By Priscilla Walmsley, Published Sep 4, 2012 by Prentice Hall. Part of the Charles F. Goldfarb Definitive XML Series series.
Both element and attribute declarations can use simple types to describe their data content. This chapter introduces simple types and explains how to define your own atomic simple types for use in your schemas.
8.1. Simple type varieties
There are three varieties of simple types: atomic types, list types, and union types.
- Atomic types have values that are indivisible, such as 10 or large.
- List types have values that are whitespace-separated lists of atomic values, such as <availableSizes>10 large 2</availableSizes>.
- Union types may have values that are either atomic values or list values. What differentiates them is that the set of valid values, or “value space,” for the type is the union of the value spaces of two or more other simple types. For example, to represent a dress size, you may define a union type that allows a value to be either an integer from 2 through 18, or one of the string values small, medium, or large.
List and union types are covered in Chapter 10.
8.1.1. Design hint: How much should I break down my data values?
Data values should be broken down to the most atomic level possible. This allows them to be processed in a variety of ways for different uses, such as display, mathematical operations, and validation. It is much easier to concatenate two data values back together than it is to split them apart. In addition, more granular data is easier to validate.
It is a fairly common practice to put a data value and its units in the same element, for example <length>3cm</length>. However, the preferred approach is to have a separate data value, preferably an attribute, for the units, for example <length units="cm">3</length>.
Using a single concatenated value is limiting because:
- It is extremely cumbersome to validate. You have to apply a complicated pattern that would need to change every time a unit type is added.
- You cannot perform comparisons, conversions, or mathematical operations on the data without splitting it apart.
- If you want to display the data item differently (for example, as “3 centimeters” or “3 cm” or just “3”, you have to split it apart. This complicates the stylesheets and applications that process instance documents.
It is possible to go too far, though. For example, you may break a date down as follows:
<orderDate>
<year>2001</year>
<month>06</month>
<day>15</day>
</orderDate>
This is probably overkill unless you have a special need to process these items separately.
8.2. Simple type definitions
8.2.1. Named simple types
Simple types can be either named or anonymous. Named simple types are always defined globally (i.e., their parent is always schema 1) and are required to have a name that is unique among the types (both simple and complex) in the schema. The syntax for a named simple type definition is shown in Table 8-1.
The name of a simple type must be an XML non-colonized name, which means that it must start with a letter or underscore, and may only contain letters, digits, underscores, hyphens, and periods. You cannot include a namespace prefix when defining the type; it takes its namespace from the target namespace of the schema document.
All examples of named types in this book have the word “Type” at the end of their names to clearly distinguish them from element and attribute names. However, this is a convention and not a requirement. You can even have a type definition and an element declaration using the same name, but this is not recommended because it can be confusing.
Example 8-1 shows the definition of a named simple type DressSizeType along with an element declaration that references it. Named types can be used in multiple element and attribute declarations.
Example 8-1. Defining and referencing a named simple type
<xs:simpleType name="DressSizeType">
<xs:restriction base="xs:integer">
<xs:minInclusive value="2"/>
<xs:maxInclusive value="18"/>
</xs:restriction>
</xs:simpleType>
<xs:element name="size" type="DressSizeType"/>
8.2.2. Anonymous simple types
Anonymous types, on the other hand, must not have names. They are always defined entirely within an element or attribute declaration, and may only be used once, by that declaration. Defining a type anonymously prevents it from ever being restricted, used in a list or union, redefined, or overridden. The syntax to define an anonymous simple type is shown in Table 8-2.
Example 8-2 shows the definition of an anonymous simple type within an element declaration.
Example 8-2. Defining an anonymous simple type
<xs:element name="size">
<xs:simpleType>
<xs:restriction base="xs:integer">
<xs:minInclusive value="2"/>
<xs:maxInclusive value="18"/>
</xs:restriction>
</xs:simpleType>
</xs:element>
8.2.3. Design hint: Should I use named or anonymous types?
The advantage of named types is that they may be defined once and used many times. For example, you may define a type named ProductCodeType that lists all of the valid product codes in your organization. This type can then be used in many element and attribute declarations in many schemas. This has the advantages of
- Encouraging consistency throughout the organization
- Reducing the possibility of error
- Requiring less time to define new schemas
- Simplifying maintenance, because new product codes need only be added in one place
If a type is named, you can also derive new types from it, which is another way to promote reuse and consistency.
Named types can also make a schema more readable when its type definitions are complicated.
An anonymous type, on the other hand, can be used only in the element or attribute declaration that contains it. It can never be redefined, overridden, have types derived from it, or be used in a list or union type. This can seriously limit its reusability, extensibility, and ability to change over time.
However, there are cases where anonymous types are preferable to named types. If the type is unlikely to ever be reused, the advantages listed above no longer apply. Also, there is such a thing as too much reuse. For example, if an element can contain the values 1 through 10, it does not make sense to define a type named OneToTenType to be reused by other unrelated element declarations with the same value space. If the value space for one of the element declarations using that named type changes but the other element declarations stay the same, it actually makes maintenance more difficult, because a new type would need to be defined at that time.
In addition, anonymous types can be more readable when they are relatively simple. It is sometimes desirable to have the definition of the type right there with the element or attribute declaration.
8.3. Simple type restrictions
Every simple type is a restriction of another simple type, known as its base type. It is not possible to extend a simple type, except by adding attributes which results in a complex type. This is described in Section 13.4.1 on p. 306.
Every new simple type restricts the value space of its base type in some way. Example 8-3 shows a definition of DressSizeType that restricts the built-in type integer.
Example 8-3. Deriving a simple type from a built-in simple type
<xs:simpleType name="DressSizeType">
<xs:restriction base="xs:integer">
<xs:minInclusive value="2"/>
<xs:maxInclusive value="18"/>
<xs:pattern value="\d{1,2}"/>
</xs:restriction>
</xs:simpleType>
Simple types may also restrict user-derived simple types that are defined in the same schema document, or even in a different schema document. For example, you could further restrict DressSizeType by defining another simple type, MediumDressSizeType, as shown in Example 8-4.
A simple type restricts its base type by applying facets to restrict its values. In Example 8-4, the facets minInclusive and maxInclusive are used to restrict the value of MediumDressSizeType to be between 8 and 12 inclusive.
Example 8-4. Deriving a simple type from a user-derived simple type
<xs:simpleType name="MediumDressSizeType">
<xs:restriction base="DressSizeType">
<xs:minInclusive value="8"/>
<xs:maxInclusive value="12"/>
</xs:restriction>
</xs:simpleType>
8.3.1. Defining a restriction
The syntax for a restriction element is shown in Table 8-3. You must specify one base type either by using the base attribute or by defining the simple type anonymously using a simpleType child. The option of using a simpleType child is generally only useful when restricting list types, as described in Section 10.3.3 on p. 190.
Within a restriction element, you can specify any of the facets, in any order. However, the only facets that may appear more than once in the same restriction are pattern, enumeration, and assertion. It is legal to define a restriction that has no facets specified. In this case, the derived type allows the same values as the base type.
8.3.2. Overview of the facets
The available facets are listed in Table 8-4.
The syntax for applying a facet is shown in Table 8-5. All facets (except assertion) must have a value attribute, which has different valid values depending on the facet. Most facets may also have a fixed attribute, as described in Section 8.3.4 on p. 140.
† The assertion facet has a different syntax that is described in Table 14-1.
Certain facets are not applicable to some types. For example, it does not make sense to apply the fractionDigits facet to a character string type. There is a defined set of applicable facets for each of the built-in types.1 If a facet is applicable to a built-in type, it is also applicable to atomic types that are derived from it. For example, since the length facet is applicable to string, if you derive a new type from string, the length facet is also applicable to your new type. Section 8.4 on p. 142 describes each of the facets in detail and lists the built-in types to which the facet can apply.
8.3.3. Inheriting and restricting facets
When a simple type restricts its base type, it inherits all of the facets of its base type, its base type’s base type, and so on back through its ancestors. Example 8-4 showed a simple type MediumDressSizeType whose base type is DressSizeType. DressSizeType has a pattern facet which restricts its value space to one- or two-digit numbers. Since MediumDressSizeType inherits all of the facets from DressSizeType, this same pattern facet applies to MediumDressSizeType also. Example 8-5 shows an equivalent definition of MediumDressSizeType where it restricts integer and has the pattern facet applied.
Example 8-5. Effective definition of MediumDressSizeType
<xs:simpleType name="MediumDressSizeType">
<xs:restriction base="xs:integer">
<xs:minInclusive value="8"/>
<xs:maxInclusive value="12"/>
<xs:pattern value="\d{1,2}"/>
</xs:restriction>
</xs:simpleType>
Sometimes a simple type definition will include facets that are also specified for one of its ancestors. In Example 8-4, MediumDressSizeType includes minInclusive and maxInclusive, which are also applied to its base type, DressSizeType. The minInclusive and maxInclusive facets of MediumDressSizeType (whose values are 8 and 12, respectively) override those of DressSizeType (2 and 18, respectively).
It is a requirement that the facets of a derived type (in this case MediumDressSizeType) be more restrictive than those of the base type. In Example 8-6, we define a new restriction of DressSizeType, called SmallDressSizeType, and set minInclusive to 0. This type definition is illegal, because it attempts to expand the value space by allowing 0, which was not valid for DressSizeType.
Example 8-6. Illegal attempt to extend a simple type
<xs:simpleType name="SmallDressSizeType">
<xs:restriction base="DressSizeType">
<xs:minInclusive value="0"/>
<xs:maxInclusive value="6"/>
</xs:restriction>
</xs:simpleType>
This rule also applies when you are restricting the built-in types. For example, the short type has a maxInclusive value of 32767. It is illegal to define a restriction of short that sets maxInclusive to 32768.
Although enumeration facets can appear multiple times in the same type definition, they are treated in much the same way. If both a derived type and its ancestor have a set of enumeration facets, the values of the derived type must be a subset of the values of the ancestor. An example of this is provided in Section 8.4.4 on p. 145.
Likewise, the pattern facets specified in a derived type must allow a subset of the values allowed by the ancestor types. A schema processor will not necessarily check that the regular expressions represent a subset; instead, it will validate instances against the patterns of both the derived type and all the ancestor types, effectively taking the intersection of the pattern values.
8.3.4. Fixed facets
When you define a simple type, you can fix one or more of the facets. This means that further restrictions of this type cannot change the value of the facet. Any of the facets may be fixed, with the exception of pattern, enumeration, and assertion. Example 8-7 shows our DressSizeType with fixed minInclusive and maxInclusive facets, as indicated by a fixed attribute set to true.
Example 8-7. Fixed facets
<xs:simpleType name="DressSizeType">
<xs:restriction base="xs:integer">
<xs:minInclusive value="2" fixed="true"/>
<xs:maxInclusive value="18" fixed="true"/>
<xs:pattern value="\d{1,2}"/>
</xs:restriction>
</xs:simpleType>
With this definition of DressSizeType, it would have been illegal to define the MediumDressSizeType as shown in Example 8-4 because it attempts to override the minInclusive and maxInclusive facets which are now fixed. Some of the built-in types have fixed facets that cannot be overridden. For example, the built-in type integer has its fractionDigits facet fixed at 0, so it is illegal to derive a type from integer and specify a fractionDigits that is not 0.
8.3.4.1. Design hint: When should I fix a facet?
Fixing facets makes your type less flexible and discourages other schema authors from reusing it. Keep in mind that any types that may be derived from your type must be more restrictive, so you are not at risk that your type will be dramatically changed if its facets are unfixed.
A justification for fixing facets might be that changing that facet value would significantly alter the meaning of the type. For example, suppose you want to define a simple type that represents price. You define a Price type and fix the fractionDigits at 2. This still allows other schema authors to restrict Price to define other types, for example, by limiting it to a certain range of values. However, they cannot modify the fractionDigits of the type, because this would result in a type not representing a price in dollars and cents.
8.4. Facets
8.4.1. Bounds facets
The four bounds facets (minInclusive, maxInclusive, minExclusive, and maxExclusive) restrict a value to a specified range. Our previous examples applied minInclusive and maxInclusive to restrict the value space of DressSizeType. While minInclusive and maxInclusive specify boundary values that are included in the valid range, minExclusive and maxExclusive specify bounds that are excluded from the valid range.
There are several constraints associated with the bounds facets:
- minInclusive and minExclusive cannot both be applied to the same type. Likewise, maxInclusive and maxExclusive cannot both be applied to the same type. You may, however, mix and match, applying, for example, minInclusive and maxExclusive together. You may also apply just one end of the range, such as minInclusive only.
- The value for the lower bound (minInclusive or minExclusive) must be less than or equal to the value for the upper bound (maxInclusive or maxExclusive).
- The facet value must be a valid value for the base type. For example, when restricting integer, it is illegal to specify a maxInclusive value of 18.5, because 18.5 is not a valid integer.
The four bounds facets can be applied only to the date/time and numeric types, and the types derived from them. Special consideration should be given to time zones when applying bounds facets to date/time types. For more information, see Section 11.4.15 on p. 235.
8.4.2. Length facets
The length facet allows you to limit values to a specific length. If it is a string-based type, length is measured in number of characters. This includes the XML DTD types and anyURI. If it is a binary type, length is measured in octets of binary data. If it is a list type, length is measured as the number of items in the list. The facet value for length must be a nonnegative integer.
The minLength and maxLength facets allow you to limit a value’s length to a specific range. Either of both of these facets may be applied. If they are both applied, minLength must be less than or equal to maxLength. If the length facet is applied, neither minLength nor maxLength may be applied. The facet values for minLength and maxLength must be nonnegative integers.
The three length facets (length, minLength, maxLength) can be applied to any string-based types (including the XML DTD types), the binary types, and anyURI. They cannot be applied to the date/time types, numeric types, or boolean.
8.4.2.1. Design hint: What if I want to allow empty values?
Many of the built-in types do not allow empty values. Types other than string, normalizedString, token, hexBinary, base64Binary, and anyURI do not allow empty values unless xsi:nil appears in the element tag.
You may have an integer that you want to be either between 2 and 18, or empty. First, consider whether you want to make the element (or attribute) optional. In this case, if the data is absent, the element will not appear at all. However, sometimes it is desirable for the element to appear, as a placeholder, or perhaps it is unavoidable because of the technology used to generate the instance.
If you do determine that the elements must be able to appear empty, you must define a union type that includes both the integer type and an empty string, as shown in Example 8-8.
Example 8-8. Union allowing an empty value
<xs:simpleType name="DressSizeType">
<xs:union>
<xs:simpleType>
<xs:restriction base="xs:integer">
<xs:minInclusive value="2"/>
<xs:maxInclusive value="18"/>
</xs:restriction>
</xs:simpleType>
<xs:simpleType>
<xs:restriction base="xs:token">
<xs:enumeration value=""/>
</xs:restriction>
</xs:simpleType>
</xs:union>
</xs:simpleType>
8.4.2.2. Design hint: What if I want to restrict the length of an integer?
The length facet only applies to the string-based types, the XML DTD types, the binary types, and anyURI. It does not make sense to try to limit the length of the date/time types because they have fixed lexical representations. But what if you want to restrict the length of an integer value?
You can restrict the lower and upper bounds of an integer by applying bounds facets, as discussed in Section 8.4.1 on p. 142. You can also control the number of significant digits in an integer using the totalDigits facet, as discussed in Section 8.4.3 on p. 145. However, these facets do not consider leading zeros as significant. Therefore, they cannot force an integer to appear in the instance with a specific number of digits. To do this, you need a pattern. For example, the pattern \d{1,2} used in our DressSizeType example forces the size to be one or two digits long, so 012 would be invalid.
Before taking this approach, however, you should reconsider whether it is really an integer or a string. See Section 11.3.3.1 on p. 220 for a discussion of this issue.
8.4.3. totalDigits and fractionDigits
The totalDigits facet allows you to specify the maximum number of digits in a number. The facet value for totalDigits must be a positive integer.
The fractionDigits facet allows you to specify the maximum number of digits in the fractional part of a number. The facet value for fractionDigits must be a nonnegative integer, and it must not exceed the value for totalDigits, if one exists.
The totalDigits facet can be applied to decimal or any of the integer types, as well as types derived from them. The fractionDigits facet may only be applied to decimal, because it is fixed at 0 for all integer types.
8.4.4. Enumeration
The enumeration facet allows you to specify a distinct set of valid values for a type. Unlike most other facets (except pattern and assertion), the enumeration facet can appear multiple times in a single restriction. Each enumerated value must be unique, and must be valid for that type. If it is a string-based or binary type, you may also specify the empty string in an enumeration value, which allows elements or attributes of that type to have empty values.
Example 8-9 shows a simple type SMLXSizeType that allows the values small, medium, large, and extra large.
Example 8-9. Applying the enumeration facet
<xs:simpleType name="SMLXSizeType">
<xs:restriction base="xs:token">
<xs:enumeration value="small"/>
<xs:enumeration value="medium"/>
<xs:enumeration value="large"/>
<xs:enumeration value="extra large"/>
</xs:restriction>
</xs:simpleType>
When restricting types that have enumerations, it is important to note that you must restrict, rather than extend, the set of enumeration values. For example, if you want to restrict the valid values of SMLSizeType to only be small, medium, and large, you could define a simple type as in Example 8-10.
Example 8-10. Restricting an enumeration
<xs:simpleType name="SMLSizeType">
<xs:restriction base="SMLXSizeType">
<xs:enumeration value="small"/>
<xs:enumeration value="medium"/>
<xs:enumeration value="large"/>
</xs:restriction>
</xs:simpleType>
Note that you need to repeat all of the enumeration values that apply to the new type. This example is legal because the values for SMLSizeType (small, medium, and large) are a subset of the values for SMLXSizeType. By contrast, Example 8-11 attempts to add an enumeration facet to allow the value extra small. This type definition is illegal because it attempts to extend rather than restrict the value space of SMLXSizeType.
Example 8-11. Illegal attempt to extend an enumeration
<xs:simpleType name="XSMLXSizeType">
<xs:restriction base="SMLXSizeType">
<xs:enumeration value="extra small"/>
<xs:enumeration value="small"/>
<xs:enumeration value="medium"/>
<xs:enumeration value="large"/>
<xs:enumeration value="extra large"/>
</xs:restriction>
</xs:simpleType>
The only way to add an enumeration value to a type is by defining a union type. Example 8-12 shows a union type that adds the value extra small to the set of valid values. Union types are described in detail in Section 10.2 on p. 183.
Example 8-12. Using a union to extend an enumeration
<xs:simpleType name="XSMLXSizeType">
<xs:union memberTypes="SMLXSizeType">
<xs:simpleType>
<xs:restriction base="xs:token">
<xs:enumeration value="extra small"/>
</xs:restriction>
</xs:simpleType>
</xs:union>
</xs:simpleType>
When enumerating numbers, it is important to remember that the enumeration facet works on the actual value of the number, not its lexical representation as it appears in an XML instance. Example 8-13 shows a simple type NewSmallDressSizeType that is based on integer, and specifies an enumeration of 2, 4, and 6. The two instance elements shown, which contain 2 and 02, are both valid. This is because 02 is equivalent to 2 for integer-based types. However, if the base type of NewSmallDressSizeType had been string, the value 02 would not be valid, because the strings 2 and 02 are not the same. If you wish to constrain the lexical representation of a numeric type, you should apply the pattern facet instead. For more information on type equality in XML Schema, see Section 11.7 on p. 253.
Example 8-13. Enumerating numeric values
Schema:
<xs:simpleType name="NewSmallDressSizeType">
<xs:restriction base="xs:integer">
<xs:enumeration value="2"/>
<xs:enumeration value="4"/>
<xs:enumeration value="6"/>
</xs:restriction>
</xs:simpleType>
Valid instances:
<size>2</size>
<size>02</size>
The enumeration facet can be applied to any type except boolean.
8.4.5. Pattern
The pattern facet allows you to restrict values to a particular pattern, represented by a regular expression. Chapter 9 provides more detail on the rules for the regular expression syntax. Unlike most other facets (except enumeration and assertion), the pattern facet can be specified multiple times in a single restriction. If multiple pattern facets are specified in the same restriction, the instance value must match at least one of the patterns. It is not required to match all of the patterns.
Example 8-14 shows a simple type DressSizeType that includes the pattern \d{1,2}, which restricts the size to one or two digits.
Example 8-14. Applying the pattern facet
<xs:simpleType name="DressSizeType">
<xs:restriction base="xs:integer">
<xs:minInclusive value="2"/>
<xs:maxInclusive value="18"/>
<xs:pattern value="\d{1,2}"/>
</xs:restriction>
</xs:simpleType>
When restricting types that have patterns, it is important to note that you must restrict, rather than extend, the set of valid values that the patterns represent. In Example 8-15, we define a simple type SmallDressSizeType that is derived from DressSizeType, and add an additional pattern facet that restricts the size to one digit.
Example 8-15. Restricting a pattern
<xs:simpleType name="SmallDressSizeType">
<xs:restriction base="DressSizeType">
<xs:minInclusive value="2"/>
<xs:maxInclusive value="6"/>
<xs:pattern value="\d{1}"/>
</xs:restriction>
</xs:simpleType>
It is not technically an error to apply a pattern facet that does not represent a subset of the ancestors’ pattern facets. However, the schema processor tries to match the instance value against the pattern facets of both the type and its ancestors, ensuring that it is in fact a subset. Example 8-16 shows an illegal attempt to define a new size type that allows the size value to be up to three digits long. While the schema is not in error, it will not have the desired effect because the schema processor will check values against both the pattern of LongerDressSizeType and the pattern of DressSizeType. The value 004 would not be considered a valid instance of LongerDressSizeType because it does not conform to the pattern of DressSizeType.
Example 8-16. Illegal attempt to extend a pattern
<xs:simpleType name="LongerDressSizeType">
<xs:restriction base="DressSizeType">
<xs:pattern value="\d{1,3}"/>
</xs:restriction>
</xs:simpleType>
Unlike the enumeration facet, the pattern facet applies to the lexical representation of the value. If the value 02 appears in an instance, the pattern is applied to the digits 02, not 2 or +2 or any other form of the integer.
The pattern facet can be applied to any type.
8.4.6. Assertion
The assertion facet allows you to specify additional constraints on values using XPath 2.0. Example 8-17 is a simple type with an assertion, namely that the value must be divisible by 2. It uses a facet named assertion with a test attribute that contains the XPath expression.
Simple type assertions are a flexible and powerful feature covered in more detail, along with complex type assertions, in Chapter 14.
Example 8-17. Simple type assertion
<xs:simpleType name="EvenDressSizeType">
<xs:restriction base="DressSizeType">
<xs:assertion test="$value mod 2 = 0" />
</xs:restriction>
</xs:simpleType>
8.4.7. Explicit Time Zone
The explicitTimezone facet allows you to control the presence of an explicit time zone on a date/time value. Example 8-18 is a simple type based on time but with an explicit time zone required. The syntax of time zones is described in more detail in Section 11.4.13 on p. 233.
The value attribute of explicitTimezone has three possible values:
- optional, making the time zone optional (the value for most built-in date/time types)
- required, making the time zone required (the value for the dateTimeStamp built-in type)
- prohibited, disallowing the time zone
Example 8-18. Explicit time zone
<xs:simpleType name="SpecificTimeType">
<xs:restriction base="xs:time">
<xs:explicitTimezone value="required"/>
</xs:restriction>
</xs:simpleType>
8.4.8. Whitespace
The whiteSpace facet allows you to specify the whitespace normalization rules which apply to this value. Unlike the other facets, which restrict the value space of the type, the whiteSpace facet is an instruction to the schema processor on to what to do with whitespace. This type of facet is known as a prelexical facet because it results in some processing of the value before the other constraining facets are applied. The valid values for the whiteSpace facet are:
- preserve: All whitespace is preserved; the value is not changed.
- replace: Each occurrence of tab (#x9), line feed (#xA), and carriage return (#xD) is replaced with a single space (#x20).
- collapse: As with replace, each occurrence of tab (#x9), line feed (#xA), and carriage return (#xD) is replaced with a single space (#x20). After the replacement, all consecutive spaces are collapsed into a single space. In addition, leading and trailing spaces are deleted.
Table 8-6 shows examples of how values of a string-based type will be handled depending on its whiteSpace facet.
The whitespace processing, if any, will happen first, before any validation takes place. In Example 8-9, the base type of SMLXSizeType is token, which has a whiteSpace facet of collapse. Example 8-19 shows valid instances of SMLXSizeType. They are valid because the leading and trailing spaces are removed, and the line feed is turned into a space. If the base type of SMLXSizeType had been string, the whitespace would have been left as is, and these values would have been invalid.
Example 8-19. Valid instances of SMLXSizeType
<size> small </size>
<size>extra
large</size>
Although you should understand what the whiteSpace facet represents, it is unlikely that you will ever apply it directly in your schemas. The whiteSpace facet is fixed at collapse for most built-in types. Only the string-based types can be restricted by a whiteSpace facet, but this is not recommended. Instead, select a base type that already has the whiteSpace facet you want. The types string, normalizedString, and token have the whiteSpace values preserve, replace, and collapse, respectively. For example, if you wish to define a string-based type that will have its whitespace collapsed, base your type on token, instead of basing it on string and applying a whiteSpace facet. Section 11.2.1 on p. 205 provides a discussion of these three types.
8.5. Preventing simple type derivation
XML Schema allows you to prevent derivation of other types from your type. By specifying the final attribute with a value of #all in your simple type definition, you prevent derivation of any kind (restriction, extension, list, or union). If you want more granular control, the value of final can be a whitespace-separated list of any of the keywords restriction, extension, list, or union. The extension value refers to the extension of simple types to derive complex types, described in Section 13.4.1 on p. 306. Example 8-20 shows some valid values for final.
Example 8-20. Valid values for the final attribute in simple type definitions
final="#all"
final="restriction list union"
final="list restriction extension"
final="union"
final=""
Example 8-21 shows a simple type that cannot be restricted by any other type or used as the item type of a list. With this definition of DressSizeType, it would have been illegal to define MediumDressSizeType in Example 8-4 because it attempts to restrict DressSizeType.
Example 8-21. Preventing type derivation
<xs:simpleType name="DressSizeType" final="restriction list">
<xs:restriction base="xs:integer">
<xs:minInclusive value="2"/>
<xs:maxInclusive value="18"/>
</xs:restriction>
</xs:simpleType>
If no final attribute is specified, it defaults to the value of the finalDefault attribute of the schema element. If neither final nor finalDefault is specified, there are no restrictions on derivation from that type. You can specify the empty string ("") for the final value if you want to override the finalDefault value.
8.6. Implementation-defined types and facets
Starting with version 1.1, additional simple types and facets may be defined and supported by a particular XML Schema implementation.
8.6.1. Implementation-defined types
An implementation can choose to support a set of primitive simple types in addition to those built into XML Schema (described in Chapter 11).
Suppose that an implementation defines a special primitive type ordinalDate that represents an ordinal date: a year, followed by a hyphen, followed by a number from 001 to 366 indicating the day of the year. Although an ordinal date value could be represented as a string, it may be beneficial to promote it to its own primitive type if it has special considerations for ordering or validation of its values, or special operations that can be performed on it (for example, subtracting two ordinal dates to get a duration).
A schema author can use an implementation-defined type just like a built-in type, except that it will be in a different namespace defined by the implementation. The schema in Example 8-22 contains two references to the ordinalDate type, which is in the hypothetical http://example.org/extensions namespace. The anyOrdinalDate element declaration refers to the type directly by its qualified name. The OrdinalDateIn2011 user-defined simple type is a restriction of ordinalDate using bounds facets to specify a range of allowed values.
Example 8-22. Using an implementation-defined type
<xs:schema xmlns:xs="http://www.w3.org/2001/XMLSchema"
xmlns:ext="http://example.org/extensions">
<xs:element name="anyOrdinalDate" type="ext:ordinalDate"/>
<xs:element name="recentOrdinalDate" type="OrdinalDateIn2011"/>
<xs:simpleType name="OrdinalDateIn2011">
<xs:restriction base="ext:ordinalDate">
<xs:minInclusive value="2011-001"/>
<xs:maxInclusive value="2011-365"/>
</xs:restriction>
</xs:simpleType>
</xs:schema>
8.6.2. Implementation-defined facets
Implementation-defined facets might specify additional constraints on the valid values, or even signal to the processor how to process the value. An example is the Saxon processor’s preprocess facet which allows you to specify an XPath expression that transforms the value in some way before validation.
In Example 8-23, the saxon:preprocess facet appears among the children of restriction. You can tell that it is an implementation-defined facet because it is in a different namespace, http://saxon.sf.net/. This particular example is telling the processor to convert the value to upper case before validating it against the enumeration facets. It is essentially implementing a case-insensitive enumeration.
Example 8-23. Using the Saxon preprocess facet
<xs:schema xmlns:xs="http://www.w3.org/2001/XMLSchema"
xmlns:saxon="http://saxon.sf.net/">
<xs:simpleType name="SMLXSizeType">
<xs:restriction base="xs:token">
<saxon:preprocess action="upper-case($value)"/>
<xs:enumeration value="SMALL"/>
<xs:enumeration value="MEDIUM"/>
<xs:enumeration value="LARGE"/>
<xs:enumeration value="EXTRA LARGE"/>
</xs:restriction>
</xs:simpleType>
</xs:schema>
Implementation-defined facets can apply to the XML Schema built-in types (and user-defined restrictions of them); they can also apply to any implementation-defined types such as the ordinalDate example type described in the previous section.
While implementation-defined types and facets can be useful, they do affect the portability of your schema. With the schema in Example 8-23, if you try to validate a document that contains lower-case “small” for a size, it would be valid when using Saxon but not when using a different implementation. Therefore, implementation-defined facets should be used only in controlled situations. Section 23.5.3 on p. 642 provides more information on how to make your schemas more portable across implementations when using implementation-defined types and facets.
