Phases of translation

< cpp‎ | language
C++ language
General topics
Flow control
Conditional execution statements
Iteration statements (loops)
range-for (C++11)
Jump statements
Function declaration
Lambda function expression
inline specifier
Dynamic exception specifications (until C++17*)
noexcept specifier (C++11)
decltype (C++11)
auto (C++11)
alignas (C++11)
Storage duration specifiers
Alternative representations
Boolean - Integer - Floating-point
Character - String - nullptr (C++11)
User-defined (C++11)
Attributes (C++11)
typedef declaration
Type alias declaration (C++11)
Implicit conversions - Explicit conversions
static_cast - dynamic_cast
const_cast - reinterpret_cast
Memory allocation
Class-specific function properties
explicit (C++11)
Special member functions

The C++ source file is processed by the compiler as if the following phases take place, in this exact order:


[edit] Phase 1

1) The individual bytes of the source code file are mapped (in implementation-defined manner) to the characters of the basic source character set. In particular, OS-dependent end-of-line indicators are replaced by newline characters.
2) The set of source file characters accepted is implementation-defined(since C++11). Any source file character that cannot be mapped to a character in the basic source character set is replaced by its universal character name (escaped with \u or \U) or by some implementation-defined form that is handled equivalently.
3) Trigraph sequences are replaced by corresponding single-character representations.
(until C++17)
(until C++23)

Input files that are a sequence of UTF-8 code units (UTF-8 files) are guaranteed to be supported. The set of other supported kinds of input files is implementation-defined. If the set is non-empty, the kind of an input file is determined in an implementation-defined manner that includes a means of designating input files as UTF-8 files, independent of their content (recognizing the byte order mark is not sufficient).

  • If an input file is determined to be a UTF-8 file, then it shall be a well-formed UTF-8 code unit sequence and it is decoded to produce a sequence of Unicode scalar values. A sequence of translation character set elements is then formed by mapping each Unicode scalar value to the corresponding translation character set element. In the resulting sequence, each pair of characters in the input sequence consisting of carriage return (U+000D) followed by line feed (U+000A), as well as each carriage return (U+000D) not immediately followed by a line feed (U+000A), is replaced by a single new-line character.
  • For any other kind of input file supported by the implementation, characters are mapped (in implementation-defined manner) to a sequence of translation character set elements. In particular, OS-dependent end-of-line indicators are replaced by new-line characters.
(since C++23)

[edit] Phase 2

1) If the first translation character is byte order mark (U+FEFF), it is deleted. (since C++23)Whenever backslash (\) appears at the end of a line (immediately followed by zero or more whitespace characters other than new-line followed by(since C++23) the newline character), these characters are deleted, combining two physical source lines into one logical source line. This is a single-pass operation; a line ending in two backslashes followed by an empty line does not combine three lines into one.
2) If a non-empty source file does not end with a newline character after this step (end-of-line backslashes are no longer splices at this point), a terminating newline character is added.

[edit] Phase 3

1) The source file is decomposed into comments, sequences of whitespace characters (space, horizontal tab, new-line, vertical tab, and form-feed), and preprocessing tokens, which are the following:
a) header names such as <iostream> or "myfile.h"
b) placeholder tokens produced by preprocessing import and module directives (i.e. import XXX; and module XXX;)
(since C++20)
d) preprocessing numbers
e) character literals, including user-defined character literals(since C++11)
f) string literals, including user-defined string literals(since C++11)
g) operators and punctuators (including alternative tokens), such as +, <<=, <%, ##, or and
h) individual non-whitespace characters that do not fit in any other category
The program is ill-formed if the character matching this category is
  • apostrophe (', U+0027),
  • quotation mark (", U+0022), or
  • a character not in the basic character set.
2) Any transformations performed during phase 1 and(until C++23) phase 2 between the initial and the final double quote of any raw string literal are reverted.
(since C++11)
3) Each comment is replaced by one space character.

Newlines are kept, and it is unspecified whether non-newline whitespace sequences may be collapsed into single space characters.

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), universal character names are recognized and replaced by the designated element of the translation character set, except when matching a character sequence in:

a) a character literal (c-char-sequence)
b) a string literal (s-char-sequence and r-char-sequence), excluding delimiters (d-char-sequence)
c) a file name for inclusion (h-char-sequence and q-char-sequence)
(since C++23)

If the input has been parsed into preprocessing tokens up to a given character, the next preprocessing token is generally taken to be the longest sequence of characters that could constitute a preprocessing token, even if that would cause subsequent analysis to fail. This is commonly known as maximal munch.

int foo = 1;
int bar = 0xE+foo;   // error, invalid preprocessing number 0xE+foo
int baz = 0xE + foo; // OK
int quux = bar+++++baz; // error: bar++ ++ +baz, not bar++ + ++baz.

The sole exceptions to the maximal munch rule are:

  • If the next character begins a sequence of characters that could be the prefix and initial double quote of a raw string literal, the next preprocessing token shall be a raw string literal. The literal consists of the shortest sequence of characters that matches the raw-string pattern.
#define R "x"
const char* s = R"y"; // ill-formed raw string literal, not "x" "y"
const char* s2 = R"(a)" "b)"; // a raw string literal followed by a normal string literal
  • If the next three characters are <:: and the subsequent character is neither : nor >, the < is treated as a preprocessing token by itself (and not as the first character of the alternative token <:).
struct Foo { static const int v = 1; };
std::vector<::Foo> x; // OK, <: not taken as the alternative token for [
extern int y<::>;     // OK, same as extern int y[].
int z<:::Foo::value:>; // OK, int z[::Foo::value];
(since C++11)
  • Header name preprocessing tokens are only formed within a #include or import(since C++20) directive or in a __has_include expression(since C++17).
std::vector<int> x; // OK, <int> not a header-name

[edit] Phase 4

1) The preprocessor is executed.
2) Each file introduced with the #include directive goes through phases 1 through 4, recursively.
3) At the end of this phase, all preprocessor directives are removed from the source.

[edit] Phase 5

1) All characters in character literals and string literals are converted from the source character set to the encoding (which may be a multibyte character encoding such as UTF-8, as long as the 96 characters of the basic character set have single-byte representations).
2) Escape sequences and universal character names in character literals and non-raw string literals are expanded and converted to the literal encoding. If the character specified by a universal character name cannot be encoded as a single code point in the corresponding literal encoding, the result is implementation-defined, but is guaranteed not to be a null (wide) character.

Note: the conversion performed at this stage can be controlled by command line options in some implementations: gcc and clang use -finput-charset to specify the encoding of the source character set, -fexec-charset and -fwide-exec-charset to specify the ordinary and wide literal encodings respectively, while Visual Studio 2015 Update 2 and later uses /source-charset and /execution-charset to specify the source character set and literal encoding respectively.

(until C++23)

For a sequence of two or more adjacent string literal tokens, a common encoding prefix is determined as described here. Each such string literal token is then considered to have that common encoding prefix. (Character conversion is moved to phase 3)

(since C++23)

[edit] Phase 6

Adjacent string literals are concatenated.

[edit] Phase 7

Compilation takes place: each preprocessing token is converted to a token. The tokens are syntactically and semantically analyzed and translated as a translation unit.

[edit] Phase 8

Each translation unit is examined to produce a list of required template instantiations, including the ones requested by explicit instantiations. The definitions of the templates are located, and the required instantiations are performed to produce instantiation units.

[edit] Phase 9

Translation units, instantiation units, and library components needed to satisfy external references are collected into a program image which contains information needed for execution in its execution environment.

[edit] Notes

Some compilers do not implement instantiation units (also known as template repositories or template registries) and simply compile each template instantiation at phase 7, storing the code in the object file where it is implicitly or explicitly requested, and then the linker collapses these compiled instantiations into one at phase 9.

[edit] Defect reports

The following behavior-changing defect reports were applied retroactively to previously published C++ standards.

DR Applied to Behavior as published Correct behavior
CWG 787 C++98 the behavior was undefined if a non-empty source file does
not end with a newline character at the end of phase 2
add a terminating newline
character in this case
CWG 1775 C++11 forming a universal character name inside a raw
string literal in phase 2 resulted in undefined behavior
made well-defined
CWG 2747 C++98 phase 2 checked the end-of-file splice after splicing, this is unnecessary removed the check
P2621R2 C++98 universal character names were not allowed to
be formed by line splicing or token concatenation

[edit] References

  • C++23 standard (ISO/IEC 14882:2023):
  • 5.2 Phases of translation [lex.phases]
  • C++20 standard (ISO/IEC 14882:2020):
  • 5.2 Phases of translation [lex.phases]
  • C++17 standard (ISO/IEC 14882:2017):
  • 5.2 Phases of translation [lex.phases]
  • C++14 standard (ISO/IEC 14882:2014):
  • 2.2 Phases of translation [lex.phases]
  • C++11 standard (ISO/IEC 14882:2011):
  • 2.2 Phases of translation [lex.phases]
  • C++03 standard (ISO/IEC 14882:2003):
  • 2.1 Phases of translation [lex.phases]
  • C++98 standard (ISO/IEC 14882:1998):
  • 2.1 Phases of translation [lex.phases]

[edit] See also

C documentation for Phases of translation