XSLT: Rendering Full-text and Metadata

1
XSLT Rendering Full-text and Metadata

• Thomas G. Habing
• Grainger Engineering Library Information Center
• University of Illinois at Urbana-Champaign
• http//dli.grainger.uiuc.edu
  thabing_at_uiuc.edu

2
Evolution of Rendering

• Full-text
• 96 Plugins Adobe Acrobat (PDF) or Softquad
  Panorma (SGML)
• 99 Custom Server-side Converter to DHTML
• 00 Standards-based XSLT Converter to DHTML
• Metadata
• 96 OpenText Livelink for short and long
  citations
• 98 Merged with full-text
• 99 XSLT for long citations
• 00 Migrating to XSLT for long short citations

3
XSLT Where Should It Happen

• Client-side
• IE5 only
• Not Netscape 6 or Mozilla (yet)
• IE5 not yet fully compliant w/ XSLT and XPath
  standard
• Can reduce the load on your servers
• But performance on low-end clients can be BAD
• Server-side
• Performance could be a problem on busy servers,
  serving large, complex documents
• More control and flexibility over the conversion
  (metamerge)
• Offline Preconversion
• Best performance
• Not good for dynamic documents (metamerge)

4
Transform Source XML into HTML or Renderable
XML?

• Our current choice is HTML
• Many browsers do not support XML at all
• Even of the browsers that do support XML, they
  still support HTML better
• Images, Linking, Tables, CSS, etc.
• As there is more support for additional XML
  standards, such as XLink, this may not always be
  the case
• Even for renderable XML, we still need to
  revert to embedded HTML for some constructs
• Eliminates the need to for dual code maintenance

5
How We Use XSLT

• Most tags are converted to HTML spans with a
  class attribute equal to the original tag
  name,i.e. ltsectgt becomes ltspan classsectgt
• ltxsltemplate matchgt ltspangt
  ltxslattribute nameclassgt ltxslvalue-of
  selectlocal-name()/gt lt/xslattributegt
  ltxslapply-templates/gt lt/spangtlt/xsltemplategt
• If the tag must be a block element, that is
  specified in the CSS span.sect displayblock
• This improves the modularity of the code

6
How We Use XSLT (cont.)

• Some tags require special conversion,i.e. ltemph
  type1gt becomes ltspan classemph_1gt
• ltxsltemplate matchemphgt ltspangt
  ltxslattribute nameclassgt ltxslvalue-of
  selectlocal-name()/gt_ ltxslvalue-of
  select_at_type/gt lt/xslattributegt
  ltxslapply-templates/gt lt/spangtlt/xsltemplategt
• CSS span.emph_1 font-styleitalic
  span.emph_2 font-weightbold

7
How We Use XSLT (cont.)

• Some elements need to have punctuation added,
  i.e.ltaggtltaugtTomlt/augtltaugtTimlt/augtltaugtBoblt/augtlt/aggt
  becomes Tom, Tim, Bob.
• ltxsltemplate matchauposition() last()gt
  ltspangtltxslcall-template nameclass-attr/gt
  ltxslapply-templates/gt. lt/spangtlt/xsltemplategt
  ltxsltemplate matchauposition() ! last()gt
  ltspangtltxslcall-template nameclass-attr/gt
  ltxslapply-templates/gt,32 lt/spangtlt/xsltempl
  ategt

8
How We Use XSLT (cont.)

• Some elements need their children rearrangedi.e.
  ltaugtltsngtHabinglt/sngtltfngtTomlt/fngtlt/augt becomes Tom
  Habing
• ltxsltemplate matchaugt ltspangtltxslcall-templ
  ate nameclass-attr/gt ltxslapply-templates
  selectfn/gt32 ltxslapply-templates
  selectsn/gt lt/spangtlt/xsltemplategt
• Real DTDs can require some fairly complex
  processing.
• So far XSLT seems to be able to handle nearly
  every case we have come across
• However, some cases have required JScript
  extensions to XSLT

9
How We Use XSLT (cont.)

• Some elements are converted into HTML elements
  other than ltspangt
• Elements representing figures or images are
  converted to the ltimg srcgt tag.
• Elements representing internal links with ID and
  IDREF attributes are usually converted into lta
  namegtlt/agt or lta hrefgtlt/agt tags.
• Table elements are converted into corresponding
  HTML lttablegt, lttrgt, or lttdgt tags.
• In all cases, a class attribute is added which is
  the original XML tag name this is used in the
  CSS.

10
Alternates to XSLT

• Custom Parser (last years workshop)
• SAX - Simple API for XML
• Performance can be better for large XML files
• Streaming parser - entire document is not loaded
  into memory as with XSLT
• Not as flexible as XSLT
• More difficult to rearrange output
• Not a W3C standard