接上一篇C语言中可变参数函数实现原理,从理论上详细介绍了C语言中可变参数函数的实现,这一篇从minix内核源码中的scanf函数入手,学习C语言经典可变参数函数的实现过程
在scanf.c文件中,可以看到scanf函数,代码如下:
#include <stdio.h> #include <stdarg.h> #include "loc_incl.h" int scanf(const char *format, ...) { va_list ap; int retval; va_start(ap, format); retval = _doscan(stdin, format, ap); va_end(ap); return retval; }
对于va_list、va_start、va_end等在stdarg.h头文件中定义的宏,都已经在(stdarg.h头文件源代码分析)一文中介绍过。
在上述代码中我们可以看到有一个_doscan函数,而这一函数在头文件loc_incl.h中定义,函数声明如下:
int _doscan(FILE * stream, const char *format, va_list ap);
_doscan函数的实现源代码如下:
int _doscan(register FILE *stream, const char *format, va_list ap) { int done = 0; /* number of items done */ int nrchars = 0; /* number of characters read */ int conv = 0; /* # of conversions */ int base; /* conversion base */ unsigned long val; /* an integer value */ register char *str; /* temporary pointer */ char *tmp_string; /* ditto */ unsigned width = 0; /* width of field */ int flags; /* some flags */ int reverse; /* reverse the checking in [...] */ int kind; register int ic = EOF; /* the input character */ #ifndef NOFLOAT long double ld_val; #endif if (!*format) return 0; while (1) { if (isspace(*format)) { while (isspace(*format)) format++; /* skip whitespace */ ic = getc(stream); nrchars++; while (isspace (ic)) { ic = getc(stream); nrchars++; } if (ic != EOF) ungetc(ic,stream); nrchars--; } if (!*format) break; /* end of format */ if (*format != '%') { ic = getc(stream); nrchars++; if (ic != *format++) break; /* error */ continue; } format++; if (*format == '%') { ic = getc(stream); nrchars++; if (ic == '%') { format++; continue; } else break; } flags = 0; if (*format == '*') { format++; flags |= FL_NOASSIGN; } if (isdigit (*format)) { flags |= FL_WIDTHSPEC; for (width = 0; isdigit (*format);) width = width * 10 + *format++ - '0'; } switch (*format) { case 'h': flags |= FL_SHORT; format++; break; case 'l': flags |= FL_LONG; format++; break; case 'L': flags |= FL_LONGDOUBLE; format++; break; } kind = *format; if ((kind != 'c') && (kind != '[') && (kind != 'n')) { do { ic = getc(stream); nrchars++; } while (isspace(ic)); if (ic == EOF) break; /* outer while */ } else if (kind != 'n') { /* %c or %[ */ ic = getc(stream); if (ic == EOF) break; /* outer while */ nrchars++; } switch (kind) { default: /* not recognized, like %q */ return conv || (ic != EOF) ? done : EOF; break; case 'n': if (!(flags & FL_NOASSIGN)) { /* silly, though */ if (flags & FL_SHORT) *va_arg(ap, short *) = (short) nrchars; else if (flags & FL_LONG) *va_arg(ap, long *) = (long) nrchars; else *va_arg(ap, int *) = (int) nrchars; } break; case 'p': /* pointer */ set_pointer(flags); /* fallthrough */ case 'b': /* binary */ case 'd': /* decimal */ case 'i': /* general integer */ case 'o': /* octal */ case 'u': /* unsigned */ case 'x': /* hexadecimal */ case 'X': /* ditto */ if (!(flags & FL_WIDTHSPEC) || width > NUMLEN) width = NUMLEN; if (!width) return done; str = o_collect(ic, stream, kind, width, &base); if (str < inp_buf || (str == inp_buf && (*str == '-' || *str == '+'))) return done; /* * Although the length of the number is str-inp_buf+1 * we don't add the 1 since we counted it already */ nrchars += str - inp_buf; if (!(flags & FL_NOASSIGN)) { if (kind == 'd' || kind == 'i') val = strtol(inp_buf, &tmp_string, base); else val = strtoul(inp_buf, &tmp_string, base); if (flags & FL_LONG) *va_arg(ap, unsigned long *) = (unsigned long) val; else if (flags & FL_SHORT) *va_arg(ap, unsigned short *) = (unsigned short) val; else *va_arg(ap, unsigned *) = (unsigned) val; } break; case 'c': if (!(flags & FL_WIDTHSPEC)) width = 1; if (!(flags & FL_NOASSIGN)) str = va_arg(ap, char *); if (!width) return done; while (width && ic != EOF) { if (!(flags & FL_NOASSIGN)) *str++ = (char) ic; if (--width) { ic = getc(stream); nrchars++; } } if (width) { if (ic != EOF) ungetc(ic,stream); nrchars--; } break; case 's': if (!(flags & FL_WIDTHSPEC)) width = 0xffff; if (!(flags & FL_NOASSIGN)) str = va_arg(ap, char *); if (!width) return done; while (width && ic != EOF && !isspace(ic)) { if (!(flags & FL_NOASSIGN)) *str++ = (char) ic; if (--width) { ic = getc(stream); nrchars++; } } /* terminate the string */ if (!(flags & FL_NOASSIGN)) *str = '\0'; if (width) { if (ic != EOF) ungetc(ic,stream); nrchars--; } break; case '[': if (!(flags & FL_WIDTHSPEC)) width = 0xffff; if (!width) return done; if ( *++format == '^' ) { reverse = 1; format++; } else reverse = 0; for (str = Xtable; str < &Xtable[NR_CHARS] ; str++) *str = 0; if (*format == ']') Xtable[*format++] = 1; while (*format && *format != ']') { Xtable[*format++] = 1; if (*format == '-') { format++; if (*format && *format != ']' && *(format) >= *(format -2)) { int c; for( c = *(format -2) + 1 ; c <= *format ; c++) Xtable[c] = 1; format++; } else Xtable['-'] = 1; } } if (!*format) return done; if (!(Xtable[ic] ^ reverse)) { /* MAT 8/9/96 no match must return character */ ungetc(ic, stream); return done; } if (!(flags & FL_NOASSIGN)) str = va_arg(ap, char *); do { if (!(flags & FL_NOASSIGN)) *str++ = (char) ic; if (--width) { ic = getc(stream); nrchars++; } } while (width && ic != EOF && (Xtable[ic] ^ reverse)); if (width) { if (ic != EOF) ungetc(ic, stream); nrchars--; } if (!(flags & FL_NOASSIGN)) { /* terminate string */ *str = '\0'; } break; #ifndef NOFLOAT case 'e': case 'E': case 'f': case 'g': case 'G': if (!(flags & FL_WIDTHSPEC) || width > NUMLEN) width = NUMLEN; if (!width) return done; str = f_collect(ic, stream, width); if (str < inp_buf || (str == inp_buf && (*str == '-' || *str == '+'))) return done; /* * Although the length of the number is str-inp_buf+1 * we don't add the 1 since we counted it already */ nrchars += str - inp_buf; if (!(flags & FL_NOASSIGN)) { ld_val = strtod(inp_buf, &tmp_string); if (flags & FL_LONGDOUBLE) *va_arg(ap, long double *) = (long double) ld_val; else if (flags & FL_LONG) *va_arg(ap, double *) = (double) ld_val; else *va_arg(ap, float *) = (float) ld_val; } break; #endif } /* end switch */ conv++; if (!(flags & FL_NOASSIGN) && kind != 'n') done++; format++; } return conv || (ic != EOF) ? done : EOF; } _doscan函数代码
在上面的源代码中,值得注意的是第26行的getc宏,定义代码如下:
#define getc(p) (--(p)->_count >= 0 ? (int) (*(p)->_ptr++) : \ __fillbuf(p))
getc的调用形式:ch=getc(fp); 功能是从文件指针指向的文件读入一个字符,并把它作为函数值返回给int型变量ch。
第4行~第17行,定义一些后面需要用到的变量
第23行~第34行,跳过format格式串中的空格,并且跳过输入流中的空格
第37行~第42行,输入流stream与format格式串中的空白符(空白符可以是空格(space)、制表符(tab)和新行符(newline))保持一致
第44行~第52行,在format中的字符为'%'的前提下,stream中的字符也为'%',则继续
第54行~第57行,format当前字符为'*',表示读指定类型的数据但不保存
第58行~第62行,指定说明最大域宽。 在百分号(%)与格式码之间的整数用于限制从对应域读入的最大字符数于宽度
第64行~第282行,switch语句,用于格式修饰符,这些修饰符包括: h、l、L、c、p、b、d、i、o、u……,还有基于扫描集的'['修饰符
-----------------------------------------------------------------------------------
对scanf函数的源码分析,需要在scanf函数的语法格式详细的理解基础上进行,由于scanf函数实现十分复杂,需要仔细的品味,这里只是比较初步的分析,具体还有待后期不断的完善