Parsing Microformats

...any way you like

Ryan King Engineer

Scope

Two Kinds of Microformats

compound vs. elemental microformats

Parsing?

• Not the best name
• A better name?

Overlaid Trees

• extracting a logical tree from a physical tree

      <>
        <>
          <>
            <>...</>
          </>
        </>
      </>
    

General Rules

fragment identifiers

• process that part of the document
• a@name is not defined
• see: Handling URLs

root class names

• purposely rare terms to reduce the chances of collision
• can be on any element that has @class (including body)

properties

• read the specs for lists
• must ignore

cardinality

• read the specs
• for singular properties, take the first one you find (document order, aka DFS)
• a few special cases

sub-properties

• example: hcard's n

      <p class="vcard">
        <span class="fn n">
          <span class="given-name">Ryan</span>
          <span class="family-name">King</span>
        </span>
      </p>
    

Extracting Properties

text

• simplest extraction
• concatenate text nodes

      <span class="vcard">
        <span class="fn n">
          <span class="given-name">Ryan</span>
          <span class="family-name">King</span>
        </span>
      </span>
    

This hCard's fn is "Ryan King".

(remember whitespace collapsing rules)

value-excerpting

• if children that match .value
• concatenate values

      <span class="vcard">
        <span class="fn n">
          ...
        </span>
        <span class="title">I'm an
          <span class="value">Engineer</span> 
          at Technorati.
        </span>
      </span>
    

title is just "Engineer"

semantic exceptions

Applying HTML semantics to make publishing smoother.

mailto links

• omit mailto URI scheme and query string

      <span class="vcard">
        <span class="fn n">
          ...
        </span>
        <a class="email" href="mailto:ryan@theryanking.com?Subject=nice+presentation">
          email me!
        </a>
      </span>
    

email is just "ryan@theryanking.com"

URL fields

URL Fields

Examples:

<a     class="url" href="foo">bar</a>

<area  class="url" href="foo">bar</a>

<img   class="url" src="foo" alt="bar" />

<object class="url" data="foo">bar</a>

the value is "foo"

non-URLs

non-URL Fields

Examples:

<img   class="fn" alt="foo" src="bar" />

<area  class="fn" alt="foo" href="bar" />

<abbr  class="fn" title="foo">bar</abbr>

the value is "foo"

table headers

rationale/history

• tabular calendars
• data in headers

table headers

example

      <table>
        <tr>
          <th>Time</th>
          <th id="location"><span class="location">Coleman</span></th>
        </tr>
        <tr>
          <th id="s1115">
            <abbr class="dtstart" title="2007-05-12T11:15-0600">11:15</abbr>
          </th>
          <td headers="location s1115" class="vevent">
          <span class="summary">Parsing Microformats...</span>
          </td>
        </tr>
      </table>
    

table headers

Processing

• virtual transclusion
• headers go at end (allow overriding)

include pattern

rationale

• hResume
• repeated hCard data

include pattern

example

      <span class="vcard">
        <span id="name" class="fn n">...</span>
      </span>

      ....
      
      <span class="vcard">
        <a class="include" href="#name"></a>
        ...
      </span>

    

include pattern

notes

• virtual transclusion
• transcluded in place

Approaches

XPath

XPath

XPath

challenge 1

    <xsl:if test="contains(
      concat (
        ' ',
        concat(
          normalize-whitespace(@class),
          ' '
        )
      ),
      ' vcard '
     )" >
    

Tokenized attributes are not easy.

XPath

challenge 2

• non-xml content
  • libxml2's html mode
  • tidy

XPath

See also

• X2V
• GRDDL

Selectors

Advantages

• easy to read
• fits the content model

Selectors

Challenges

• expressiveness of selectors
• ".fn with only one .vcard ancestor"?

Event-based (like SAX)

Advantages

• can deal with some content that is not-well-formed outside of the microformat
• fast, low memory footprint

Event-based

Disadvantages

• hard to maintain

Test Suite

• hCard -> vCard
• hCalendar -> iCalendar
• hReview -> JSON

Test Suite

Future Work

• more JSON
• better web version (side-by-side comparison, etc.)
• more documentation

Resources

• hcard parsing
• parsing microformats
• test suite
• microformats-dev list

?