Creating hierarchical dimension for products

Question

I have to customize the hierarchical dimension id generation using property accessor for products. OOB category-dim-output-config.xml includes the definition to generate the hierarchical dimension ids till category, using custom property accessor I'm able to create the dimension ids for categories.
My Endeca application needs the hierarchical dimension ids till product hierarchy, so I have added the following <item /> definition for child product.
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE item PUBLIC "-//Art Technology Group, Inc.//DTD Repository Ouput Specifier 1.0//EN" "http://www.atg.com/dtds/search/indexing-dependency-schema.dtd">
<item item-descriptor-name="category" xml-combine="replace">
<title property-name="displayName"/>
<meta-properties>
<property name="dimval.spec" type="string" property-accessor="/atg/commerce/endeca/index/accessor/MyCustomDimensionIdGenerator" output-name="dimval.spec" is-non-repository-property="true"/>
<property name="$url" suppress="true"/>
<property name="$baseUrl" suppress="true"/>
<property name="$siteId" output-name="dimval.prop.category.siteId" filter="/atg/commerce/endeca/index/ConcatFilter"/>
<property name="rootCatalogId" property-accessor="/atg/commerce/endeca/index/accessor/RootCatalogPropertyAccessor" output-name="dimval.prop.category.rootCatalogId" is-non-repository-property="true"/>
<property name="ancestorCatalogIds" property-accessor="/atg/commerce/endeca/index/accessor/AncestorCatalogsPropertyAccessor" output-name="dimval.prop.category.ancestorCatalogIds" is-non-repository-property="true"/>
<property name="$repository.repositoryName" suppress="true"/>
<property name="$itemDescriptor.itemDescriptorName" suppress="true"/>
<property name="dimval.dimension_spec" default-value="product.parentNode.id" type="string" property-accessor="/atg/commerce/endeca/index/accessor/DimensionSpecPropertyAccessor" output-name="dimval.dimension_spec" is-non-repository-property="true"/>
<property name="dimval.parent_spec" type="string" output-name="dimval.parent_spec" property-accessor="/atg/commerce/endeca/index/accessor/ParentSpecPropertyAccessor" is-non-repository-property="true"/>
<property name="dimval.display_order" type="integer" output-name="dimval.display_order" property-accessor="/atg/commerce/endeca/index/accessor/DisplayOrderPropertyAccessor" is-non-repository-property="true"/>
<property name="$repositoryId" output-name="dimval.prop.category.repositoryId"/> <!-- override $sideId output name, since it's irritatingly defined as
having an explicit output name. -->
</meta-properties>
<text-properties>
<property name="displayName" output-name="dimval.display_name"/>
<!-- <property name="description"/> <property name="longDescription"/> -->
<property name="keywords" suppress="true" filter="concat"/>
</text-properties>
<item property-name="parentCategoriesForCatalog">
<meta-properties>
<property name="$repositoryId" suppress="true"/>
<property name="$repository.repositoryName" suppress="true"/>
<property name="$itemDescriptor.itemDescriptorName" type="string" suppress="true" filter="unique"/>
</meta-properties>
<monitor>
<!-- monitor repository id, since it goes into CategoryParentPropertyAccessor -->
<property name="$repositoryId"/>
</monitor>
</item>
<item property-name="parentCatalog">
<meta-properties>
<property name="$repositoryId" suppress="true"/>
<property name="$repository.repositoryName" suppress="true"/>
<property name="$itemDescriptor.itemDescriptorName" suppress="true"/>
</meta-properties>
<monitor>
<!-- monitor repository id, since it goes into CategoryParentPropertyAccessor -->
<property name="$repositoryId"/>
</monitor>
</item>
<item is-multi="true" property-name="catalogs">
<meta-properties>
<property name="$repositoryId" output-name="dimval.prop.category.catalogs.repositoryId" filter="/atg/commerce/endeca/index/ConcatFilter"/>
<property name="$repository.repositoryName" suppress="true"/>
<property name="$itemDescriptor.itemDescriptorName" suppress="true"/>
</meta-properties>
</item>
<item is-multi="true" property-name="childProducts" is-document="true">
<meta-properties>
<property name="$repositoryId" output-name="dimval.prop.category.childProduct.repositoryId"/>
<property name="$repository.repositoryName" suppress="true"/>
<property name="$itemDescriptor.itemDescriptorName" suppress="true"/>
</meta-properties>
<monitor>
<!-- monitor repository id, since it goes into CategoryParentPropertyAccessor -->
<property name="$repositoryId"/>
</monitor>
</item>
</item>
This is generating the dimension id for categories only but not for products as I have added "dimval_spec" at category level. I i add the "dimval_spec" property in product item then product record includes two "dimval_spec" property values. Actually a record should have only one "dimval_spec" property and its value.
My queries are "
1) does the definition file I have defined is correct?
1) How can I have only one dimension id property in the definition xml for both category and product, so that each record will contain only one dimension id?
Appreciate the help if any one has worked on this and provide some guidance ... !

Answer 1

Hi,
There is as no thmb rule for it. Generally we try to balance the dimension when we create them. For example, It is not advisable to include Material and Vendor in a single dimension.(otherwise it will lead to dimension explosion...large no. of records in a single dimension.)
The idea of dimension tables is to group associated characteristics to reduce the number of fields in the fact table and to reduce redundancies in storing frequent combinations of
characteristics values in a table. Examples of characteristics typically grouped into one dimension include Customer, Customer Group, and Zip Code grouped into a Customer
dimension, and Material, Material Group, Color grouped into a Material dimension.
A special case of a dimension is the line-item dimension, which is used for characteristics with a very high cardinality. Atypical example is the document number in a lineitem-
level InfoCube.
Regards,
Gunjan.