Transforming XML The XSLT Language

1
Transforming XMLThe XSLT Language

• Michael H. Kay

2
Topics
XSLTas a language
Usability and fitness for purpose
Academicinterest
Open Sourceand Investment
Successin the market
Successin the market
Successin the market
TechnologyArchitecture
SAXON as a product
SAXON as a product
SoftwareEngineering
3
Why did XML happen?

• The Web needed something better than HTML
• Data Interchange needed something better than CSV
  files
• SGML was around, subsetting it was easier than
  reinventing
• It had to be cheap, and it was
• By luck, there was no competition

4
XSLTExtensible Stylesheet Language -
Transformations

• A declarative language for transforming XML
• Widely used in
• publishing applications
• messaging applications
• anywhere else where XML is found
• XSLT 1.0 (1999) was widely implemented
• XSLT 2.0 (2007) is popular with users, but there
  are few products

5
What kind of language is XSLT?

• Declarative, functional
• Rule-based
• Uses XML syntax
• Data model is abstract XML tree
• Type system based on XML Schema
• A two-language system XSLTXPath

6
Template Rules
ltxsltemplate matchbibliographygt lth1gtlta
namebiblgtBibliographylt/agtlt/h1gt ltdlgt
ltxslapply-templates/gt lt/dlgt lt/xsltemplategt ltx
sltemplate matchbibl-entrygt ltdtgt
ltxslvalue-of select_at_ref/gt lt/dtgt ltddgt
ltxslapply-templates/gt lt/ddgt lt/xsltemplategt
7
The XSLT Processing Model
8
Pros and ConsDeclarative, Functional

• PRO
• Optimizable
• Safe
• Robust
• Productive

• CON
• Script kiddies hate it
• Slow
• Recursion is mind-numbing

9
Pros and ConsRule-based

• PRO
• Great for text and semi-structured data
• Potential for change

• CON
• Script kiddies hate it
• Makes static analysis hard

10
Pros and ConsUses XML Syntax

• PRO
• Templates fill in the blanks
• Common infrastructure (editors, parsers)
• Extensibility

• CON
• Verbose
• Ugly

11
Pros and ConsXML Tree Data Model

• PRO
• Abstracts away from the lexical XML detail
• But its still distinctively XML

• CON
• Too abstract for some
• In-memory assumption
• Inadequate for complex algorithms

12
The XPath Axes (1)
parent
preceding-sibling
following-sibling
child
self
13
The XPath Axes (2)
ancestor
preceding
following
descendant
14
Pros and ConsUse of XML Schema

• PRO
• Everyone uses XML Schema
• Type safety
• Optimization
• Better diagnostics

• CON
• Everyone hates XML Schema
• Strong typing is for wimps

15
Success Factors for XSLT

• People needed it
• There wasnt much competition
• Good open-source implementations appeared early
• High level of spec conformance
• Adequate performance
• Browser support
• Endorsement/credibility

16
Architectureof an XSLT Processor
Stylesheet
Parser
Builder
Compiler
CompiledStylesheet
ResultDocument
SourceDocument
Outputter
XPath
Serializer
Builder
Parser
17
Factors driving Performance

• Tree model searching and matching
• Pipelining, streaming
• Static code optimization
• Use of schema
• Code generation
• Basic good programming
• Engineering Methodology

18
XSLT and XQuery

• XQuery 1.0 overlaps in capability
• Much smaller language, less power
• Clean design, easier to learn
• Backed by database vendors
• who have money
• Backed by academics
• who want money
• More oriented to data than documents
• which is where the money is

19
New in XSLT 2.0

• Grouping
• Regular Expressions
• Schema-awareness
• Functions
• Multiple output
• Date/time handling
• ... and much more

20
Where are we today?

• XSLT 2.0 came out in Jan 2007
• 2½ implementations (Saxon, Altova)
• Highly popular with users
• but not with non-users
• The big vendors have failed to produce products
• not for want of trying

21
Saxon (XSLT and XQuery)
10 years
300 emails/month
500downloadsper day
xxxK revenue
1 developer
180K LOC
300K test cases
10 bugs/month
22
Engineering Model

• Alpine mountaineering
• Agile, high-speed, high-risk
• Small teams, fast decision making
• Importance of tooling
• IDEs
• test automation
• Ship frequently
• low half-life for bugs
• Support community

23
Business Model

• Open Source has driven out the profit
• Only the low-cost operators are able to make
  money
• IBM, Oracle, Microsoft, Intel are failing to
  deliver product
• Is this good for users?

24
Summary XSLT

• XML concentrates on information interchange
• This creates a need for transformation
• Many processors available
• Excellent conformance to standards
• Good performance and reliability
• Often open source and/or free
• Role of XSLT
• Styling XML for presentation
• Transforming XML for application interworking
• Middle-tier business logic

25
Wider lessons

• Open source changes everything
• Ultra-low-cost vendors have a significant
  advantage
• But this is reducing investment and reducing
  quality (on some measures)
• The future is uncertain...