
The string can contain character data and markup.
E.g.:
<!ENTITY
<!ENTITY
<!ENTITY
<!ENTITY
<!ENTITY
me "Bryan" >
lt "&#38;#60;" >
Sigma "&#x3A3;" >
flag "Stars &amp; Stripes" >
icon '<image xlink:href="icon.jpg" />' >
These declarations allow the shorthand forms “&me;”,
“&lt;”, “&Sigma;”, “&flag;”, “&icon;” in the body of the
document.
114
Replacement Text for Internal Entities


Character references (and also parameter entity
references, see later) appearing in the entity definition
are expanded when the declaration is processed. So the
replacement text for lt is “&#60” and for Sigma is “Σ”.
References to other entities appearing in an entity
definition are not expanded at the time the declaration
is processed (see next slide). The replacement text for
flag is “Stars &amp; Stripes”.
115
Expansion of General Entities



When the parser encounters an entity reference in the
content of a document, the reference is expanded to the
replacement text for the entity.
The parser then resumes processing, starting at the
beginning of the inserted text. If the replacement text
contains further entity references, these are replaced in
turn, as they are encountered.
An entity reference must not expand to fragmentary
markup. For example, the replacement text may
include complete elements, but not isolated start tags,
or isolated “<” characters.
116
External Entities


For an external entity the Definition appearing in the
ENTITY declaration is an external ID—same syntax as
in DOCTYPE declarations.
There are two forms (for parsed external entities):
<!ENTITY Name SYSTEM URI-Literal >
<!ENTITY Name PUBLIC Public-ID-Literal
URI-Literal >

The first form is fairly self-explanatory: URI-Literal is
typically an absolute or relative URL.
• If relative, it is relative to the location of the document entity
or external entity containing the ENTITY declaration.
117
References to External Entities

External entities may be referenced anywhere markup
can appear in document content, except in the literal
value of an attribute.
• Internal entities can be referenced in attribute specifications.
118
Public Identifiers

Where an external entity is likely to widely used one
can give a PUBLIC identifier.
• This acts as a logical identifier, something like a URN.
• XML standard itself doesn’t specify a syntax for public
identifiers, but XML-based standards usually follow the
SGML format for Formal Public Identifiers.

For example the DOCTYPE declaration for an
XHMTL 1.1 document should be:
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
"http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
• When an document type declaration or an entity declaration
gives a public identifier, this must be followed by a URI,
which acts as a fall back for locating the entity.
119
Format of Parsed External Entities

A parsed external entity may optionally start with a text
declaration. The rest of the file should be Content, such
as may appear in an XML element.
• The replacement text for an external entity is the unprocessed
content (minus text declaration).

An external DTD subset may optionally start with a
text declaration. The rest of the file should be
Declarations, such as may appear in an internal DTD
subset.
• We will see that external DTD subsets allow a couple of
processing options beyond those allowed in internal subsets.
120
Text Declaration


It is recommended to start a parsed external entity with
a text declaration.
The text declaration has syntax:
<?xml …optional version declaration… encoding=Literal ?>
i.e. syntax is identical to an XML declaration, except
that now it is the encoding declaration that is
mandatory (and no standalone declaration is allowed).
• It is allowed for the entity character encoding to be different
from the document that references it.
121
XHTML Example

A document entity:
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
"http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd" [
<!ENTITY myfooter SYSTEM "myfooter.xml" >
]>
<html>
<body> This is one page in a complicated Web site.
&myfooter; </body>
</html>

An external entity “myfooter.xml”:
<hr/>Maintained by
<a xlink:href="mailto:[email protected]">
Bryan Carpenter</a>.
122
Combining DTD Subsets


The example above is illustrative—sadly no known
Web browser will recognize this kind of DTD.
In principle it shows how one can use both an external
subset and an internal subset in the same DTD.
• We will see shortly that one can go further with this using
parameter entities.
• Although the internal subset appears after the reference to the
external subset, the internal subset is processed first. If there
are conflicting declarations of entities, the internal subset
takes precedence.
• In general if an entity or attribute is declared more than once,
declarations after the first are ignored.
123
Unparsed Entities



Unparsed entities are declared in a similar way to
parsed entities, but used in a completely different way.
The XML parser does not include the contents of the
entity in the XML data, it just forwards a reference to
the application, e.g. as a URI.
An unparsed entity must be annotated with information
about what type of content it holds (or, equivalently,
what kind of application can process it).
• This notation is also passed to the application.

We give the syntax on the following slides—there is a
quite lot of syntax with modest payoff.
124
Notation Declarations

Before using a notation to label an unparsed entity, its
notation is declared. Declaration syntax is quite free,
e.g. the following are legal:
<!NOTATION myformat >
<!NOTATION jpeg SYSTEM "image/jpeg" >
<!NOTATION perl SYSTEM "/usr/bin/perl" >
<!NOTATION Name PUBLIC Public-ID >
• According to the specification, public and system identifiers
may “locate a helper application capable of processing data”.
• Some authors interpret that they may be, say, MIME types.
• A declared notation may also be the target of a processing
instruction, or specified in the value of an attribute of type
NOTATION or NOTATIONS.
125
Declaring and Using Unparsed Entities

Declaration similar to parsed external entities, except it
ends with “NDATA Name” where Name is a notation.
E.g.:
<!ENTITY notafake SYSTEM "notafake.jpg" NDATA jpeg >

The only way to reference a declared unparsed entity is
in an attribute declared of type ENTITY or ENTITIES,
e.g. if we have:
<!ATTLIST figure source req ENTITY #REQUIRED >
then the name notafake can finally be used as follows:
<figure source="notafake"/>
126
Parameter Entities

General entity references are expanded when they
appear to in the document content. They are not useful
for abstracting sections within a DTD.
• The only time general entities are expanded while processing
a DTD is if they appear in the default value of an attribute.

There is a separate “macro-expansion” mechanism that
is designed for use within DTDs. This uses parameter
entities.
127
Declaration and Reference

Declaration of parameter entities follows a syntax
similar to parsed general entities, except the name is
preceded by a “%” sign, e.g.:
<!ENTITY % my-dec '<!ENTITY me "Bryan" >' >
This declares the internal parameter entity my-dec. Its
value is the declaration of the general entity me.
 To actually insert the declaration of me, later in the
DTD, one would use a parameter entity reference, which
looks like an entity reference except it uses “%”, e.g.:
%my-dec;
128
External Parameter Entities

Like general entities, parameter entities may be
external, e.g.
<!DOCTYPE my-root [
<!ENTITY % my-type SYSTEM "my-type.dtd">
%my-type;
]>
is for all intents and purposes equivalent to:
<!DOCTYPE my-root SYSTEM "my-type.dtd">
129
Attribute Groups


Parameter entities can be used to modularize a DTD,
for example by splitting it over multiple files which may
be selectively included (c.f. conditional sections,
described shortly).
Another common use is to collect together lists of
related attribute definitions that appear in more than
one attribute list, e.g.
<!ENTITY % image-atts
"source CDATA #REQUIRED
height NMTOKEN #IMPLIED">
…
<!ATTLIST image %image-atts; >
<!ATTLIST figure %image-atts; caption CDATA #IMPLIED >
130
Restrictions

Some reasonable restrictions are placed on references
to parameter references, to ensure their replacement
text is a “complete” fragment of DTD syntax.
• This is similar to the requirement that a parsed general entity
must expand to a “complete” fragment of markup.

There is a more ad hoc requirement that in an internal
DTD subset any parameter references must expand to
one or more complete DTD declarations
• Example of previous slide would be illegal in internal subset
because %image-atts; isn’t a complete declaration.
• This restriction does not apply inside external parameter
entities referenced from the internal subset.
131
Parameter Entity Expansion


When a parameter entity is referenced anywhere in the
DTD except in the literal text of another entity
declaration, extra space characters are added at the
ends of the replacement text.
E.g., this will not work:
<!ENTITY % my-prefix "MYAPP_" >
<!ELEMENT %my-prefix;Root (#PCDATA) >
as element “name” expands to “ MYAPP_ Root”, with
illegal space. Following is OK:
<!ENTITY % my-prefix "MYAPP_" >
<!ENTITY % my-root "%my-prefix;Root" >
<!ELEMENT %my-root (#PCDATA) >
132
Conditional Sections

Two special kinds of section can appear in an external
DTD subset:
<![IGNORE[ Text ]]>
<![INCLUDE[ Declarations ]]>

These are designed for use in the following the idiom:
<![%my-control;[ … conditional Declarations … ]]>
where my-control is the name of a parameter entity
defined earlier with value IGNORE or INCLUDE.
• For example my-control can be declared in the internal
subset, affecting what parts of an external DDT are effective.
• The IGNORE section has the unusual property that other
sections can nest within it. Thus also convenient for
“commenting out”.
133
Processing Miscellany

The XML parser will present character data extracted
from element bodies and attributes to the application.
It does this after suitable processing. E.g.
• Character references may be replaced by actual character.
• CDATA sections may be replaced by their contents.
• Entity references may be expanded.

Various other normalizations take place.
• Every CR LF sequence, or CR not followed by LF, is replaced
by a single LF.
• A validating parser will normalize the values of attributes
according to their type (next slide).
134
Normalization of Attribute Values

For every attribute value, the following normalizations
are applied by a validating parser:
• Convert every literal white space character (Line Feed, etc) to
a space character (#x20).
• Expand character references
• Expand entity references (and apply normalizations above to
replacement text).

For attributes that have type other than CDATA, all
leading or trailing spaces are then removed. Also
sequences of spaces are replaced by single spaces.
• Note that these transformations generally will not be applied
by a non-validating parser (so the application will see
different data).
135
Standalone Documents

Various markup declarations affect the data passed to
the application.
• E.g. declarations of entities, default values of attributes, types
of attributes affecting normalization,…


The XML declaration may include standalone="yes" if
all those markup declarations are in the internal DTD
subset (and moreover not in parameter entities).
Note this restricts only external declarations, not
external entities: somewhat mysteriously, a document
that is “standalone” can have references to external
entities, provided they are declared internally.
136
xml:space and xml:lang

If these attributes are used or declared, the
specification restricts what values they may take.
• xml:space may only be declared with an enumerated type
restricted to one or both of “default” and “preserve”, and
may only be specified with one or other of these values.
• xml:lang may only be specified with a value that is a standard
IETF RFC 3066 language tag (see also ISO-639).

Intended uses are to specify handling of white space,
and the natural language of the document. This
handling is up to the application: a parser will likely
forward these the same as other attributes, and they
don’t affect its handling of text.
137
Conclusion


We have covered everything in the basic XML
specification, but there is still a long way to go with
XML.
Coming up:
•
•
•
•
The Document Object Model
Namespaces in XML
Introductory XPath
XML Schema
138
The Document Object Model:
Programming Interfaces
139
The Document Object Model



The Document Object Model or DOM is a set of W3C
specifications that define standard ways to access parts
of an XML or HTML document from within a program.
This addresses, for example, how a Java program
extracts the data from an XML document, after the
document has been processed by an XML parser.
But the origins of the DOM lie in JavaScript
programming for interactive and dynamic Web pages.
140
Dynamic HTML Example
<html>
<head>
<script language="javascript">
function appendText() {
paraNode = document.getElementsByTagName("p") [0] ;
textNode = paraNode.firstChild ;
textNode.nodeValue += " Ouch!" ;
}
</script>
</head>
<body>
<p>Hello.</p>
<form>
<input type="button" value="Push Me" onclick="appendText()"/>
</form>
</body>
</html>
141
The DOM from JavaScript


The DOM represents an XML (or HTML) document as
a series of nodes.
In the example, the JavaScript method appendText() is
called when the user clicks on the button labeled “Push
Me”.
• This method identifies the DOM node representing the <p>
element, extracts the nested text node, and modifies the data
(the text) associated with that text node.
• The function getElementsByTagName() is a method defined
in the DOM, and firstChild and nodeValue are node
properties in the DOM.
142
A Document is a Tree

Here is how
an example
fragment of
HTML and
can be
thought of
as a tree.
143
DOM Nodes


In the DOM, one builds the document tree out of a set
of Node objects
Each Node object has a set of capabilities (properties
and methods) and implements specific interfaces.
Node
Node
Node
Node
Node
Node
Node
Node
Node
Node
Node
Node
Node
Node
Node
Node
Node
144
Node Types in Core DOM


DOM APIs are defined as a series of interfaces (this term
meant in a similar way to how Java uses it).
Node itself is an interface, and the following interfaces
extend Node:
• Document: node representing entire document, including
prolog.
• DocumentFragment: convenience node, representing an
incomplete document.
• Element: node representing an element.
• Attr: node representing an attribute attached to an element.
• Text: node representing “character data”.
• Comment: node representing a comment.
145
Further Node Types

Nodes on previous slide cover requirements of HTML.
Following extra node types are needed to represent
general XML:
• EntityReference: node representing an entity reference.
• Entity: node representing content of a parsed external entity.
• ProcessingInstruction: node representing a processing
instruction.
• CDATASection: node representing a CDATA section.
• Notation: node representing a notation declared in the DTD.
146
Constants
The
Node
Interface
Properties
Methods
147
Remarks

DOM interfaces are traditionally defined in the OMG’s
Interface Definition Language.
• IDL interfaces can be mechanically converted to interface
definitions for languages like Java, ECMAScript, C++, …

By design, the Node interface is sufficiently generic
that any DOM tree can be manipulated using the
methods and properties of this interface alone.
• Use of the more specialized derived interfaces—Element,
Attribute, etc—is at programmer option. Those interfaces
provide more type-safety, and various type-specific
convenience methods and properties.
148
nodeName, nodeValue Properties


Each node type has different rules for values of some of the
properties—most importantly nodeName and nodeValue.
The attributes property is only relevant for an element node.
Node Type
149
Node Children

The following node types may have children, as
indicated, in the DOM tree:
• Document: DocumentType, Element (maximum of one),
ProcessingInstruction, Comment.
• DocumentFragment: Element, Text, CDATASection,
EntityReference, ProcessingInstruction, Comment.
• Element: Element, Text, CDATASection, EntityReference,
ProcessingInstruction, Comment.
• Attr: Text, EntityReference.
• EntityReference: Element, Text, CDATASection,
EntityReference, ProcessingInstruction, Comment.
• Entity: Element, Text, CDATASection, EntityReference,
ProcessingInstruction, Comment.
150
The Document Interface

The Document interface has some indispensable factory methods
for creating other node types:
151
Levels of the DOM

W3C recognizes 4 stages of evolution of the DOM,
called levels:
• Level 0: Legacy HTML “DOM” features from Navigator 3.0
and IE 3.0. No W3C specification.
• Level 1: Specification completed 1998. Basic tree structure for
HTML and XML:
152
Level 2 DOM

Specification completed 2000. Notably added support for Events.
Also XML Namespaces, DOM representation of Cascading Style
Sheets, and new facilities for manipulating the tree.
153
Level 3 DOM

Specification still in progress. It will add, for example, support
for XPath search operations on a DOM tree, and formally define
the mapping between DOM and XML Infoset. Load/Save
standardizes parsing/serialization APIs.
154
Using Xerces


Earlier we used the dom.Writer sample application
from the Xerces project to validate XML documents.
In the following slides we illustrate how to use the
Xerces parser from your own Java program.
• We will use the conventional Java JAXP API to control
parsing.
• In future the Load/Save features of Level 3 DOM are likely to
be the preferred API (Xerces already provides an
experimental implementation).
• To control certain aspects of parsing, it may be necessary to
use the org.apache.xerces.parsers.DOMParser
implementation class directly.
155
Checking Well-Formedness
import javax.xml.parsers.DocumentBuilder;
import javax.xml.parsers.DocumentBuilderFactory;
import org.w3c.dom.Document;
import java.io.File;
public class MyChecker {
public static void main(String [] args) throws Exception {
File source = new File(args [0]) ;
DocumentBuilderFactory factory =
DocumentBuilderFactory.newInstance();
factory.setNamespaceAware(true) ;
DocumentBuilder builder = factory.newDocumentBuilder();
Document document = builder.parse(source) ;
}
}
156
Remarks

In JAXP one first obtains a parser factory, then uses the
factory to create a parser instance (called a document
builder in JAXP).
• Behaviors of the parser (e.g. whether namespace-aware,
validating/non-validating, …) are controlled by setting
properties of the factory before creating the instance.

Actual parsing is done by the parse() method, which
returns a DOM instance (Document node).
• Example above does nothing if document is well-formed;
prints an exception if it is not.
157
Validating
public static void main(String [] args) throws Exception {
File source = new File(args [0]) ;
DocumentBuilderFactory factory =
DocumentBuilderFactory.newInstance();
factory.setNamespaceAware(true) ;
factory.setValidating(true) ;
DocumentBuilder builder = factory.newDocumentBuilder();
builder.setErrorHandler(new MyErrorHandler()) ;
Document document = builder.parse(source) ;
}
158
Simple Error Recovery
import org.xml.sax.* ;
class MyErrorHandler implements ErrorHandler {
public void warning(SAXParseException e) throws SAXException {
System.out.println(e.getMessage());
}
public void error(SAXParseException e) throws SAXException {
System.out.println(e.getMessage());
}
public void fatalError(SAXParseException e) throws SAXException {
System.out.println(e.getMessage());
System.exit(1);
}
}
159
Processing the DOM


The two previous examples ended by creating a
Document object, then did nothing with it.
The following example summarizes several important
Java DOM classes and methods in one recursive
method that extracts all useful information in the
elements, attributes, and text nodes of a document.
• Assumes a declaration like
import org.w3c.dom.* ;
is in effect.
• After parsing the document, this method could be invoked
from main() by:
process(document.getDocumentElement()) ;
160
Recursive Processing of DOM
static void process(Node node) {
System.out.println(node.getNodeName()) ;
switch(node.getNodeType()) {
case Node.ELEMENT_NODE:
NamedNodeMap attributes = node.getAttributes() ;
for(int i = 0 ; i < attributes.getLength() ; i++)
process(attributes.item(i)) ;
NodeList children = node.getChildNodes() ;
for(int i = 0 ; i < children.getLength() ; i++)
process(children.item(i)) ;
break ;
case Node.TEXT_NODE:
case Node.ATTRIBUTE_NODE:
System.out.println(node.getNodeValue()) ;
}
}
161
Interactive XML





We have been discussing use of the DOM in Java as a convenient
intermediate representation for handling XML data.
This is different from the original application of the DOM,
illustrated in our introductory dynamic HTML example.
In that application, JavaScript event handlers manipulate the
DOM embedded in a Web browser, dynamically altering the text
displayed by the browser.
This technique can also be applied to XML documents that have
a visual representation, for example to Scalable Vector Graphics
documents.
Following screen captures are from www.svgarena.org.
162
XML Chess
163
XML Solitaire
164
Namespaces in XML
165
Motivation



XML DTDs allow us to define a set of customized
element, attribute, and entity names to support a
particular application.
But in a document of any complexity one may want to
mix and match markup from different application
domains.
In particular in a Web page authored using XHTML
tags one might want to embed pictures represented
with SVG elements; in a technical Web page one might
want to embed mathematical formulae represented
using MathML elements.
166
Namespaces



Typically different XML vocabularies are designed by
different committees, and it is impractical to avoid
name clashes.
In general we can’t simply merge the DTDs—unclear
this would be desirable anyway: loss of modularity.
The Namespaces in XML specification address this issue
by allowing each application to define names in a
particular namespace. A single XML document can
then include markup from more than one namespace.
167
Relation to the XML Spec



The namespaces specification was released a year after
the XML 1.0 specification.
It is formally 100% compatible with the XML
specification—it “merely” puts some restrictions on the
document form.
Significant practical issues arise. DTDs—set in stone in
the XML/SGML standards—are not an ideal match for
namespaces: DTDs can be used with namespaces, but it
takes some hacks.
• The likely intention was to move away from DTD validation
towards XML Schema validation, which is fully namespace
compatible. So far this transition is incomplete?
168
Qualified Names


The simple general idea is that names occurring in
document instances may be prefixed in a way that
identifies their name space.
Here is an (imaginary!) example including SVG tags in
an XHTML document:
<body>
<h1>What are circles really like?</h1>
This is what a circle really looks like:
<object>
<svg:svg>
<svg:circle cx='150' cy='50' r='25'/>
</svg:svg>
</object>
</body>
169
Name Syntax



In the namespaces specification use of the colon, “:”, in
XML names is reserved.
Namespaces qualified names have at most one colon; if
there is a colon it separates the prefix from the local
part:
Prefix:Local-part
In the example above we had qualified SVG names
svg:svg
svg:circle
Here the prefixes are svg and the local parts are svg
and circle respectively. HTML names had no prefix,
only local parts.
170
Namespace Names

Each namespace (e.g. the XHTML namespace or the
SVG namespace) itself has a “name”.
• This is neither the prefix appearing on qualified names, nor in
general a legal XML name!

Instead a namespace name is an IRI.
• An IRI is an Internationalized Resource Identifier: it is a
generalization of a URI allowing non-ASCII characters.

E.g. the name of the XHTML 1.0 namespace is:
http://www.w3.org/1999/xhtml
and the name of the SVG 1.1 namespace is:
http://www.w3.org/2000/svg
171
Use of URIs




The use of URIs (or IRIs) for namespace names is quite
confusing.
Though namespace names follow the syntax of resource
identifiers, the namespace is not a resource; there need
not be any resource at the location—all that is relevant
is the sequence of characters in the URI. This string of
characters identifies the namespace.
One day this odd situation may be resolved to allow
applications to automatically find schema at namespace
URIs—see for example www.rddl.org.
For now namespace names and schema locations are
two independent IRIs.
172
Defining Prefixes

In principle an instance document can choose any
convenient prefix to use within that document
• Though an existing DTD may restrict the choice.

A prefix is declared in an attribute named xmlns:Prefix.
The scope is the element the declaration appears on,
and any nested elements. Value specified for attribute
must be the namespace name. E.g.:
<object xmlns:svg='http://www.w3.org/2000/svg' >
<svg:svg>
<svg:circle cx='150' cy='50' r='25'/>
</svg:svg>
</object>
173
Default Namespaces


If a “vocabulary” has been defined in a namespace, a
document instance must acknowledge this, even if
prefixes aren’t needed.
Use the attribute xmlns to declare a default namespace.
E.g. all XHTML documents should start:
<html xmlns="http://www.w3.org/1999/xhtml" >
<head>
…
</html>
174
Scope of Declarations

So far as the namespaces specification is concerned:
• Declarations of prefixes and default namespaces can appear
on any elements in a document. Scope is the declaration is
limited to the element it appears on.
• Different namespaces can be the default in different elements.
• The same prefix could be used to represent different
namespaces in different parts of the document, or vice versa
(these are bad ideas!)

DTDs may limit these possibilities.
Default namespace declarations don’t affect attribute
names. Interpretation of an un-prefixed attribute name
is determined by the element it is attached to.
• Means attribute is in same namespace as element? No!
175
Namespaces and DTDs



DTDs are not “namespace aware”. So far as DTDs are
concerned a qualified name with a colon is just an
atomic XML name.
A partial workaround is to make all names introduced
in DTD declarations parameter entity references, then
factor out the prefix as a single parameter entity that
can be set in a document’s internal DTD subset.
This is ugly, and it doesn’t solve all problems.
• E.g. DTDs still won’t recognize an equivalence between
names in the same namespace, if they have the same local
name but different prefixes: so far as DTDs are concerned
they are different XML names.
176
Namespaces and XML Schema


Later lectures discuss XML Schema—one of the
alternative validation frameworks to DTDs—in detail.
You will probably find the issues associated with
namespaces become much more concrete once XML
Schema (which are fully namespace-aware) are
understood.
177
Elementary XPath
178
XML Path Language

The XML Path Language, or XPath, is a language for
addressing nodes of an XML document.
• Understand “nodes” as in the DOM (although technically there
are minor differences).


XPath is an important part of XML Schema, XSLT, and
XPointer, and is used as a query language in some XML
databases.
In simple cases an XPath expression looks like a UNIX
path name, with nested directory names replaced by
nested element names, so for example:
• “/” corresponds to the document node.
• Expressions may be absolute (relative to /) or relative to some
context node.
179
Simple Examples

The XPath:
/planets

represents a document element called planets.
The XPath:
/planets/planet/density

represents the set of all elements named density that are
directly nested in any element named planet that is
directly nested in a document element planets.
The XPath:
/planets/planet/*
represents all elements directly nested in an element
planet directly nested in document element planets.
180
<?xml version="1.0" encoding="UTF-8"?>
<planets>
<planet>
<name>Mercury</name>
<mass>0.0553</mass>
<day units="days">58.65</day>
<radius units="miles">1516</radius>
<density>0.983</density>
Document is a simplified </planet>
version of an example
<planet>
<name>Venus</name>
given the in the book
<mass>0.815</mass>
Inside XML, by Holzner.
<day units="days">116.75</day>
<radius units="miles">3716</radius>
Highlighted node is
<density>0.943</density>
evaluation of
</planet>
<planet>
/planets
<name>Earth</name>
<mass>1</mass>
<day units="days">1</day>
<radius units="miles">3960</radius>
<density>1</density>
</planet>
</planets>
Example
181
<?xml version="1.0" encoding="UTF-8"?>
<planets>
<planet>
<name>Mercury</name>
<mass>0.0553</mass>
<day units="days">58.65</day>
<radius units="miles">1516</radius>
<density>0.983</density>
</planet>
<planet>
Highlighted nodes are
<name>Venus</name>
evaluation of
<mass>0.815</mass>
<day units="days">116.75</day>
/planets/planet/density
<radius units="miles">3716</radius>
<density>0.943</density>
</planet>
<planet>
<name>Earth</name>
<mass>1</mass>
<day units="days">1</day>
<radius units="miles">3960</radius>
<density>1</density>
</planet>
</planets>
Example
182
Example
Highlighted nodes are
evaluation of
/planets/planet/*
<?xml version="1.0" encoding="UTF-8"?>
<planets>
<planet>
<name>Mercury</name>
<mass>0.0553</mass>
<day units="days">58.65</day>
<radius units="miles">1516</radius>
<density>0.983</density>
</planet>
<planet>
<name>Venus</name>
<mass>0.815</mass>
<day units="days">116.75</day>
<radius units="miles">3716</radius>
<density>0.943</density>
</planet>
<planet>
<name>Earth</name>
<mass>1</mass>
<day units="days">1</day>
<radius units="miles">3960</radius>
<density>1</density>
</planet>
</planets>
183
Types of XPath Expressions



In full generality XPath expressions can evaluate to
various kinds of thing (including numbers, Booleans,
and strings) but one is usually interested in expressions
that evaluate to node-sets.
A node-set is a collection of nodes in an XML document.
In general he nodes that can appear in a node-set
include:
• element nodes, attribute nodes,
• text nodes (plain text children of some element), root nodes,
namespace nodes, processing instruction nodes and comment
nodes.

We will only discuss the first two cases.
184
Location Paths and Location Steps


The most important kind of XPath expression is
the location path.
In general a location path consists of a series of
location steps separated by the slash “/”.
• Compare to a UNIX path, where the individual “step”
is the name of an immediately nested subdirectory or
file, or a wildcard (*), or a move up into the parent
directory (..), etc.
185
Steps and Context Nodes


While a UNIX path may be relative to some current
directory, an XPath expression is generally evaluated
relative to some context node.
Starting from this context node of the path, the location
path takes a series of steps in various possible
“directions”, e.g.
• into the set of child elements of the current element,
• into the set of attributes of the current element,
• into the set of siblings of the current element, etc.


This “direction” is called the axis of the step.
Each individual step has its own context node,
determined by preceding steps in the path.
186
Syntax for Location Steps

The commonest example of a location step—analogous
to a UNIX directory name—is an XML element name.
• This should be the name of an element that is an immediate
child of the context node.

Recalling this example:
/planets/planet
it has two steps, planets and planet.
• Actually this is an example of what is called abbreviated syntax.
In the full, unabbreviated, syntax, that location path would be:
/child :: planets/child :: planet
child being the name of the axis. In this brief exposition we
only cover abbreviated syntax.
187
Abbreviated Syntax for Steps

The location step Name selects some element nodes: the
children of the context node called Name.
• “*” selects all children of the context node.



The location step “.” represents the context node.
The location step “..” represents the parent node of the
context node.
The location step @Name represents an attribute
node—the attribute of the context node named Name.
• @* selects all attributes of the context node.

A blank location step effectively selects the context node
or any descendant element, e.g.
• //Name selects all Name elements in the document.
• .//Name selects all Name descendants of the context node.
188
<?xml version="1.0" encoding="UTF-8"?>
<planets>
<planet>
<name>Mercury</name>
<mass>0.0553</mass>
<day units="days">58.65</day>
<radius units="miles">1516</radius>
<density>0.983</density>
</planet>
<planet>
Highlighted nodes are
<name>Venus</name>
evaluation of
<mass>0.815</mass>
<day units="days">116.75</day>
[email protected]
<radius units="miles">3716</radius>
<density>0.943</density>
</planet>
<planet>
<name>Earth</name>
<mass>1</mass>
<day units="days">1</day>
<radius units="miles">3960</radius>
<density>1</density>
</planet>
</planets>
Example
189
Example
Highlighted nodes are
evaluation of
//mass
<?xml version="1.0" encoding="UTF-8"?>
<planets>
<planet>
<name>Mercury</name>
<mass>0.0553</mass>
<day units="days">58.65</day>
<radius units="miles">1516</radius>
<density>0.983</density>
</planet>
<planet>
<name>Venus</name>
<mass>0.815</mass>
<day units="days">116.75</day>
<radius units="miles">3716</radius>
<density>0.943</density>
</planet>
<planet>
<name>Earth</name>
<mass>1</mass>
<day units="days">1</day>
<radius units="miles">3960</radius>
<density>1</density>
</planet>
</planets>
190
<?xml version="1.0" encoding="UTF-8"?>
<planets>
<planet>
<name>Mercury</name>
<mass>0.0553</mass>
<day units="days">58.65</day>
<radius units="miles">1516</radius>
<density>0.983</density>
</planet>
<planet>
If the context node is the
<name>Venus</name>
second <planet/> element,
<mass>0.815</mass>
the Highlighted nodes are
<day units="days">116.75</day>
<radius units="miles">3716</radius>
evaluation of
<density>0.943</density>
[email protected]
</planet>
<planet>
<name>Earth</name>
<mass>1</mass>
<day units="days">1</day>
<radius units="miles">3960</radius>
<density>1</density>
</planet>
</planets>
Example
191
Unions


The operator “|” forms the union of two node sets.
For example:
//radius | //density
represents the set of all radius elements and all density
elements in the document.
192
Example
Highlighted nodes are
evaluation of
//radius | //density
<?xml version="1.0" encoding="UTF-8"?>
<planets>
<planet>
<name>Mercury</name>
<mass>0.0553</mass>
<day units="days">58.65</day>
<radius units="miles">1516</radius>
<density>0.983</density>
</planet>
<planet>
<name>Venus</name>
<mass>0.815</mass>
<day units="days">116.75</day>
<radius units="miles">3716</radius>
<density>0.943</density>
</planet>
<planet>
<name>Earth</name>
<mass>1</mass>
<day units="days">1</day>
<radius units="miles">3960</radius>
<density>1</density>
</planet>
</planets>
193
XPath and Namespaces

If the elements or attributes named an XPath
expression belong to a namespace, then in general those
names must have a namespace prefix.
• You cannot use a default namespace in an XPath expression.

When XPath expressions are embedded in XML
documents (which is normal), the prefixes for the
qualified names are scoped in the usual way—by
namespace declarations (i.e. xmlns:Prefix attribute
specifications) in the XML.
194
Summary

In its full generality, XPath is a fairly complex (but
powerful) language for addressing subsets of the nodes
of an XML document.
• It incorporates extensive computational and filtering
functionalities that we have not described here.

The subset we covered in this lecture is nevertheless
useful in itself.
• And happens to include all the XPath needed to understand
XML Schema.
195
Descargar

Document