It is highly recommended to split the output into many RDF files with clear separation criteria. It makes it easier to see the effect of each converter change, when used together with version control for output RDF files.

AnnoCultor supports two dimensions of output files splitting: horizontal and vertical.

Split horizontally

Fits the case when we want a specific property to be written to a specific file, e.g. to store all preferred labels in one file, and all alternative labels in another file. Technically, it can be done by creating two destinations and using them in the corresponding property rules:

<ac:RenameLiteralPropertyRule-default>
   <ac:srcPath rdf:datatype="ac:Path">prefLabel</ac:srcPath>
   <ac:dstProperty rdf:datatype="ac:Property">skos:prefLabel</ac:dstProperty>
   <ac:dstGraph rdf:datatype="ac:Graph">PreferredLabels</ac:dstGraph>
</ac:RenameLiteralPropertyRule-default>
<ac:RenameLiteralPropertyRule-default>
   <ac:srcPath rdf:datatype="ac:Path">altLabel</ac:srcPath>
   <ac:dstProperty rdf:datatype="ac:Property">skos:altLabel</ac:dstProperty>
   <ac:dstGraph rdf:datatype="ac:Graph">AlternativeLabels</ac:dstGraph>
</ac:RenameLiteralPropertyRule-default>

Thus, skos:prefLabel of all records will go to the PreferredLabels RDF file, and skos:altLabel will go to the AlternativeLabels file.

Split vertically

Fits the case when we want records to be distributed among different files, keeping their properties intact. For example, let us separate all records by gender.

Technically, it can be done by creating a proxy destination, use it in property rules, and having a piece of Java code that assigns the correct destination per-record. 

<ac:destinations rdf:parseType="Collection">
  <ac:RdfGraph rdf:about="proxy">
    <ac:comment>Proxy destination, it will be always replaced by male or female, and never written to</ac:comment>
  </ac:RdfGraph>
  <ac:RdfGraph rdf:about="male">
        <ac:comment>Males</ac:comment>
  </ac:RdfGraph>
  <ac:RdfGraph rdf:about="female">
        <ac:comment>Females</ac:comment>
  </ac:RdfGraph>
</ac:destinations>

<ac:ObjectRule rdf:about="objectRule">
   <ac:recordSeparator></ac:recordSeparator>
   <ac:recordIdentifier>record_id</ac:recordIdentifier>
   <ac:recordInformalIdentifier>name</ac:recordInformalIdentifier>
   <ac:listeners rdf:parseType="Collection">
      <ac:OnPreCondition>
         <ac:java><![CDATA[
      
         // assume there is a field called 'gender'
         // with values 'male' or 'female'
         // that accidentally coincides with the corresponding graph names 
         Value value = sourceDataObject.getFirstValue(factory.makePath("gender"));
         if (value == null) {
           throw new Exception("No gender specified in record " + sourceDataObject);
         }
         String graphName = value.getValue();
         Graph destination = factory.makeGraph(graphName);
         proxy.setRealGraph(destination);
         ]]></ac:java>
      </ac:OnPreCondition>
   </ac:listeners>
   <ac:propertyRules>
      <ac:RenameLiteralPropertyRule-default>
         <ac:srcPath rdf:datatype="ac:Path">name</ac:srcPath>
         <ac:dstProperty rdf:datatype="ac:Property">skos:prefLabel</ac:dstProperty>
         <ac:dstGraph rdf:datatype="ac:Graph">proxy</ac:dstGraph>
      </ac:RenameLiteralPropertyRule-default>
   </ac:propertyRules>
</ac:ObjectRule>