Normalization

The parser applies several normalizations before mapping to the typed model. This ensures consistent output regardless of how the DMN file was produced.

CDATA and entities

<![CDATA[> 700]]> and > 700 produce identical text values. Both are normalized to the entity form before parsing.

Defaults

Attribute	Default	Applied to
@hitPolicy	UNIQUE	decisionTable
@preferredOrientation	Rule-as-Row	decisionTable
@isCollection	false	itemDefinition, itemComponent, inputData
@textFormat	text/plain	textAnnotation
@associationDirection	None	association
@kind	FEEL	functionDefinition, encapsulatedLogic

Wildcard inputEntry

An empty <inputEntry> and one containing <text>-</text> are both normalized to text: '-'.

Field promotion

Some tools write child text elements instead of attributes:

<!-- Both forms produce the same model -->
<variable name="score" typeRef="number"/>
<variable name="score"><typeRef>number</typeRef></variable>

The parser promotes typeRef, description, and locationURI child elements to their attribute form.

UTF-8 BOM

A UTF-8 BOM () at the start of the file is silently stripped before parsing.

DMNDI

DMNDI diagram elements are excluded by default. Pass { includeDMNDI: true } to include them in model.dmndi.