printf.c

sqlite 嵌入式数据库的源码

/*
** The "printf" code that follows dates from the 1980's.  It is in
** the public domain.  The original comments are included here for
** completeness.  They are very out-of-date but might be useful as
** an historical reference.  Most of the "enhancements" have been backed
** out so that the functionality is now the same as standard printf().
**
**************************************************************************
**
** The following modules is an enhanced replacement for the "printf" subroutines
** found in the standard C library.  The following enhancements are
** supported:
**
**      +  Additional functions.  The standard set of "printf" functions
**         includes printf, fprintf, sprintf, vprintf, vfprintf, and
**         vsprintf.  This module adds the following:
**
**           *  snprintf -- Works like sprintf, but has an extra argument
**                          which is the size of the buffer written to.
**
**           *  mprintf --  Similar to sprintf.  Writes output to memory
**                          obtained from malloc.
**
**           *  xprintf --  Calls a function to dispose of output.
**
**           *  nprintf --  No output, but returns the number of characters
**                          that would have been output by printf.
**
**           *  A v- version (ex: vsnprintf) of every function is also
**              supplied.
**
**      +  A few extensions to the formatting notation are supported:
**
**           *  The "=" flag (similar to "-") causes the output to be
**              be centered in the appropriately sized field.
**
**           *  The %b field outputs an integer in binary notation.
**
**           *  The %c field now accepts a precision.  The character output
**              is repeated by the number of times the precision specifies.
**
**           *  The %' field works like %c, but takes as its character the
**              next character of the format string, instead of the next
**              argument.  For example,  printf("%.78'-")  prints 78 minus
**              signs, the same as  printf("%.78c",'-').
**
**      +  When compiled using GCC on a SPARC, this version of printf is
**         faster than the library printf for SUN OS 4.1.
**
**      +  All functions are fully reentrant.
**
*/
#include "sqliteInt.h"

/*
** Conversion types fall into various categories as defined by the
** following enumeration.
*/
#define etRADIX       1 /* Integer types.  %d, %x, %o, and so forth */
#define etFLOAT       2 /* Floating point.  %f */
#define etEXP         3 /* Exponentional notation. %e and %E */
#define etGENERIC     4 /* Floating or exponential, depending on exponent. %g */
#define etSIZE        5 /* Return number of characters processed so far. %n */
#define etSTRING      6 /* Strings. %s */
#define etDYNSTRING   7 /* Dynamically allocated strings. %z */
#define etPERCENT     8 /* Percent symbol. %% */
#define etCHARX       9 /* Characters. %c */
#define etERROR      10 /* Used to indicate no such conversion type */
/* The rest are extensions, not normally found in printf() */
#define etCHARLIT    11 /* Literal characters.  %' */
#define etSQLESCAPE  12 /* Strings with '\'' doubled.  %q */
#define etSQLESCAPE2 13 /* Strings with '\'' doubled and enclosed in '',
                          NULL pointers replaced by SQL NULL.  %Q */
#define etTOKEN      14 /* a pointer to a Token structure */
#define etSRCLIST    15 /* a pointer to a SrcList */
#define etPOINTER    16 /* The %p conversion */


/*
** An "etByte" is an 8-bit unsigned value.
*/
typedef unsigned char etByte;

/*
** Each builtin conversion character (ex: the 'd' in "%d") is described
** by an instance of the following structure
*/
typedef struct et_info {   /* Information about each format field */
  char fmttype;            /* The format field code letter */
  etByte base;             /* The base for radix conversion */
  etByte flags;            /* One or more of FLAG_ constants below */
  etByte type;             /* Conversion paradigm */
  etByte charset;          /* Offset into aDigits[] of the digits string */
  etByte prefix;           /* Offset into aPrefix[] of the prefix string */
} et_info;

/*
** Allowed values for et_info.flags
*/
#define FLAG_SIGNED  1     /* True if the value to convert is signed */
#define FLAG_INTERN  2     /* True if for internal use only */
#define FLAG_STRING  4     /* Allow infinity precision */


/*
** The following table is searched linearly, so it is good to put the
** most frequently used conversion types first.
*/
static const char aDigits[] = "0123456789ABCDEF0123456789abcdef";
static const char aPrefix[] = "-x0\000X0";
static const et_info fmtinfo[] = {
  {  'd', 10, 1, etRADIX,      0,  0 },
  {  's',  0, 4, etSTRING,     0,  0 },
  {  'z',  0, 6, etDYNSTRING,  0,  0 },
  {  'q',  0, 4, etSQLESCAPE,  0,  0 },
  {  'Q',  0, 4, etSQLESCAPE2, 0,  0 },
  {  'c',  0, 0, etCHARX,      0,  0 },
  {  'o',  8, 0, etRADIX,      0,  2 },
  {  'u', 10, 0, etRADIX,      0,  0 },
  {  'x', 16, 0, etRADIX,      16, 1 },
  {  'X', 16, 0, etRADIX,      0,  4 },
  {  'f',  0, 1, etFLOAT,      0,  0 },
  {  'e',  0, 1, etEXP,        30, 0 },
  {  'E',  0, 1, etEXP,        14, 0 },
  {  'g',  0, 1, etGENERIC,    30, 0 },
  {  'G',  0, 1, etGENERIC,    14, 0 },
  {  'i', 10, 1, etRADIX,      0,  0 },
  {  'n',  0, 0, etSIZE,       0,  0 },
  {  '%',  0, 0, etPERCENT,    0,  0 },
  {  'p', 16, 0, etPOINTER,    0,  1 },
  {  'T',  0, 2, etTOKEN,      0,  0 },
  {  'S',  0, 2, etSRCLIST,    0,  0 },
};
#define etNINFO  (sizeof(fmtinfo)/sizeof(fmtinfo[0]))

/*
** If NOFLOATINGPOINT is defined, then none of the floating point
** conversions will work.
*/
#ifndef etNOFLOATINGPOINT
/*
** "*val" is a double such that 0.1 <= *val < 10.0
** Return the ascii code for the leading digit of *val, then
** multiply "*val" by 10.0 to renormalize.
**
** Example:
**     input:     *val = 3.14159
**     output:    *val = 1.4159    function return = '3'
**
** The counter *cnt is incremented each time.  After counter exceeds
** 16 (the number of significant digits in a 64-bit float) '0' is
** always returned.
*/
static int et_getdigit(LONGDOUBLE_TYPE *val, int *cnt){
  int digit;
  LONGDOUBLE_TYPE d;
  if( (*cnt)++ >= 16 ) return '0';
  digit = (int)*val;
  d = digit;
  digit += '0';
  *val = (*val - d)*10.0;
  return digit;
}
#endif

#define etBUFSIZE 1000  /* Size of the output buffer */

/*
** The root program.  All variations call this core.
**
** INPUTS:
**   func   This is a pointer to a function taking three arguments
**            1. A pointer to anything.  Same as the "arg" parameter.
**            2. A pointer to the list of characters to be output
**               (Note, this list is NOT null terminated.)
**            3. An integer number of characters to be output.
**               (Note: This number might be zero.)
**
**   arg    This is the pointer to anything which will be passed as the
**          first argument to "func".  Use it for whatever you like.
**
**   fmt    This is the format string, as in the usual print.
**
**   ap     This is a pointer to a list of arguments.  Same as in
**          vfprint.
**
** OUTPUTS:
**          The return value is the total number of characters sent to
**          the function "func".  Returns -1 on a error.
**
** Note that the order in which automatic variables are declared below
** seems to make a big difference in determining how fast this beast
** will run.
*/
static int vxprintf(
  void (*func)(void*,const char*,int),     /* Consumer of text */
  void *arg,                         /* First argument to the consumer */
  int useExtended,                   /* Allow extended %-conversions */
  const char *fmt,                   /* Format string */
  va_list ap                         /* arguments */
){
  int c;                     /* Next character in the format string */
  char *bufpt;               /* Pointer to the conversion buffer */
  int precision;             /* Precision of the current field */
  int length;                /* Length of the field */
  int idx;                   /* A general purpose loop counter */
  int count;                 /* Total number of characters output */
  int width;                 /* Width of the current field */
  etByte flag_leftjustify;   /* True if "-" flag is present */
  etByte flag_plussign;      /* True if "+" flag is present */
  etByte flag_blanksign;     /* True if " " flag is present */
  etByte flag_alternateform; /* True if "#" flag is present */
  etByte flag_zeropad;       /* True if field width constant starts with zero */
  etByte flag_long;          /* True if "l" flag is present */
  etByte flag_longlong;      /* True if the "ll" flag is present */
  UINT64_TYPE longvalue;     /* Value for integer types */
  LONGDOUBLE_TYPE realvalue; /* Value for real types */
  const et_info *infop;      /* Pointer to the appropriate info structure */
  char buf[etBUFSIZE];       /* Conversion buffer */
  char prefix;               /* Prefix character.  "+" or "-" or " " or '\0'. */
  etByte errorflag = 0;      /* True if an error is encountered */
  etByte xtype;              /* Conversion paradigm */
  char *zExtra;              /* Extra memory used for etTCLESCAPE conversions */
  static const char spaces[] =
   "                                                                         ";
#define etSPACESIZE (sizeof(spaces)-1)
#ifndef etNOFLOATINGPOINT
  int  exp;                  /* exponent of real numbers */
  double rounder;            /* Used for rounding floating point values */
  etByte flag_dp;            /* True if decimal point should be shown */
  etByte flag_rtz;           /* True if trailing zeros should be removed */
  etByte flag_exp;           /* True to force display of the exponent */
  int nsd;                   /* Number of significant digits returned */
#endif

  func(arg,"",0);
  count = length = 0;
  bufpt = 0;
  for(; (c=(*fmt))!=0; ++fmt){
    if( c!='%' ){
      int amt;
      bufpt = (char *)fmt;
      amt = 1;
      while( (c=(*++fmt))!='%' && c!=0 ) amt++;
      (*func)(arg,bufpt,amt);
      count += amt;
      if( c==0 ) break;
    }
    if( (c=(*++fmt))==0 ){
      errorflag = 1;
      (*func)(arg,"%",1);
      count++;
      break;
    }
    /* Find out what flags are present */
    flag_leftjustify = flag_plussign = flag_blanksign = 
     flag_alternateform = flag_zeropad = 0;
    do{
      switch( c ){
        case '-':   flag_leftjustify = 1;     c = 0;   break;
        case '+':   flag_plussign = 1;        c = 0;   break;
        case ' ':   flag_blanksign = 1;       c = 0;   break;
        case '#':   flag_alternateform = 1;   c = 0;   break;
        case '0':   flag_zeropad = 1;         c = 0;   break;
        default:                                       break;
      }
    }while( c==0 && (c=(*++fmt))!=0 );
    /* Get the field width */
    width = 0;
    if( c=='*' ){
      width = va_arg(ap,int);
      if( width<0 ){
        flag_leftjustify = 1;
        width = -width;
      }
      c = *++fmt;
    }else{
      while( c>='0' && c<='9' ){
        width = width*10 + c - '0';
        c = *++fmt;
      }
    }
    if( width > etBUFSIZE-10 ){
      width = etBUFSIZE-10;
    }
    /* Get the precision */
    if( c=='.' ){
      precision = 0;
      c = *++fmt;
      if( c=='*' ){
        precision = va_arg(ap,int);
        if( precision<0 ) precision = -precision;
        c = *++fmt;
      }else{
        while( c>='0' && c<='9' ){
          precision = precision*10 + c - '0';
          c = *++fmt;
        }
      }
    }else{
      precision = -1;
    }
    /* Get the conversion type modifier */
    if( c=='l' ){
      flag_long = 1;
      c = *++fmt;
      if( c=='l' ){
        flag_longlong = 1;
        c = *++fmt;
      }else{
        flag_longlong = 0;
      }
    }else{
      flag_long = flag_longlong = 0;
    }
    /* Fetch the info entry for the field */
    infop = 0;
    xtype = etERROR;
    for(idx=0; idx<etNINFO; idx++){
      if( c==fmtinfo[idx].fmttype ){
        infop = &fmtinfo[idx];
        if( useExtended || (infop->flags & FLAG_INTERN)==0 ){
          xtype = infop->type;
        }
        break;
      }
    }
    zExtra = 0;

    /* Limit the precision to prevent overflowing buf[] during conversion */
    if( precision>etBUFSIZE-40 && (infop->flags & FLAG_STRING)==0 ){
      precision = etBUFSIZE-40;
    }

    /*
    ** At this point, variables are initialized as follows:
    **
    **   flag_alternateform          TRUE if a '#' is present.
    **   flag_plussign               TRUE if a '+' is present.
    **   flag_leftjustify            TRUE if a '-' is present or if the
    **                               field width was negative.
    **   flag_zeropad                TRUE if the width began with 0.
    **   flag_long                   TRUE if the letter 'l' (ell) prefixed
    **                               the conversion character.
    **   flag_longlong               TRUE if the letter 'll' (ell ell) prefixed
    **                               the conversion character.
    **   flag_blanksign              TRUE if a ' ' is present.
    **   width                       The specified field width.  This is
    **                               always non-negative.  Zero is the default.
    **   precision                   The specified precision.  The default
    **                               is -1.
    **   xtype                       The class of the conversion.
    **   infop                       Pointer to the appropriate info struct.
    */
    switch( xtype ){
      case etPOINTER:
        flag_longlong = sizeof(char*)==sizeof(i64);
        flag_long = sizeof(char*)==sizeof(long int);
        /* Fall through into the next case */
      case etRADIX:
        if( infop->flags & FLAG_SIGNED ){
          i64 v;
          if( flag_longlong )   v = va_arg(ap,i64);
          else if( flag_long )  v = va_arg(ap,long int);
          else                  v = va_arg(ap,int);
          if( v<0 ){
            longvalue = -v;
            prefix = '-';
          }else{
            longvalue = v;
            if( flag_plussign )        prefix = '+';
            else if( flag_blanksign )  prefix = ' ';
            else                       prefix = 0;
          }
        }else{
          if( flag_longlong )   longvalue = va_arg(ap,u64);
          else if( flag_long )  longvalue = va_arg(ap,unsigned long int);
          else                  longvalue = va_arg(ap,unsigned int);
          prefix = 0;
        }
        if( longvalue==0 ) flag_alternateform = 0;
        if( flag_zeropad && precision<width-(prefix!=0) ){
          precision = width-(prefix!=0);
        }
        bufpt = &buf[etBUFSIZE-1];
        {
          register const char *cset;      /* Use registers for speed */
          register int base;
          cset = &aDigits[infop->charset];
          base = infop->base;
          do{                                           /* Convert to ascii */
            *(--bufpt) = cset[longvalue%base];
            longvalue = longvalue/base;
          }while( longvalue>0 );
        }
        length = &buf[etBUFSIZE-1]-bufpt;
        for(idx=precision-length; idx>0; idx--){
          *(--bufpt) = '0';                             /* Zero pad */
        }
        if( prefix ) *(--bufpt) = prefix;               /* Add sign */
        if( flag_alternateform && infop->prefix ){      /* Add "0" or "0x" */
          const char *pre;
          char x;
          pre = &aPrefix[infop->prefix];
          if( *bufpt!=pre[0] ){
            for(; (x=(*pre))!=0; pre++) *(--bufpt) = x;
          }
        }
        length = &buf[etBUFSIZE-1]-bufpt;
        break;
      case etFLOAT:
      case etEXP:
      case etGENERIC:
        realvalue = va_arg(ap,double);
#ifndef etNOFLOATINGPOINT
        if( precision<0 ) precision = 6;         /* Set default precision */