C Compilation Model

Your Task

gcc turns C into an executable in four stages: preprocess (-E), compile to assembly (-S), assemble to an object file (-c), and link. The first stage is pure text rewriting, and most "why does my macro do that?" bugs live there. Implement the core of it: a mini C preprocessor that reads a source file on stdin and prints what gcc -E -P would print.

Supported directives

A directive is a line whose first non-blank character is # (spaces are allowed after it).

• #define NAME body: object-like macro.
• #define NAME(a, b) body: function-like, only if ( directly touches the name. #define F (x) x is object-like with body (x) x. Up to 8 parameters, and NAME() takes none. Redefining a macro replaces it.
• #undef NAME.
• #ifdef NAME, #ifndef NAME, #else, #endif (nestable). Lines in a false branch are dropped, including #defines.
• Any other directive (such as #include <stdio.h>) is printed unchanged, since there are no headers here.

Expansion rules

1. Comments are replaced by a space: // to the end of the line, and /* */. A block comment that spans lines joins them into one line, as in clang's output. String and character literals are never touched, and neither are the insides of numbers like 0x1F or 1e10.
2. An identifier that names a macro is replaced by its body. A function-like macro is replaced only when its name is followed by (: collect the arguments up to the matching ) on the same line, split at top-level commas (parentheses nest, and commas inside literals don't count), and trim them. Fully expand each argument first, then substitute it for its parameter.
3. The result is rescanned for more macros, but a macro is never expanded inside its own expansion. That is why #define LOOP LOOP + 1 stops after one step.

Output

• Every non-directive line, after expansion, if it is not blank. Leading indentation is kept, any other run of spaces or tabs (outside literals) becomes one space, and trailing whitespace is dropped.
• Errors, printed where they occur (the unexpanded text is kept):
  • line N: error: MAC expects K argument(s), got M;
  • line N: error: unterminated argument list for MAC;
  • line N: error: #else without #if, #else after #else, #endif without #if;
  • line N: error: macro name missing;
  • line N: error: bad parameter list for MAC;
  • line N: error: #ifdef needs a macro name (or #ifndef).
• At the end: error: unterminated comment at end of input if a block comment never closes, then error: K unterminated conditional(s) at end of input if any are still open, then always -- N macro(s) defined, M expansion(s). N counts the macros still defined, and M counts every replacement, nested ones included.

Example

Input:

cCopy
Loading…

Output:

cCopy
Loading…

Requirements

1. Expected outputs match clang -E -P on the same input (apart from #include, which is passed through).
2. Track the "disabled" macros per expansion, not globally. MAX(MAX(1, 2), 3) must expand both calls.

Hidden tests cover self-referential and mutually recursive macros, macros passed as arguments, commas inside parentheses and strings, numbers that look like identifiers, #undef, empty-parameter macros, nested conditionals, a comment spanning lines, and every error.