XPath and XQuery Functions and Operators 4.0

F Checklist of implementation-defined features (Non-Normative)

1. It is ·implementation-defined· which version of Unicode is supported, but it is recommended that the most recent version of Unicode be used. (See Conformance.)

2. It is ·implementation-defined· whether the type system is based on XML Schema 1.0 or XML Schema 1.1. (See Conformance.)

3. It is ·implementation-defined· whether definitions that rely on XML (for example, the set of valid XML characters) should use the definitions in XML 1.0 or XML 1.1. (See Conformance.)

4. Implementations may attach an ·implementation-defined· meaning to options in the map that are not described in this specification. These options should use values of type xs:QName as the option names, using an appropriate namespace. (See Options.)

5. It is ·implementation-defined· which version of [The Unicode Standard] is supported, but it is recommended that the most recent version of Unicode be used. (See Strings, characters, and codepoints.)

6. [Definition] Some functions (such as fn:distinct-values, fn:unordered, map:keys, and map:for-each) produce results in an ·implementation-defined· or ·implementation-dependent· order. In such cases two calls with the same arguments are not guaranteed to produce the results in the same order. These functions are said to be nondeterministic with respect to ordering. (See .)[Definition] Some functions (such as fn:distinct-values, fn:unordered, map:keys, and map:for-each) produce results in an ·implementation-defined· or ·implementation-dependent· order. In such cases two calls with the same arguments are not guaranteed to produce the results in the same order. These functions are said to be nondeterministic with respect to ordering. (See Properties of functions.)[Definition] Some functions (such as fn:distinct-values, fn:unordered, map:keys, and map:for-each) produce results in an ·implementation-defined· or ·implementation-dependent· order. In such cases two calls with the same arguments are not guaranteed to produce the results in the same order. These functions are said to be nondeterministic with respect to ordering. (See Properties of functions.)

7. Where the results of a function are described as being (to a greater or lesser extent) ·implementation-defined· or ·implementation-dependent·, this does not by itself remove the requirement that the results should be deterministic: that is, that repeated calls with the same explicit and implicit arguments must return identical results. (See Properties of functions.)

8. In addition, the values of $input, typically serialized and converted to an xs:string, and $label (if supplied and non-empty) may be output to an ·implementation-defined· destination. (See fn:trace.)

9. Consider a situation in which a user wants to investigate the actual value passed to a function. Assume that in a particular execution, $v is an xs:decimal with value 124.84. Writing fn:trace($v, 'the value of $v is:') will return $v. The processor may output "124.84" and "the value of $v is:" to an ·implementation-defined· destination. (See fn:trace.)

10. Similar to fn:trace, the values of $input, typically serialized and converted to an xs:string, and $label (if supplied and non-empty) may be output to an ·implementation-defined· destination. (See fn:message.)

11. They may provide an ·implementation-defined· mechanism that allows users to choose between raising an error and returning a result that is modulo the largest representable integer value. See [ISO 10967]. (See Arithmetic operators on numeric values.)

12. For xs:decimal values, let N be the number of digits of precision supported by the implementation, and let M (M <= N) be the minimum limit on the number of digits required for conformance (18 digits for XSD 1.0, 16 digits for XSD 1.1). Then for addition, subtraction, and multiplication operations, the returned result should be accurate to N digits of precision, and for division and modulus operations, the returned result should be accurate to at least M digits of precision. The actual precision is ·implementation-defined·. If the number of digits in the mathematical result exceeds the number of digits that the implementation retains for that operation, the result is truncated or rounded in an ·implementation-defined· manner. (See Arithmetic operators on numeric values.)

13. The [IEEE 754-2019] specification also describes handling of two exception conditions called divideByZero and invalidOperation. The IEEE divideByZero exception is raised not only by a direct attempt to divide by zero, but also by operations such as log(0). The IEEE invalidOperation exception is raised by attempts to call a function with an argument that is outside the function’s domain (for example, sqrt(-1) or log(-1)). Although IEEE defines these as exceptions, it also defines “default non-stop exception handling” in which the operation returns a defined result, typically positive or negative infinity, or NaN. With this function library, these IEEE exceptions do not cause a dynamic error at the application level; rather they result in the relevant function or operator returning the defined non-error result. The underlying IEEE exception may be notified to the application or to the user by some ·implementation-defined· warning condition, but the observable effect on an application using the functions and operators defined in this specification is simply to return the defined result (typically -INF, +INF, or NaN) with no error. (See Arithmetic operators on numeric values.)

14. The [IEEE 754-2019] specification distinguishes two NaN values: a quiet NaN and a signaling NaN. These two values are not distinguishable in the XDM model: the value spaces of xs:float and xs:double each include only a single NaN value. This does not prevent the implementation distinguishing them internally, and triggering different ·implementation-defined· warning conditions, but such distinctions do not affect the observable behavior of an application using the functions and operators defined in this specification. (See Arithmetic operators on numeric values.)

15. The implementation may adopt a different algorithm provided that it is equivalent to this formulation in all cases where ·implementation-dependent· or ·implementation-defined· behavior does not affect the outcome, for example, the implementation-defined precision of the result of xs:decimal division. (See op:numeric-integer-divide.)

16. XSD 1.1 allows the string +INF as a representation of positive infinity; XSD 1.0 does not. It is ·implementation-defined· whether XSD 1.1 is supported. (See fn:number.)

17. Any other format token, which indicates a numbering sequence in which that token represents the number 1 (one) (but see the note below). It is ·implementation-defined· which numbering sequences, additional to those listed above, are supported. If an implementation does not support a numbering sequence represented by the given token, it must use a format token of 1. (See fn:format-integer.)

18. For all format tokens other than a digit-pattern, there may be ·implementation-defined· lower and upper bounds on the range of numbers that can be formatted using this format token; indeed, for some numbering sequences there may be intrinsic limits. For example, the format token U+2460 (CIRCLED DIGIT ONE, ①) has a range imposed by the Unicode character repertoire — zero to 20 in Unicode versions prior to 3.2, or zero to 50 in subsequent versions. For the numbering sequences described above any upper bound imposed by the implementation must not be less than 1000 (one thousand) and any lower bound must not be greater than 1. Numbers that fall outside this range must be formatted using the format token 1. (See fn:format-integer.)

19. The set of languages for which numbering is supported is ·implementation-defined·. If the $language argument is absent, or is set to an empty sequence, or is invalid, or is not a language supported by the implementation, then the number is formatted using the default language from the dynamic context. (See fn:format-integer.)

20. ...either a or t, to indicate alphabetic or traditional numbering respectively, the default being ·implementation-defined·. (See fn:format-integer.)

21. The string of characters between the parentheses, if present, is used to select between other possible variations of cardinal or ordinal numbering sequences. The interpretation of this string is ·implementation-defined·. No error occurs if the implementation does not define any interpretation for the defined string. (See fn:format-integer.)

22. It is ·implementation-defined· what combinations of values of the format token, the language, and the cardinal/ordinal modifier are supported. If ordinal numbering is not supported for the combination of the format token, the language, and the string appearing in parentheses, the request is ignored and cardinal numbers are generated instead. (See fn:format-integer.)

23. The use of the a or t modifier disambiguates between numbering sequences that use letters. In many languages there are two commonly used numbering sequences that use letters. One numbering sequence assigns numeric values to letters in alphabetic sequence, and the other assigns numeric values to each letter in some other manner traditional in that language. In English, these would correspond to the numbering sequences specified by the format tokens a and i. In some languages, the first member of each sequence is the same, and so the format token alone would be ambiguous. In the absence of the a or t modifier, the default is ·implementation-defined·. (See fn:format-integer.)

24. The static context provides a set of decimal formats. One of the decimal formats is unnamed, the others (if any) are identified by a QName. There is always an unnamed decimal format available, but its contents are ·implementation-defined·. (See Defining a decimal format.)

25. IEEE states that the preferred quantum is language-defined. In this specification, it is ·implementation-defined·. (See Trigonometric and exponential functions.)

26. IEEE defines various rounding algorithms for inexact results, and states that the choice of rounding direction, and the mechanisms for influencing this choice, are language-defined. In this specification, the rounding direction and any mechanisms for influencing it are ·implementation-defined·. (See Trigonometric and exponential functions.)

27. The map returned by the fn:random-number-generator function may contain additional entries beyond those specified here, but it must match the record type defined above. The meaning of any additional entries is ·implementation-defined·. To avoid conflict with any future version of this specification, the keys of any such entries should start with an underscore character. (See fn:random-number-generator.)

28. If two query parameters use the same keyword then the last one wins. If a query parameter uses a keyword or value which is not defined in this specification then the meaning is ·implementation-defined·. If the implementation recognizes the meaning of the keyword and value then it should interpret it accordingly; if it does not recognize the keyword or value then if the fallback parameter is present with the value no it should reject the collation as unsupported, otherwise it should ignore the unrecognized parameter. (See The Unicode Collation Algorithm.)

29. The following query parameters are defined. If any parameter is absent, the default is ·implementation-defined· except where otherwise stated. The meaning given for each parameter is non-normative; the normative specification is found in [UTS #35]. (See The Unicode Collation Algorithm.)

30. Because the set of collations that are supported is ·implementation-defined·, an implementation has the option to support all collation URIs, in which case it will never raise this error. (See Choosing a collation.)

31. The properties available are as defined for the Unicode Collation Algorithm (see 5.3.4 The Unicode Collation Algorithm). Additional ·implementation-defined· properties may be specified as described in the rules for UCA collation URIs. (See fn:collation.)

32. It is possible to define collations that do not have the ability to generate collation keys. Supplying such a collation will cause the function to fail. The ability to generate collation keys is an ·implementation-defined· property of the collation. (See fn:collation-key.)

33. Conforming implementations must support normalization form NFC and may support normalization forms NFD, NFKC, NFKD, and FULLY-NORMALIZED. They may also support other normalization forms with ·implementation-defined· semantics. (See fn:normalize-unicode.)

34. It is ·implementation-defined· which version of Unicode (and therefore, of the normalization algorithms and their underlying data) is supported by the implementation. See [UAX #15] for details of the stability policy regarding changes to the normalization rules in future versions of Unicode. If the input string contains codepoints that are unassigned in the relevant version of Unicode, or for which no normalization rules are defined, the fn:normalize-unicode function leaves such codepoints unchanged. If the implementation supports the requested normalization form then it must be able to handle every input string without raising an error. (See fn:normalize-unicode.)

35. It is possible to define collations that do not have the ability to decompose a string into units suitable for substring matching. An argument to a function defined in this section may be a URI that identifies a collation that is able to compare two strings, but that does not have the capability to split the string into collation units. Such a collation may cause the function to fail, or to give unexpected results, or it may be rejected as an unsuitable argument. The ability to decompose strings into collation units is an ·implementation-defined· property of the collation. The fn:collation-available function can be used to ask whether a particular collation has this property. (See Functions based on substring matching.)

36. The result of the function will always be such that validation against this schema would succeed. However, it is ·implementation-defined· whether the result is typed or untyped, that is, whether the elements and attributes in the returned tree have type annotations that reflect the result of validating against this schema. (See fn:analyze-string.)

37. Some URI schemes are hierarchical and some are non-hierarchical. Implementations must treat the following schemes as non-hierarchical: jar, mailto, news, tag, tel, and urn. Whether additional schemes are known to be non-hierarchical ·implementation-defined·. If a scheme is not known to be non-hierarchical, it must be treated as hierarchical. (See Parsing and building URIs.)

38. If the omit-default-ports option is true, the port is discarded and set to the empty sequence if the port number is the same as the default port for the given scheme. Implementations should recognize the default ports for http (80), https (443), ftp (21), and ssh (22). Exactly which ports are recognized is ·implementation-defined·. (See fn:parse-uri.)

39. If the omit-default-ports option is true then the $port is set to the empty sequence if the port number is the same as the default port for the given scheme. Implementations should recognize the default ports for http (80), https (443), ftp (21), and ssh (22). Exactly which ports are recognized is ·implementation-defined·. (See fn:build-uri.)

40. Processors may support a greater range and/or precision. The limits are ·implementation-defined·. (See Limits and precision.)

41. Similarly, a processor may be unable accurately to represent the result of dividing a duration by 2, or multiplying a duration by 0.5. A processor that limits the precision of the seconds component of duration values must deliver a result that is as close as possible to the mathematically precise result, given these limits; if two values are equally close, the one that is chosen is ·implementation-defined·. (See Limits and precision.)

42. All conforming processors must support year values in the range 1 to 9999, and a minimum fractional second precision of 1 millisecond or three digits (i.e., s.sss). However, processors may set larger ·implementation-defined· limits on the maximum number of digits they support in these two situations. Processors may also choose to support the year 0 and years with negative values. The results of operations on dates that cross the year 0 are ·implementation-defined·. (See Limits and precision.)

43. Similarly, a processor that limits the precision of the seconds component of date and time or duration values may need to deliver a rounded result for arithmetic operations. Such a processor must deliver a result that is as close as possible to the mathematically precise result, given these limits: if two values are equally close, the one that is chosen is ·implementation-defined·. (See Limits and precision.)

44. ...the format token n, N, or Nn, indicating that the value of the component is to be output by name, in lower-case, upper-case, or title-case respectively. Components that can be output by name include (but are not limited to) months, days of the week, timezones, and eras. If the processor cannot output these components by name for the chosen calendar and language then it must use an ·implementation-defined· fallback representation. (See The picture string.)

45. ...indicates alphabetic or traditional numbering respectively, the default being ·implementation-defined·. This has the same meaning as in the second argument of fn:format-integer. (See The picture string.)

46. The sequence of characters in the (adjusted) first presentation modifier is reversed (for example, 999'### becomes ###'999). If the result is not a valid decimal digit pattern, then the output is ·implementation-defined·. (See Formatting Fractional Seconds.)

47. The output for these components is entirely ·implementation-defined·. The default presentation modifier for these components is n, indicating that they are output as names (or conventional abbreviations), and the chosen names will in many cases depend on the chosen language: see 9.8.4.8 The language, calendar, and place arguments. (See Formatting Other Components.)

48. The set of languages, calendars, and places that are supported in the ·date formatting functions· is ·implementation-defined·. When any of these arguments is omitted or is an empty sequence, an ·implementation-defined· default value is used. (See The language, calendar, and place arguments.)

49. The choice of the names and abbreviations used in any given language is ·implementation-defined·. For example, one implementation might abbreviate July as Jul while another uses Jly. In German, one implementation might represent Saturday as Samstag while another uses Sonnabend. Implementations may provide mechanisms allowing users to control such choices. (See The language, calendar, and place arguments.)

50. The choice of the names and abbreviations used in any given language for calendar units such as days of the week and months of the year is ·implementation-defined·. (See The language, calendar, and place arguments.)

51. The calendar value if present must be a valid EQName (dynamic error: [err:FOFD1340]). If it is a lexical QName then it is expanded into an expanded QName using the statically known namespaces; if it has no prefix then it represents an expanded-QName in no namespace. If the expanded QName is in no namespace, then it must identify a calendar with a designator specified below (dynamic error: [err:FOFD1340]). If the expanded QName is in a namespace then it identifies the calendar in an ·implementation-defined· way. (See The language, calendar, and place arguments.)

52. At least one of the above calendars must be supported. It is ·implementation-defined· which calendars are supported. (See The language, calendar, and place arguments.)

53. The requirement to deliver a deterministic result has performance implications, and for this reason implementations may provide a user option to evaluate the function without a guarantee of determinism. The manner in which any such option is provided is ·implementation-defined·. If the user has not selected such an option, a call of the function must either return a deterministic result or must raise a dynamic error [err:FODC0003]. (See fn:doc.)

54. Various aspects of this processing are ·implementation-defined·. Implementations may provide external configuration options that allow any aspect of the processing to be controlled by the user. In particular:... (See fn:doc.)

55. It is ·implementation-defined· whether DTD validation and/or schema validation is applied to the source document. (See fn:doc.)

56. The effect of a fragment identifier in the supplied URI is ·implementation-defined·. One possible interpretation is to treat the fragment identifier as an ID attribute value, and to return a document node having the element with the selected ID value as its only child. (See fn:doc.)

57. By default, this function is ·deterministic·. This means that repeated calls on the function with the same argument will return the same result. However, for performance reasons, implementations may provide a user option to evaluate the function without a guarantee of determinism. The manner in which any such option is provided is ·implementation-defined·. If the user has not selected such an option, a call to this function must either return a deterministic result or must raise a dynamic error [err:FODC0003]. (See fn:collection.)

58. By default, this function is ·deterministic·. This means that repeated calls on the function with the same argument will return the same result. However, for performance reasons, implementations may provide a user option to evaluate the function without a guarantee of determinism. The manner in which any such option is provided is ·implementation-defined·. If the user has not selected such an option, a call to this function must either return a deterministic result or must raise a dynamic error [err:FODC0003]. (See fn:uri-collection.)

59. ...the processor may use ·implementation-defined· heuristics to determine the likely encoding, otherwise... (See fn:unparsed-text.)

60. The fact that the resolution of URIs is defined by a mapping in the dynamic context means that in effect, various aspects of the behavior of this function are ·implementation-defined·. Implementations may provide external configuration options that allow any aspect of the processing to be controlled by the user. In particular:... (See fn:unparsed-text.)

61. The collation used for matching names is ·implementation-defined·, but must be the same as the collation used to ensure that the names of all environment variables are unique. (See fn:environment-variable.)

62. Except to the extent defined by these options, the precise process used to construct the XDM instance is ·implementation-defined·. In particular, it is implementation-defined whether an XML 1.0 or XML 1.1 parser is used. (See fn:parse-xml.)

63. Except as explicitly defined, the precise process used to construct the XDM instance is ·implementation-defined·. In particular, it is implementation-defined whether an XML 1.0 or XML 1.1 parser is used. (See fn:parse-xml-fragment.)

64. If the second argument is omitted, or is supplied in the form of an output:serialization-parameters element, then the values of any serialization parameters that are not explicitly specified is ·implementation-defined·, and may depend on the context. (See fn:serialize.)

65. Because the [DOM: Living Standard] and [HTML: Living Standard] are not fixed, it is ·implementation-defined· which versions are used. (See XDM Mapping from HTML DOM Nodes.)

66. If an implementation allows these nodes to be passed in via an API or similar mechanism, their behaviour is ·implementation-defined·. (See XDM Mapping from HTML DOM Nodes.)

67. If the local name contains a character that is not a valid XML NameStartChar or NameChar, then an ·implementation-defined· replacement string is used. The result must be a valid NCName. (See node-name Accessor.)

68. If the local name contains a character that is not a valid XML NameStartChar or NameChar, then an ·implementation-defined· replacement string is used. The result must be a valid NCName. (See node-name Accessor.)

69. Additional ·implementation-defined· parser options are allowed. (See HTML parser options.)

70. Any other method and html-version combinations are ·implementation-defined·. (See fn:parse-html.)

71. The default behaviour is ·implementation-defined·. (See fn:parse-html.)

72. An ·implementation-defined· parsing algorithm, tree construction, and validation consistent with the specified HTML version. (See fn:parse-html.)

73. The input may contain deviations from the grammar of [RFC 7159], which are handled in an ·implementation-defined· way. (Note: some popular extensions include allowing quotes on keys to be omitted, allowing a comma to appear after the last item in an array, allowing leading zeroes in numbers, and allowing control characters such as tab and newline to be present in unescaped form.) Since the extensions accepted are implementation-defined, an error may be raised [err:FOJS0001] if the input does not conform to the grammar. (See fn:parse-json.)

74. The supplied function is called to process the string value of any JSON number in the input. By default, numbers are processed by converting to xs:double using the XPath casting rules. Supplying the value xs:decimal#1 will instead convert to xs:decimal (which potentially retains more precision, but disallows exponential notation), while supplying a function that casts to (xs:decimal | xs:double) will treat the value as xs:decimal if there is no exponent, or as xs:double otherwise. Supplying the value fn:identity#1 causes the value to be retained unchanged as an xs:untypedAtomic. If the liberal option is false (the default), then the supplied number-parser is called if and only if the value conforms to the JSON grammar for numbers (for example, a leading plus sign and redundant leading zeroes are not allowed). If the liberal option is true then it is also called if the value conforms to an ·implementation-defined· extension of this grammar. (See fn:parse-json.)

75. The input may contain deviations from the grammar of [RFC 7159], which are handled in an ·implementation-defined· way. (Note: some popular extensions include allowing quotes on keys to be omitted, allowing a comma to appear after the last item in an array, allowing leading zeroes in numbers, and allowing control characters such as tab and newline to be present in unescaped form.) Since the extensions accepted are implementation-defined, an error may be raised (see below) if the input does not conform to the grammar. (See fn:json-to-xml.)

76. Default: ·Implementation-defined·. (See fn:json-to-xml.)

77. Indicates that the resulting XDM instance must be typed; that is, the element and attribute nodes must carry the type annotations that result from validation against the schema given at C.2 Schema for the result of fn:json-to-xml, or against an ·implementation-defined· schema if the liberal option has the value true. (See fn:json-to-xml.)

78. The result of the function will always be such that validation against this schema would succeed. However, it is ·implementation-defined· whether the result is typed or untyped, that is, whether the elements and attributes in the returned tree have type annotations that reflect the result of validating against this schema. (See fn:csv-to-xml.)

79. Additional, ·implementation-defined· options may be available, for example, to control aspects of the XML serialization, to specify the grammar start symbol, or to produce output formats other than XML. (See fn:invisible-xml.)

80. If the arguments to fn:function-lookup identify a function that is present in the static context of the function call, the function will always return the same function that a static reference to this function would bind to. If there is no such function in the static context, then the results depend on what is present in the dynamic context, which is ·implementation-defined·. (See fn:function-lookup.)

81. Default: The version given in the prolog of the library module; or ·implementation-defined· if this is absent. (See fn:load-xquery-module.)

82. A sequence of URIs (in the form of xs:string values) which may be used or ignored in an ·implementation-defined· way.... (See fn:load-xquery-module.)

83. Values for vendor-defined configuration options for the XQuery processor used to process the request. The key is the name of an option, expressed as a QName: the namespace URI of the QName should be a URI controlled by the vendor of the XQuery processor. The meaning of the associated value is ·implementation-defined·. Implementations should ignore options whose names are in an unrecognized namespace. The ·option parameter conventions· do not apply to this contained map.... (See fn:load-xquery-module.)

84. It is ·implementation-defined· whether constructs in the library module are evaluated in the same ·execution scope· as the calling module. (See fn:load-xquery-module.)

85. The library module that is loaded may import schema declarations using an import schema declaration. It is ·implementation-defined· whether schema components in the in-scope schema definitions of the calling module are automatically added to the in-scope schema definitions of the dynamically loaded module. The in-scope schema definitions of the calling and called modules must be consistent, according to the rules defined in Section 2.2.5 Consistency Constraints ^XQ31. (See fn:load-xquery-module.)

86. Default: ·Implementation-defined·. (See fn:transform.)

87. Default: ·Implementation-defined·. (See fn:transform.)

88. If the implementation provides a way of writing or invoking functions with side-effects, this post-processing function might be used to save a copy of the result document to persistent storage. For example, if the implementation provides access to the EXPath File library [EXPath], then a serialized document might be written to filestore by calling the file:write function. Similar mechanisms might be used to issue an HTTP POST request that posts the result to an HTTP server, or to send the document to an email recipient. The semantics of calling functions with side-effects are entirely ·implementation-defined·. (See fn:transform.)

89. Calls to fn:transform can potentially have side-effects even in the absence of the post-processing option, because the XSLT specification allows a stylesheet to invoke extension functions that have side-effects. The semantics in this case are ·implementation-defined·. (See fn:transform.)

90. A string intended to be used as the static base URI of the principal stylesheet module. This value must be used if no other static base URI is available. If the supplied stylesheet already has a base URI (which will generally be the case if the stylesheet is supplied using stylesheet-node or stylesheet-location) then it is ·implementation-defined· whether this parameter has any effect. If the value is a relative reference, it is resolved against the static base URI of the fn:transform function call.... (See fn:transform.)

91. Values for vendor-defined configuration options for the XSLT processor used to process the request. The key is the name of an option, expressed as a QName: the namespace URI of the QName should be a URI controlled by the vendor of the XSLT processor. The meaning of the associated value is ·implementation-defined·. Implementations should ignore options whose names are in an unrecognized namespace. Default is an empty map.... (See fn:transform.)

92. It is ·implementation-defined· whether the XSLT transformation is executed within the same ·execution scope· as the calling code. (See fn:transform.)

93. XSLT 1.0 does not define any error codes, so this is the likely outcome with an XSLT 1.0 processor. XSLT 2.0 and 3.0 do define error codes, but some APIs do not expose them. If multiple errors are signaled by the transformation (which is most likely to happen with static errors) then the error code should where possible be that of one of these errors, chosen arbitrarily; the processor may make details of additional errors available to the application in an ·implementation-defined· way. (See fn:transform.)

94. If ST is xs:float or xs:double, then TV is the xs:decimal value, within the set of xs:decimal values that the implementation is capable of representing, that is numerically closest to SV. If two values are equally close, then the one that is closest to zero is chosen. If SV is too large to be accommodated as an xs:decimal, (see [XML Schema Part 2: Datatypes Second Edition] for ·implementation-defined· limits on numeric values) a dynamic error is raised [err:FOCA0001]. If SV is one of the special xs:float or xs:double values NaN, INF, or -INF, a dynamic error is raised [err:FOCA0002]. (See Casting to xs:decimal.)

95. In casting to xs:decimal or to a type derived from xs:decimal, if the value is not too large or too small but nevertheless cannot be represented accurately with the number of decimal digits available to the implementation, the implementation may round to the nearest representable value or may raise a dynamic error [err:FOCA0006]. The choice of rounding algorithm and the choice between rounding and error behavior is ·implementation-defined·. (See Casting from xs:string and xs:untypedAtomic.)

96. If ST is xs:decimal, xs:float or xs:double, then TV is SV with the fractional part discarded and the value converted to xs:integer. Thus, casting 3.1456 returns 3 while -17.89 returns -17. Casting 3.124E1 returns 31. If SV is too large to be accommodated as an integer, (see [XML Schema Part 2: Datatypes Second Edition] for ·implementation-defined· limits on numeric values) a dynamic error is raised [err:FOCA0003]. If SV is one of the special xs:float or xs:double values NaN, INF, or -INF, a dynamic error is raised [err:FOCA0002]. (See Casting to xs:integer.)

97. The tz timezone database, available at http://www.iana.org/time-zones. It is ·implementation-defined· which version of the database is used. (See IANA Timezone Database.)

98. Unicode Standard Annex #15: Unicode Normalization Forms. Ed. Mark Davis and Ken Whistler, Unicode Consortium. The current version is 9.0.0, dated 2016-02-24. As with [The Unicode Standard], the version to be used is ·implementation-defined·. Available at: http://www.unicode.org/reports/tr15/. (See UAX #15.)

99. Unicode Standard Annex #29: Unicode Text Segmentation. Ed. Josh Hadley, Unicode Consortium. The current version is 15.1.0, dated 2023-08-16. As with [The Unicode Standard], the version to be used is ·implementation-defined·. Available at: http://www.unicode.org/reports/tr29/. (See UAX #29.)

100. The Unicode Consortium, Reading, MA, Addison-Wesley, 2016. The Unicode Standard as updated from time to time by the publication of new versions. See http://www.unicode.org/standard/versions/ for the latest version and additional information on versions of the standard and of the Unicode Character Database. The version of Unicode to be used is ·implementation-defined·, but implementations are recommended to use the latest Unicode version; currently, Version 9.0.0. (See The Unicode Standard.)

101. Unicode Technical Standard #10: Unicode Collation Algorithm. Ed. Mark Davis and Ken Whistler, Unicode Consortium. The current version is 9.0.0, dated 2016-05-18. As with [The Unicode Standard], the version to be used is ·implementation-defined·. Available at: http://www.unicode.org/reports/tr10/. (See UTS #10.)

102. Unicode Technical Standard #35: Unicode Locale Data Markup Language. Ed Mark Davis et al, Unicode Consortium. The current version is 29, dated 2016-03-15. As with [The Unicode Standard], the version to be used is ·implementation-defined·. Available at: http://www.unicode.org/reports/tr35/. (See UTS #35.)