value-class-pattern-issues

From Microformats Wiki
Revision as of 17:28, 17 June 2008 by BenWard (talk | contribs) (→‎Depth of Parsing: Added another depth of parsing example)
Jump to navigation Jump to search

Value Excerption Pattern Issues

Open issues concerning the parsing of the value excerption pattern.

Open Issues

These issues are awaiting resolution and reflection in the specification, but may not be blockers on the implementation of the specification.

Excluded Fields

There seem to be some properties within which value excerpting is NOT allowed (or should not be allowed!) e.g. "type" in hCard. TobyInk 07:38, 22 May 2008 (PDT)

  • You mean type as a sub-property of tel? That's one of the identified machine-data items that needs a means of including the publisher's choice text along with the microformat specified one. Not to conflate two separate issues, but just noting that separation of type text and type value needs to be handled somewhere, and value-excerption-pattern could be considered as part of the solution. BenWard 07:54, 22 May 2008 (PDT)
  • Some fields make sense to exclude this, as it seems unintuitive, and can be used to avoid many of nested-microformat problems that may avoid a messier mfo pattern. E.g. entry-summary and entry-content in hAtom, both could very feasibly have nested formats or any kind, but doesn't strike me as useful to segregate into "value" at all. BenWard 10:41, 6 Jun 2008 (PDT)
  • Total other alternative, make value-excerption opt-in. Would need a bit of effort to go through all the specs and clarify, but actually might make more sense. It's a useful pattern for some properties (especially those with data patterns). BenWard 10:41, 6 Jun 2008 (PDT)

White-space behaviour when concatenating value nodes.

We specify that no characters get inserted between concatenated occurrences of ‘value’. Need to audit all properties to ensure that this behaviour would be correct in all cases.

Possibly specify that individual properties can override this behaviour, specifying a separator character. Possibly specify that this should be a provision of parsing implementations, so as to maintain flexibility for future publishing.

Depth of Parsing

Currently any descendent is parsed, which causes issues if a microformat field using the value-excerption-pattern is nested within another.

Example:

<code><div class="hentry vevent">
    <h1 class="entry-title summary">Party on Sunday!</h1>
    <div class="updated published">Tuesday <span class="value">2008-06-17</span></div>
    <p class="entry-content description">We're having a party on <span class="dtstart">Sunday, at 7pm! <span class="value">2008-06-22T19:00:00+0100</span></span>. Please bring your friends!</p>
</div></code>

In this example, hAtom and hCalendar are interleaved. The DTSTART property of the event is contained within the entry-content of the hAtom entry, using the value-excerption-pattern to include the machine-data datetime. However, with full descendent parsing, the hAtom model will come out as the following:

ENTRY
    ENTRY-TITLE=Party on Sunday!
    UPDATED= 2008-06-17
    PUBLISHED=2008-06-17
    ENTRY-CONTENT=2008-06-22T19:00:00+0100
  • e.g. an hCalendar vevent nested inside hAtom entry-content must not result in entry-content parsing as 20080627T12:34:00+100.
  • e.g. hCalendar defines organizer, which may be an hCard, which may have a tel property containing a sub-property value. Under these parsing rules, the entire organizer field would be parsed as the telephone number.
  • Cognition copes with this OK -- the organizer is parsed as a full contact with an hCard - not just a number. TobyInk 07:38, 22 May 2008 (PDT)
    • Ah, but if a parser were written just as an hCalendar parser, with no hCard support? Cognition already understands that hCard is something special. BenWard 07:54, 22 May 2008 (PDT)

Possible resolutions:

  • Specify the mfo (‘microformat object’) class be used when nesting microformats, as a processing instruction to parsers not to parse unrelated nested items
  • Specify that value must only be read from children, not from all descendants. Restrictive, unlikely to work for existing hCard TEL usage.
  • Specify the above (parse children, not all descendants), but allow individual properties (such as tel to override and parse all descendants. This would result in a parse-depth flag on all fields, and many getting overridden for all descendants, but again, seems to be a well structured solution. Property name dictionaries in parsers would have to include the depth flag with the property.

Parsing title from Empty value Elements

As a solution to the invisible data requirements sometimes presented by machine-data in microformats, a parsing rule is proposed where the value element is empty (contains no non-whitespace characters), the title attribute instead be parsed.

e.g. <span class="dtstart">Tuesday the 24th at 6pm <span class="value" title="20080624T180000+1000"></span>lt;/span>

Possible resolutions:

  • This is parsable, should be specced.
  • Suggest restricting to instances where a single value element exists, e.g. Disallow concatenation of multiple embedded values, and disallow embedded values from being appended to visible data. This pattern exists to solve the machine data problem, and restricting it more will discourage it being used for hiding other, useful data.
  • Standard builds of HTMLTidy drop empty elements, which is unfortunate. However, it is trivial to compile tidy with a patch to not drop empty elements which have class attributes (See tidy-microformats.zip). The linked file contains an intel binary, and a diff for patching against the HTML Tidy source.

Closed Issues

These issues are closed, and either dismissed with reason, or the specification has been updated in resolution.

Nested value

Should <span class="value">Foo <span class="value">Bar</span></span> parse as foo bar or bar? Should value elements be allowed to be nested within value elements?

Resolution: Disallowed. Deemed complex to parse, and unnecessary when publishing.

Related Pages