After the textual transformations are finished, the input file is converted into a sequence of preprocessing tokens. These mostly correspond to the syntactic tokens used by the C compiler, but there are a few differences. White space separates tokens; it is not itself a token of any kind. Tokens do not have to be separated by white space, but it is often necessary to avoid ambiguities.
When faced with a sequence of characters that has more than one possible
tokenization, the preprocessor is greedy. It always makes each token,
starting from the left, as big as possible before moving on to the next
token. For instance, a+++++b is interpreted as
a ++ ++ + b, not as a ++ + ++ b, even though the
latter tokenization could be part of a valid C program and the former
could not.
Once the input file is broken into tokens, the token boundaries never change, except when the ## preprocessing operator is used to paste tokens together. See Concatenation. For example,
#define foo() bar
foo()baz
==> bar baz
not
==> barbaz
The compiler does not re-tokenize the preprocessor's output. Each preprocessing token becomes one compiler token.
Preprocessing tokens fall into five broad classes: identifiers,
preprocessing numbers, string literals, punctuators, and other. An
identifier is the same as an identifier in C: any sequence of
letters, digits, or underscores, which begins with a letter or
underscore. Keywords of C have no significance to the preprocessor;
they are ordinary identifiers. You can define a macro whose name is a
keyword, for instance. The only identifier which can be considered a
preprocessing keyword is defined. See Defined.
This is mostly true of other languages which use the C preprocessor. However, a few of the keywords of C++ are significant even in the preprocessor. See C++ Named Operators.
In the 1999 C standard, identifiers may contain letters which are not part of the “basic source character set”, at the implementation's discretion (such as accented Latin letters, Greek letters, or Chinese ideograms). This may be done with an extended character set, or the \u and \U escape sequences. The implementation of this feature in GCC is experimental; such characters are only accepted in the \u and \U forms and only if -fextended-identifiers is used.
As an extension, GCC treats $ as a letter. This is for compatibility with some systems, such as VMS, where $ is commonly used in system-defined function and object names. $ is not a letter in strictly conforming mode, or if you specify the -$ option. See Invocation.
A preprocessing number has a rather bizarre definition. The category includes all the normal integer and floating point constants one expects of C, but also a number of other things one might not initially recognize as a number. Formally, preprocessing numbers begin with an optional period, a required decimal digit, and then continue with any sequence of letters, digits, underscores, periods, and exponents. Exponents are the two-character sequences e+, e-, E+, E-, p+, p-, P+, and P-. (The exponents that begin with p or P are new to C99. They are used for hexadecimal floating-point constants.)
The purpose of this unusual definition is to isolate the preprocessor from the full complexity of numeric constants. It does not have to distinguish between lexically valid and invalid floating-point numbers, which is complicated. The definition also permits you to split an identifier at any position and get exactly two tokens, which can then be pasted back together with the ## operator.
It's possible for preprocessing numbers to cause programs to be
misinterpreted. For example, 0xE+12 is a preprocessing number
which does not translate to any valid numeric constant, therefore a
syntax error. It does not mean 0xE + 12, which is what you
might have intended.
String literals are string constants, character constants, and header file names (the argument of #include).1 String constants and character constants are straightforward: "..." or '...'. In either case embedded quotes should be escaped with a backslash: '\'' is the character constant for '. There is no limit on the length of a character constant, but the value of a character constant that contains more than one character is implementation-defined. See Implementation Details.
Header file names either look like string constants, "...", or are written with angle brackets instead, <...>. In either case, backslash is an ordinary character. There is no way to escape the closing quote or angle bracket. The preprocessor looks for the header file in different places depending on which form you use. See Include Operation.
No string literal may extend past the end of a line. Older versions of GCC accepted multi-line string constants. You may use continued lines instead, or string constant concatenation. See Differences from previous versions.
Punctuators are all the usual bits of punctuation which are meaningful to C and C++. All but three of the punctuation characters in ASCII are C punctuators. The exceptions are @, $, and `. In addition, all the two- and three-character operators are punctuators. There are also six digraphs, which the C++ standard calls alternative tokens, which are merely alternate ways to spell other punctuators. This is a second attempt to work around missing punctuation in obsolete systems. It has no negative side effects, unlike trigraphs, but does not cover as much ground. The digraphs and their corresponding normal punctuators are:
Digraph: <% %> <: :> %: %:%:
Punctuator: { } [ ] # ##
Any other single character is considered “other”. It is passed on to the preprocessor's output unmolested. The C compiler will almost certainly reject source code containing “other” tokens. In ASCII, the only other characters are @, $, `, and control characters other than NUL (all bits zero). (Note that $ is normally considered a letter.) All characters with the high bit set (numeric range 0x7F–0xFF) are also “other” in the present implementation. This will change when proper support for international character sets is added to GCC.
NUL is a special case because of the high probability that its appearance is accidental, and because it may be invisible to the user (many terminals do not display NUL at all). Within comments, NULs are silently ignored, just as any other character would be. In running text, NUL is considered white space. For example, these two directives have the same meaning.
#define X^@1
#define X 1
(where ^@ is ASCII NUL). Within string or character constants, NULs are preserved. In the latter two cases the preprocessor emits a warning message.
[1] The C standard uses the term string literal to refer only to what we are calling string constants.