P2314R4: Character sets and encodings

Introduction

• Switch C++ to a modified "model C" approach for universal-character-names as described in the C99 Rationale v5.10, section 5.2.1.
• Introduce the term "literal encoding". For purposes of the C++ specification, the actual set of characters is not relevant, but the sequence of code units (i.e. the encoding) specified by a given character or string literal are. The terms "execution (wide) character set" are retained to describe the locale-dependent runtime character set used by functions such as isalpha.
• (Not a wording change) Do not attempt to treat all string literals the same; their treatment depends on (phase 7) context.

• 578. Phase 1 replacement of characters with universal-character-names
• 1332. Handling of invalid universal-character-names
• 1335. Stringizing, extended characters, and universal-character-names
• 1403. Universal-character-names in comments
• 2455. Concatenation of string literals vs translation phases 5 and 6

Change history

Changes since R0

• missed edits in the normative wording
• add comparison with P2297R0
• retained "execution (wide) character set" for locale-dependent runtime encoding, but moved the definition to the library wording

Changes since R1

• missed edits in the normative wording
• Clarify that an r-char-sequence never contains a universal-character-name.
• Remove words that claim string literal objects are initialized in translation phase 5.
• Add SG16 and EWG poll results.
• Fix typo in [lex.ccon] table 10.

Changes since R2

• Feedback from CWG teleconference on 2021-09-14

Changes since R3

• Update for applied P1949
• Reflected review comments from Hubert Tong
• Feedback from CWG teleconference on 2021-09-21

Terminology changes

• translation character set: the abstract character set used during translation; can represent the character equivalent of all valid universal-character-names
• basic character set: minimum character set needed to express C++ program source
• basic literal character set: minimum set of characters expressible by literals
• ordinary / wide literal encoding: compile-time encoding used for initializing string literal objects

Behavior changes

This changes the specified behavior of the stringizing preprocessor operator [cpp.stringize] as follows

C++20	this paper
#define S(x) # x const char * s1 = S(Köppe); // "K\\u00f6ppe" const char * s2 = S(K\u00f6ppe); // "K\\u00f6ppe"	#define S(x) # x const char * s1 = S(Köppe); // "Köppe" const char * s2 = S(K\u00f6ppe); // "Köppe"

However, it turns out that all major implementations already implement what this paper specifies, i.e. no implementation provides an escaped UCN.

Not all string-literals are the same

Context	Destination
asm-declaration	build environment
#include "fn" or #include <fn>	file name
language linkage	translation
operator "" [over.literal]	translation
#line directive	diagnostic
argument for [[nodiscard]] and [[deprecated]]	diagnostic
#error, static_assert	diagnostic
__FILE__, __func__	literal encoding
std::typeinfo::name()	literal encoding
character-literal or string-literal appearing elsewhere	literal encoding
user-defined-literal	literal encoding

The destinations have the following meaning:

• build environment: A string likely passed as text (program input) to another component in the build environment.
• file name: A file name suitable for the build environment.
• translation: No use outside the compiler.
• diagnostic: Diagnostic text; kept in the translation character set until needed for output; then treated the same as e.g. identifiers appearing in diagnostic messages
• literal encoding: The implementation-defined encoding for the runtime environment, used for string-literals that appear in usual program text.

Comparison with P2297R0

The paper P2297R0 "Wording improvements for encodings and character sets" by Corentin Jabot has overlap with this paper. The main differences are:

• terminology: This paper uses "translation character set"; P2297R0 uses "Unicode" to describe the C++ compile-time character set. Note that ISO 10646 does not seem to define "Unicode" as a stand-alone term, and the term seems unclear regarding the inclusion of non-assigned code points.
• alert and backspace for literals: This paper retains the status quo, after reformulation to refer to Unicode code points; P2297R0 removes the requirement to represent these. In my view, since simple-escape-sequences are expressly specified to represent alert and backspace, those characters should be required to be representable in the literal encoding. The presence or absence of such a specified requirement is believed not to have an impact on currently existing implementations.

Poll results

SG16

  SF F N A SA
   2 4 1 0  1

  SF F N A SA
   3 5 0 0  0

EWG

SF 	F 	N 	A 	SA
5 	6 	0 	0 	0

Wording changes

sequence of one or more bytes representing a member of the extended character set of either the source or the execution environment the code unit sequence for an encoded character of the execution character set

...

3. The source file is decomposed into preprocessing tokens (5.4 [lex.pptoken]) and sequences of white-space characters (including comments). A source file shall not end in a partial preprocessing token or in a partial comment. [ Footnote: ... ] Each comment is replaced by one space character. New-line characters are retained. Whether each nonempty sequence of white-space characters other than new-line is retained or replaced by one space character is unspecified. As characters from the source file are consumed to form the next preprocessing token (i.e., not being consumed as part of a comment or other forms of whitespace), except when matching a c-char-sequence, s-char-sequence, r-char-sequence, h-char-sequence, or q-char-sequence, universal-character-names are recognized and replaced by the designated element of the translation character set. The process of dividing a source file’s characters into preprocessing tokens is context-dependent. [Example: See the handling of < within a #include preprocessing directive. — end example]

4. Preprocessing directives are executed, macro invocations are expanded, and _Pragma unary operator expressions are executed. If a character sequence that matches the syntax of a universal-character-name is produced by token concatenation (15.6.3 [cpp.concat]), the behavior is undefined. A #include preprocessing directive causes the named header or source file to be processed from phase 1 through phase 4, recursively. All preprocessing directives are then deleted.

5. For a sequence of two or more adjacent string-literal tokens, a common encoding-prefix is determined as specified in 5.13.5 [lex.string]. Each such string-literal token is then considered to have that common encoding-prefix. Each basic-c-char, basic-s-char, and r-char in a character-literal or a string-literal, as well as each escape-sequence and universal-character-name in a character-literal or a non-raw string literal, is encoded in the literal’s associated character encoding as specified in 5.13.3 [lex.ccon] and 5.13.5 [lex.string].

• each character named by ISO/IEC 10646, as identified by its unique UCS scalar value, and
• a distinct character for each UCS scalar value where no named character is assigned.

2 The basic character set is a subset of the translation character set, consisting of 96 characters as specified in table X. [ Note: Unicode short names are given only as a means to identifying the character; the numerical value has no other meaning in this context. -- end note ]

The universal-character-name construct provides a way to name other characters.

hex-quad :
    hexadecimal-digit hexadecimal-digit hexadecimal-digit hexadecimal-digit
  
universal-character-name :
    \u hex-quad
    \U hex-quad hex-quad

A universal-character-name designates the character in ISO/IEC 10646 (if any) the translation character set whose Unicode code point UCS scalar value is the hexadecimal number represented by the sequence of hexadecimal-digits in the universal-character-name. The program is ill-formed if that number is not a Unicode code point or if it is a surrogate code point UCS scalar value. Noncharacter code points and reserved code points are considered to designate separate characters distinct from any ISO/IEC 10646 character. If a universal-character-name outside the c-char-sequence, s-char-sequence, or r-char-sequence of a character-literal or string-literal (in either case, including within a user-defined-literal) corresponds to a control character or to a character in the basic source character set, the program is ill-formed. [ Footnote: Note: A sequence of characters resembling a universal-character-name in an r-char-sequence (5.13.5) does not form a universal-character-name. ] [Note: ISO/IEC 10646 code points are integers in the range [0, 10FFFF] (hexadecimal). A surrogate code point is a value in the range [D800, DFFF] (hexadecimal). A control character is a character whose code point is in either of the ranges [0, 1F] or [7F, 9F] (hexadecimal). — end note]

A code unit is an integer value of character type (6.8.1 [basic.fundamental]). Characters in a character-literal other than a multicharacter or non-encodable character literal or in a string-literal are encoded as a sequence of one or more code units, as determined by the encoding-prefix ([lex.ccon], [lex.string]); this is termed the respective literal encoding. The ordinary literal encoding is the encoding applied to an ordinary character or string literal. The wide literal encoding is the encoding applied to a wide character or string literal.

The basic execution character set and the basic execution wide-character set shall each contain all the members of the basic source character set, plus control characters representing alert, backspace, and carriage return, plus a null character (respectively, null wide character), whose value is 0. For each basic execution character set, the values of the members shall be non-negative and distinct from one another. In both the source and execution basic character sets, the value of each character after 0 in the above list of decimal digits shall be one greater than the value of the previous. The execution character set and the execution wide-character set are implementation-defined supersets of the basic execution character set and the basic execution wide-character set, respectively. The values of the members of the execution character sets and the sets of additional members are locale-specific.

preprocessing-token:
  header-name
  import-keyword
  module-keyword
  export-keyword
  identifier
  pp-number
  character-literal
  user-defined-character-literal
  string-literal
  user-defined-string-literal
  preprocessing-op-or-punc
  each universal-character-name that cannot be one of the above
  each non-whitespace character that cannot be one of the above

A preprocessing token is the minimal lexical element of the language in translation phases 3 through 6. In this document, glyphs are used to identify elements of the basic character set ([lex.charset]). The categories of preprocessing token are: header names, placeholder tokens produced by preprocessing import and module directives (import-keyword, module-keyword, and export-keyword), identifiers, preprocessing numbers, character literals (including user-defined character literals), string literals (including user-defined string literals), preprocessing operators and punctuators, and single universal-character-names and non-whitespace characters that do not lexically match the other preprocessing token categories. If a single universal-character-name does not match any of the other preprocessing token categories, the program is ill-formed. If a ' or a "U+0027 APOSTROPHE or a U+0022 QUOTATION MARK character matches the last category, the behavior is undefined. If any character not in the basic character set matches the last category, the program is ill-formed. Preprocessing tokens can be separated by whitespace; this consists of comments (5.7), or whitespace characters (space, horizontal tabU+0020 SPACE, U+0009 CHARACTER TABULATION, new-line, vertical tab, and form-feedU+000B LINE TABULATION, and U+000C FORM FEED), or both. ...

h-char:
    any member of the source translation character set except new-line and > U+003E GREATER-THAN SIGN
...
q-char:
    any member of the source translation character set except new-line and " U+0022 QUOTATION MARK

identifier-start:
    nondigit
    universal-character-name
    an element of the translation character set of class XID_Start

identifier-continue:
    digit
    nondigit
    universal-character-name
    an element of the translation character set of class XID_Continue

basic-c-char:
    any member of the basic source translation character set
    except the single-quote ’, backslash \ U+0027 APOSTROPHE, U+005C REVERSE SOLIDUS, or new-line character
...
conditional-escape-sequence-char:
    any member of the basic source character set that is not an octal-digit, a simple-escape-sequence-char, or
    the characters u, U, or x

[Note 1 : The associated character encoding for ordinary and wide character literals determines encodability, but does not determine the value of non-encodable ordinary or wide character literals or ordinary or wide multicharacter literals. The examples in Table 9 for non-encodable ordinary and wide character literals assume that the specified character lacks representation in the execution character set ordinary literal encoding or execution wide-character set wide literal encoding, respectively, or that encoding it would require more than one code unit. — end note]

Encoding prefix	...	Associated character encoding
none	...	encoding of the execution character set ordinary literal encoding
L	...	encoding of the execution wide-character set wide literal encoding

basic-s-char:
    any member of the basic source translation character set
    except the double-quote ", backslash \U+0022 QUOTATION MARK, U+005C REVERSE SOLIDUS, or new-line character
...
r-char:
    any member of the source translation character set, except a right parenthesis ) U+0029 RIGHT PARENTHESIS followed by
    the initial d-char-sequence (which may be empty) followed by a double quote " U+0022 QUOTATION MARK.
...
d-char:
    any member of the basic source character set except:
    space, the left parenthesis (, the right parenthesis ), the backslash \, and the control characters
    representing horizontal tab, vertical tab, form feed
    U+0020 SPACE, U+0028 LEFT PARENTHESIS, U+0029 RIGHT PARENTHESIS, U+005C REVERSE SOLIDUS,
    U+0009 CHARACTER TABULATION, U+000B LINE TABULATION, U+000C FORM FEED (FF), and new-line

Encoding prefix	...	Associated character encoding
none	...	encoding of the execution character set ordinary literal encoding
L	...	encoding of the execution widecharacter set wide literal encoding

Table 13 has some examples of valid concatenations.

- 8 - In translation phase 6 (5.2 [lex.phases]), adjacent string-literals are concatenated. The lexical structure and grouping of the contents of the individual string-literals is retained. Characters in concatenated strings are kept distinct. [Example:

"\xA" "B"

R"(\u00)" "41"

Table 13 has some examples of valid concatenations. — end example]

When encoding a stateful character encoding, ...

[ Note: The sequence c₁ c₂ ...c_k can only contain characters from the basic source character set. — end note]

[ Note: The sequence c₁ c₂ ...c_k can only contain characters from the basic source character set. — end note]

The fundamental storage unit in the memory model is the byte. A byte is at least large enough to contain any member the ordinary literal encoding of any element of the basic execution literal character set (5.3) and the eight-bit code units of the Unicode UTF-8 encoding form and is composed of a contiguous sequence of bits, [ Footnote: ... ] the number of which is implementation-defined.

Type char is a distinct type that has an implementation-defined choice of “signed char” or “unsigned char” as its underlying type. The values of type char can represent distinct codes for all members of the implementation’s basic character set. ...

Type wchar_t is a distinct type that has an implementation-defined signed or unsigned integer type as its underlying type. The values of type wchar_t can represent distinct codes for all members of the largest extended character set specified among the supported locales (28.3.1).

The types char, wchar_t, char8_t, char16_t, char32_t are collectively called character types. The character types, Types bool, char, wchar_t, char8_t, char16_t, char32_t, and the signed and unsigned integer types are collectively called integral types. A synonym for integral type is integer type. [Note: Enumerations (9.7.1) are not integral; however, unscoped enumerations can be promoted to integral types as specified in 7.3.6. — end note]

A literal is a primary expression. The type of a literal is determined based on its form as specified in 5.13 [lex.literal]. A string-literal is an lvalue designating a corresponding string literal object ([lex.string]), a user-defined-literal has the same value category as the corresponding operator call expression described in 5.13.8 [lex.ext], and any other literal is a prvalue.

The resulting tokens comprise the controlling constant expression which is evaluated according to the rules of 7.7 using arithmetic that has at least the ranges specified in 17.3. For the purposes of this token conversion and evaluation all signed and unsigned integer types act as if they have the same representation as, respectively, intmax_t or uintmax_t (17.4). [Note: ... -- end note] This includes interpreting character-literals, which may involve converting escape sequences into execution character set members interpreting escape-sequences and universal-character-names (5.13.3 [lex.ccon]). Whether the numeric value for these character-literals matches the value obtained when an identical character-literal occurs in an expression (other than within a #if or #elif directive) is implementation-defined. [Note: ... -- end note] Also, whether a single-character character-literal may have a negative value is implementation-defined. Each subexpression with type bool is subjected to integral promotion before processing continues.

For both object-like and function-like macro invocations, before the replacement list is reexamined for more macro names to replace, each instance of a ## preprocessing token in the replacement list (not from an argument) is deleted and the preceding preprocessing token is concatenated with the following preprocessing token. Placemarker preprocessing tokens are handled specially: concatenation of two placemarkers results in a single placemarker preprocessing token, and concatenation of a placemarker with a non-placemarker preprocessing token results in the non-placemarker preprocessing token. If the result begins with a sequence matching the syntax of universal-character-name, the behavior is undefined. [ Note: This determination does not consider the replacement of universal-character-names in translation phase 3 ([lex.phases]). ] If the result is not a valid preprocessing token, the behavior is undefined. The resulting token is available for further macro replacement. The order of evaluation of ## operators is unspecified.

The C standard library makes widespread use of characters and character sequences that follow a few uniform conventions:

• Properties specified as locale-specific may change during program execution by a call to setlocale(int, const char*) (28.5.1 [clocale.syn]), or by a change to a locale object, as described in 28.3 [locales] and Clause 29 [input.output].
• The execution character set and the execution wide-character set are supersets of the basic literal character set (5.3 [lex.charset]). The encodings of the execution character sets and the sets of additional elements (if any) are locale-specific. [ Note: The encodings of the execution character sets can be unrelated to any literal encoding. -- end note ]
• A letter is any of the 26 lowercase or 26 uppercase letters in the basic execution character set.
• The decimal-point character is the locale-specific (single-byte) character used by functions that convert between a (single-byte) character sequence and a value of one of the floating-point types. It is used in the character sequence to denote the beginning of a fractional part. It is represented in Clause 17 through Clause 32 and Annex D by a period, ’.’, which is also its value in the "C" locale, but may change during program execution by a call to setlocale(int, const char*), [ Footnote: ... ] or by a change to a locale object, as described in 28.3 and Clause 29.

A null-terminated multibyte string, or ntmbs, is an ntbs that constitutes a sequence of valid multibyte characters, beginning and ending in the initial shift state. [ Footnote: An NTBS that contains characters only from the basic execution literal character set is also an NTMBS. Each multibyte character then consists of a single byte. ]

%Z	The time zone abbreviation or name. A single word is parsed. This word can only contain characters from the basic source character set (5.3 [lex.charset]) that are alphanumeric, or one of ’_’, ’/’, ’-’, or ’+’.

[...]

do_widen(do_narrow(c, 0)) == c

Affected subclause: 5.2
Change: Removal of trigraph support as a required feature.
Rationale: Prevents accidental uses of trigraphs in non-raw string literals and comments. Effect on original feature: Valid C ++ 2014 code that uses trigraphs may not be valid or may have different semantics in this revision of C ++ . Implementations may choose to translate trigraphs as specified in C ++ 2014 if they appear outside of a raw string literal, as part of the implementation-defined mapping from physical source file characters to the basic source character set.

Acknowledgements