RELAX

RELAX (REgular LAnguage description for XML)

2000 Feb 24

MURATA Makoto

1. Political

RELAX is expected to become a Japanese national standard, and it is already recieving an ovation.
Already welcomed by some influential people in the XML community
Talk at XTech, 2000 March
JIS technical report for RELAX Core, 2000 March
JIS technical report for RELAX Namespace, 2000 June
Submission to the fast track procedure of ISO, late 2000 or early 2001

More about this, see RELAX FAQ

2. Technical

RELAX Core = Datatypes + Horn clauses + Regular Hedge Grammars

2.1 Datatypes

Borrowed from XML Schema Part 2

Built-in datatypes as well as facets.

2.2 Horn Clauses

A clause describes conditions on tag names and attribute values.

(1) Simple example

foo(x) :- 
  getTag(x)="foo",
  getAttr(x, "bar") ∈ Domain´(integer)
  common(x).

common(x) :- 
  getAttr(x, "lang")  ∈ Domain´(lang).

<tag name="foo">
  <attribute name="bar" type="integer"/>
  <ref role="common"/>
</tag>

<attList role="common">
  <attribute name="lang" type="language"/>
</tag>

(2) Condition on attribute values

divAsSection(x) :- 
  getTag(x)="div",
  getAttr(x, "class")="section".

divAsSubsection(x) :- 
  getTag(x)="div",
  getAttr(x, "class")="subsection".

<tag name="div" role="divAsSection">
  <attribute name="class">
    <enumeration value="section"/>
  </attribute>
</tag>

<tag name="div" role="divAsSubsection">
  <attribute name="class">
    <enumeration value="subsection"/>
  </attribute>
</tag>

(3) Mutually exclusive attributes

aWithName(x) :- 
  getTag(x)="a",
  getAttr(x, "name")= Domain(NMTOKEN)
  getAttr(x, "href")= φ.

aWithHref(x) :- 
  getTag(x)="a",
  getAttr(x, "href")= Domain(uri_reference)
  getAttr(x, "name")= φ.

<tag name="a" role="aWithHref">
  <attribute
    name="href"
    type="uri_reference"
    required="true"/>
  <attribute name="name" type="none"/>
</tag>

<tag name="a" role="aWithName">
  <attribute
    name="name"
    type="NMTOKEN"
    required="true"/>
  <attribute name="href" type="none"/>
</tag>

2.3 Regular Hedge Grammar

A hedge is an ordered sequence of ordered trees.

(1) Regular (String) Grammars, revisited

n ::= ε
n ::= m

n ::= a
n ::= a m

(2) Regular Hedge Grammars

n ::= r

n ::= a [r]

where r is a regular expression containing non-terminals.

(3) hedgeRule in RELAX Core

<hedgeRule
   label="non-terminal-name"
  ... hedge model ...
<elementRule>

LHS: a non-terminal.

RHS: a regular expression containing non-terminals.

(4) elementRule in RELAX Core

<elementRule
   role="predicate-name"
   label="non-terminal-name">
  ... hedge model ...
<elementRule>

LHS: a non-terminal.

RHS: a predicate name followed by one of the fowlloing:

a regular expression containing non-terminals,
a mixed hedge model (a mixed content model), or
a datatype reference.

Recall predicates divAsSection and divAsSubsection. Permissible structures can be described as below:

<elementRule
   role="divAsSection"
   label="section">
  <sequence>
    <ref label="paragraph" occurs="*"/>
    <ref label="subsection" occurs="+"/>
  </sequence>
<elementRule>

<elementRule
   role="divAsSubsection"
   label="subsection">
  ... hedge model ...
<elementRule>

3. Future work

Conversion from DTD to RELAX (a beta version is already available)
Release of a RELAX Core processor written in C++ (an alpha version is already available)
Generation of Java classes (an alpha version is already available)
The design of RELAX namespace (already three year history as INSTAC).