Pre-Processing through Flattening

Complex modeling constructs that are not supported by the ArcGIS 10.x geodatabase model can be removed using the Flattener transformation. The sequence of applicable transformation procedures is described in the following sections. A configuration example for the Flattener is also provided (see last subsection).

Remove type

This optional flattening step allows removing application schema types that would lead to unnecessarily complex classes in the resulting ArcGIS workspace model. For example, if the application schema contained a construct that supported metadata for properties, the according metadata type can be removed using the type removal rule.

Flatten ONINAs

This flattening step removes XxxReason <<union>> types from the model and ensures that nil reason codes (other, noInformation, notApplicable) are incorporated in Xxx enumerations. Execution of this flattening rule is required for application schema that use the XxxReason <<union>> modeling construct, and optional otherwise .

Flatten inheritance

In this optional flattening step, ShapeChange can flatten particular inheritance trees contained in the model. The ArcGIS workspace (UML) model supports inheritance to a certain extent, which will be explained later on. In specific cases, for example to prepare the execution of the homogeneous geometry flattening rule, inheritance flattening should be performed.

If ArcGIS subtypes are explicitly modelled, inheritance flattening should NOT include these types. In order to prevent their inheritance relationships from being flattened, include rule-trf-cls-flatten-inheritance-ignore-arcgis-subtypes in the rule set of the flattener transformation (that flattens inheritance).

Flatten multiplicity

A database schema supports unlimited cardinality for a property only in the case of 1:n and n:m relationships between object or feature types. Accordingly, the multiplicity of properties where the value type is an object or feature type does not need to be flattened. In all other cases, this flattening step ensures that the maximum cardinality is restricted to a limited size (three, for example).

For creating an ArcGIS workspace model, it is also important that the maximum cardinality of geometry typed properties is not greater than one.

This flattening step is required if the application schema contains properties with unlimited maximum cardinality and a value type that is not a feature type (or object type, if the object type is not flattened in the following flattening step).

Flatten types

This flattening step is used to move the contents of <<dataType>> and <<union>> types into feature and object types. The content of particular object types can also be flattened, if desired (for example in the case of a complex geometry type model). Type flattening can also apply a type mapping, for example to replace the type “Measure” (from ISO 19103) with type “Real”, or “CI_Citation” with “CharacterString”. This further reduces the complexity of the model.The flattening step is required in case that the application schema contains properties with a value type of <<dataType>> or <<union>> (or object types, in case they are not omitted from type flattening) that are not mapped to a simple type.

Particularly in the case of flattening the “Measure” type, this mapping can lead to a loss of expressiveness – and should thus be well documented and the implications understood by the user of the ArcGIS workspace model. In this example, the community may have defined a recommended unit of measure for particular properties, which could then considered to be required for property values stored in an Esri geodatabase.

Flatten name

In this optional flattening step, the name of a model element (class or property, i.e. attribute or association end role) and its code (either stored in the alias or a tagged value of the model element) will be switched – if the code exists. This can be omitted for enumeration values, which is useful for the creation of the ArcGIS workspace model in case that the code represents the initial value for the enumeration. This will be explained in more detail later on.

Remove name and code component

This optional flattening step can remove particular components of model element names. The previous flattening steps, in particular type flattening, can lead to rather complex names (through concatenation of or addition to model element names). This flattening step can reduce that complexity by removing unnecessary name components.

Flatten homogeneous geometries

If an application schema contains features with properties whose value type is a choice between different geometry types (either modelled as a <<union>> or through inheritance) then this choice must be resolved. The reason is that the ArcGIS workspace model does not support the conversion of feature types that can have geometry of different types.

Additional Considerations

Constraints

Constraints are not flattened so that information to create range domains and identify the desired length for textual attributes is available in the model.

Configuration Example

The following code shows how the Flattener transformation can be configured to pre-process the application schema so that it can be converted to an ArcGIS workspace model. Note that the process map entries defined for the homogeneous geometry rule can be extended to take into account further geometry types that may occur in a given application schema.

<Transformer class="de.interactive_instruments.ShapeChange.Transformation.Flattening.Flattener"
  id="flattenedModel" input="input" mode="enabled">
  <parameters>
    <ProcessParameter name="removeType" value="TypeMeta,FeatureAttMetadata,GeometryMetadata"/>
    <ProcessParameter name="maxOccurs" value="3"/>
    <ProcessParameter name="maxOccursForSpecificProperties"
      value="FeatureEntity::propGeometry::1,SourceInfo::prop9::2"/>
    <ProcessParameter name="ignoreFeatureOrObjectTypedProperties" value="true"/>
    <ProcessParameter name="flattenObjectTypes" value="false"/>
    <ProcessParameter name="flattenObjectTypesIncludeRegex" value=".*GeometryInfo"/>
    <ProcessParameter name="lowerCaseCodeForProperties" value="true"/>
    <ProcessParameter name="removePropertyNameAndCodeComponent" value="(?i)\.valueOrReason"/>
    <ProcessParameter name="flattenInheritanceIncludeRegex" value=".*GeometryInfo"/>
    <ProcessParameter name="omitHomogeneousGeometriesForTypesWithSingleGeometryProperty"
      value="true"/>
    <ProcessParameter name="applyHomogeneousGeometriesOnSubtypes" value="true"/>
    <ProcessParameter name="codeForEnumerationValues" value="false"/>
  </parameters>
  <rules>
    <ProcessRuleSet name="flattener">
      <rule name="rule-trf-all-removeType"/>
      <rule name="rule-trf-prop-flatten-ONINAs"/>
      <rule name="rule-trf-cls-flatten-inheritance"/>
      <rule name="rule-trf-prop-flatten-multiplicity"/>
      <rule name="rule-trf-prop-flatten-types"/>
      <rule name="rule-trf-all-flatten-name"/>
      <rule name="rule-trf-prop-remove-name-and-code-component"/>
      <rule name="rule-trf-prop-flatten-homogeneousgeometries"/>
    </ProcessRuleSet>
  </rules>
  <mapEntries>
    <ProcessMapEntry rule="rule-trf-prop-flatten-types" targetType="CharacterString"
      type="CI_Citation"/>
    <ProcessMapEntry rule="rule-trf-prop-flatten-types" targetType="Real" type="Measure"/>
    <ProcessMapEntry param="_P" rule="rule-trf-prop-flatten-homogeneousgeometries"
      targetType="GM_Point" type="GM_Point"/>
    <ProcessMapEntry param="_MP" rule="rule-trf-prop-flatten-homogeneousgeometries"
      targetType="GM_MultiPoint" type="GM_MultiPoint"/>
    <ProcessMapEntry param="_C" rule="rule-trf-prop-flatten-homogeneousgeometries"
      targetType="GM_Curve" type="GM_Curve"/>
    <ProcessMapEntry param="_S" rule="rule-trf-prop-flatten-homogeneousgeometries"
      targetType="GM_Surface" type="GM_Surface"/>
  </mapEntries>
</Transformer>