PDF Object Types

Objects in PDF have a type and attributes of PDF objects expect to hold objects of a specific type. The types form a hierarchy with Object as root.

The PDF type hierarchy is independent of the Smalltalk class hierarchy. Often, the Smalltalk class hierarchy coincides with the PDF type hierarchy, but this is not required. Also, the name of the PDF type can be different from the name of the Smalltalk class, but often they are the same. PDF types and Smalltalk classes are connected by a 1 to 1 relationship: a class can implement only one type and a type can only be implemented by one class.

A Smalltalk class is declared to represent a PDF type by implementing the class method pdfTypeDefinition which returns a PDFTypeDefinition object like:

pdfTypeDefinition
  ^PDFTypeDefinition
    name: #Integer
    supertype: #Number
    section: '7.3.3'
    documentation: 'An integer shall be written as one or more decimal digits optionally preceded by a sign.
The value shall be interpreted as a signed decimal integer and shall be converted to an integer object.'
    version: 0

• name is a symbol used as type for the type declarations of an attribute. The name should be identical or very similar to the name used in the PDF specification.
• supertype denotes the parent type. It is nil for the top PDF type Object.
• section specifies the section of the PDF specification where the type is defined. This will allow the PDF specification to be opened on the documentation of the type.
• documentation holds a short description of the type, usually copied (and slightly edited) from the first paragraph in the PDF specification.
• version is optional and holds an integer specifying the minor PDF version (mayor version is 1) where the type was introduced. When omitted the version defaults to 0 (corresponding to PDF version 1.0).

pdfTypeDefinition is a special method used to construct the PDF type hierarchy. ALL pdfTypeDefinition methods shall only construct and return a PDFTypeDefinition object. In order to do so, ALL implementers of pdfTypeDefinition are collected, evaluated and sorted into the type hierarchy. Therefore, none of the pdfTypeDefinition methods shall have any side effects. This allows to define a PDF object type independent of the Smalltalk class hierarchy.

Attributes of Dictionaries are typed. The definition of an Attribute specifies the class/type of the object and wheather the object is directly stored or a reference to it.

The following pragmas are used for declaring a type:

• #type: *aSymbol* The value (or referenced value) should be of type aSymbol
• #typeDirect: *aSymbol* The value should be of type aSymbol
• #typeIndirect: *aSymbol* The value should be a reference to an object of type aSymbol
• #typeArrayOf: *aSymbol* The value should be an array with values (or referenced values) of type aSymbol
• #typeDictionaryOf: *aSymbol* The value should be a dictionary with values (or referenced values) of type aSymbol

The pragmas can occur several times. One of the given types should apply.

Page>>Contents
  <type: #Contents>
  <typeArrayOf: #Contents>

This declares that the attribute #Contents of a Page can hold either

• a Contents object or
• an array of Contents objects.

A type of an object is used when it gets assigned to an attribute. With the type requirement of the attribute, the object is specialized to the attribute type if possible. If this is not possible, a TypeError is raised.

The attribute /Root of /Trailer requires a reference to a /Catalog:

Trailer>>Root
  <typeIndirect: #Catalog>

When you assign a reference to a plain /Dictionary to it

aTrailer at: #Root put: Dictionary new newReference

the Dictionary is changed to be a /Catalog:

aTrailer Root class name = #Catalog

If something else gets assigned, a TypeError is raised.

aTrailer at: #Root put: Trailer new newReference

⇒ TypeError

As /Dictionary can be subclassed, so can /Array. The attributes are not named, but accessed positionally with their index. The attributes in subclasses of /Array should be typed and documented like dictionary attributes. For this, the same pragmas as in /Dictionary are used. In /Dictionary the pragma's #attribute:documentation:, the attribute number was only used for sorting. In /Array it is the index of the attribute.

Example:

/Rectangle (subclass of /Array)
 
Rectangle>>llx
  <attribute: 1 documentation: 'lower left corner x'>
  ^self at: 1
 
Rectangle>>lly
  <attribute: 2 documentation: 'lower left corner y'>
  ^self at: 2
 
Rectangle>>urx
  <attribute: 3 documentation: 'upper right corner x'>
  ^self at: 3
 
Rectangle>>ury
  <attribute: 4 documentation: 'upper right corner y'>
  ^self at: 4

Every /Array subclass must implement #numberOfAttributes which is the size of the array. This allows further type checking and the creation of #empty instances with the right number of slots.

From the pragmas above types (instances of Type) are created when sending #typesOf: to a Dictionary with the attribute name as argument. Example:

Trailer new typesOf: #Root

returns

Array with: (IndirectType on: Catalog)

Typing is extended to be able to group several classes in different implementation hierarchies into an abstract PDF Type.

/ColourSpace shall be the super type of all the different /Array and /Name subclasses.

/Function shall be the super type of /FunctionDictionary and /FunctionStream.

/Shading shall be the super type of /ShadingDictionary and /ShadingStream.

PDFObject isType: aType

and

PDFObject class subsumes: aPDFObject

First, we need to know if an object is already of a specific type. If this is true, we leave it alone, since it is already what we wanted.

A PDF object class implements #subsumes: which is simply the test if the object #isKindOf: the class. The supertypes implement this with a list of top classes to which they delegate.

aPDFObject asType: aType

and

PDFObject class specializes: aPDFObject

When an object is not a subclass of a type, it can be specialized if the objects class subsumes the type.