* Copyright (c) 1990 The Regents of the University of California.
* Redistribution and use in source and binary forms are permitted
* provided that the above copyright notice and this paragraph are
* duplicated in all such forms and that any documentation,
* advertising materials, and other materials related to such
* distribution and use acknowledge that the software was developed
* by the University of California, Berkeley. The name of the
* University may not be used to endorse or promote products derived
* from this software without specific prior written permission.
* THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
// Extensively hacked for GNU iostream by Per Bothner 1991, 1992.
// Changes copyright Per Bothner 1992.
#if defined(LIBC_SCCS) && !defined(lint)
static char sccsid[] = "%W% (Berkeley) %G%";
#endif /* LIBC_SCCS and not lint */
#ifndef NO_FLOATING_POINT
#define BUF (MAXEXP+MAXFRACT+3) /* 3 = sign + decimal point + NUL */
* Flags used during conversion.
#define LONG 0x01 /* l: long or double */
#define LONGDBL 0x02 /* L: long double; unimplemented */
#define SHORT 0x04 /* h: short */
#define SUPPRESS 0x08 /* suppress assignment */
#define POINTER 0x10 /* weird %p pointer (`fake hex') */
#define NOSKIP 0x20 /* do not skip blanks */
* The following are used in numeric conversions only:
* SIGNOK, NDIGITS, DPTOK, and EXPOK are for floating point;
* SIGNOK, NDIGITS, PFXOK, and NZDIGITS are for integral.
#define SIGNOK 0x40 /* +/- is (still) legal */
#define NDIGITS 0x80 /* no digits detected */
#define DPTOK 0x100 /* (float) decimal point is still legal */
#define EXPOK 0x200 /* (float) exponent (e+3, etc) still legal */
#define PFXOK 0x100 /* 0x prefix is (still) legal */
#define NZDIGITS 0x200 /* no zero digits detected */
#define CT_CHAR 0 /* %c conversion */
#define CT_CCL 1 /* %[...] conversion */
#define CT_STRING 2 /* %s conversion */
#define CT_INT 3 /* integer, i.e., strtol or strtoul */
#define CT_FLOAT 4 /* floating, i.e., strtod */
#define u_char unsigned char
#define u_long unsigned long
extern "C" u_long strtoul(const char*, char**, int);
static const u_char *__sccl(register char *tab, register const u_char *fmt);
// If state is non-NULL, set failbit and/or eofbit as appropriate.
int streambuf::vscan(char const *fmt0,
ios *stream /* = NULL */)
register const u_char *fmt = (const u_char *)fmt0;
register int c; /* character from format, or conversion */
register size_t width; /* field width, or 0 */
register char *p; /* points into all kinds of strings */
register int n; /* handy integer */
register int flags; /* flags as defined above */
register char *p0; /* saves original value of p when necessary */
int nassigned; /* number of fields assigned */
int nread; /* number of characters consumed from fp */
// Assignments to base and ccfn are just to suppress warnings from gcc.
int base = 0; /* base argument to strtol/strtoul */
typedef u_long (*strtoulfn)(const char*, char**, int);
// conversion function (strtol/strtoul)
char ccltab[256]; /* character class table for %[...] */
char buf[BUF]; /* buffer for numeric conversions */
/* `basefix' is used to avoid `if' tests in the integer scanner */
static short basefix[17] =
{ 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 };
* switch on the format. continue if done;
* break once format type is derived.
case '0': case '1': case '2': case '3': case '4':
case '5': case '6': case '7': case '8': case '9':
width = width * 10 + c - '0';
* Those marked `compat' are for 4.[123]BSD compatibility.
* (According to ANSI, E and X formats are supposed
* to the same as e and x. Sorry about that.)
ccfn = (strtoulfn)strtol;
ccfn = (strtoulfn)strtol;
flags |= PFXOK; /* enable 0x prefixing */
case 'e': case 'f': case 'g':
fmt = __sccl(ccltab, fmt);
case 'p': /* pointer format is like hex */
flags |= POINTER | PFXOK;
if (flags & SUPPRESS) /* ??? */
*va_arg(ap, short *) = nread;
*va_arg(ap, long *) = nread;
*va_arg(ap, int *) = nread;
* Disgusting backwards compatibility hacks. XXX
ccfn = (strtoulfn)strtol;
* We have a conversion that requires input.
* Consume leading white space, except for formats
if ((flags & NOSKIP) == 0) {
n = (unsigned char)*_gptr;
// Note that there is at least one character in
// the buffer, so conversions that do not set NOSKIP
// can no longer result in an input failure.
/* scan arbitrary characters (sets NOSKIP) */
if (width == 0) // FIXME!
if ((n = _egptr - _gptr) < (int)width) {
size_t r = sgetn((char*)va_arg(ap, char*),
/* scan a (nonempty) character class (sets NOSKIP) */
width = ~0; /* `infinity' */
/* take only those things in the class */
while (ccltab[(unsigned char)*_gptr]) {
p0 = p = va_arg(ap, char *);
while (ccltab[(unsigned char)*_gptr]) {
/* like CCL, but zero-length string OK, & no NOSKIP */
while (!isspace((unsigned char)*_gptr)) {
p0 = p = va_arg(ap, char *);
while (!isspace((unsigned char)*_gptr)) {
/* scan an integer as if by strtol/strtoul */
if (width == 0 || width > sizeof(buf) - 1)
flags |= SIGNOK | NDIGITS | NZDIGITS;
for (p = buf; width; width--) {
c = (unsigned char)*_gptr;
* Switch on the character; `goto ok'
* if we accept it as a part of number.
* The digit 0 is always legal, but is
* special. For %i conversions, if no
* digits (zero or nonzero) have been
* scanned (only signs), we will have
* base==0. In that case, we should set
* it to 8 and enable 0x prefixing.
* Also, if we have not scanned zero digits
* before this, do not turn off prefixing
* (someone else will turn it off if we
* have scanned any nonzero digits).
flags &= ~(SIGNOK|NZDIGITS|NDIGITS);
flags &= ~(SIGNOK|PFXOK|NDIGITS);
/* 1 through 7 always legal */
case '1': case '2': case '3':
case '4': case '5': case '6': case '7':
flags &= ~(SIGNOK | PFXOK | NDIGITS);
/* digits 8 and 9 ok iff decimal or hex */
break; /* not legal here */
flags &= ~(SIGNOK | PFXOK | NDIGITS);
case 'A': case 'B': case 'C':
case 'D': case 'E': case 'F':
case 'a': case 'b': case 'c':
case 'd': case 'e': case 'f':
/* no need to fix base here */
break; /* not legal here */
flags &= ~(SIGNOK | PFXOK | NDIGITS);
/* sign ok only as first character */
/* x ok iff flag still set & 2nd char */
if (flags & PFXOK && p == buf + 1) {
* If we got here, c is not a legal character
* for a number. Stop accumulating digits.
* c is legal: store it and look at the next.
* If we had only a sign, it is no good; push
* back the sign. If the number ends in `x',
* it was [sign] '0' 'x', so push back the x
* and treat it as [sign] '0'.
(void) sputbackc(*(u_char *)--p);
if (c == 'x' || c == 'X') {
if ((flags & SUPPRESS) == 0) {
res = (*ccfn)(buf, (char **)NULL, base);
*va_arg(ap, void **) = (void *)res;
*va_arg(ap, short *) = res;
*va_arg(ap, long *) = res;
*va_arg(ap, int *) = res;
/* scan a floating point number as if by strtod */
if (width == 0 || width > sizeof(buf) - 1)
flags |= SIGNOK | NDIGITS | DPTOK | EXPOK;
for (p = buf; width; width--) {
c = (unsigned char)*_gptr;
* This code mimicks the integer conversion
* code, but is much simpler.
case '0': case '1': case '2': case '3':
case '4': case '5': case '6': case '7':
flags &= ~(SIGNOK | NDIGITS);
flags &= ~(SIGNOK | DPTOK);
/* no exponent without some digits */
if ((flags&(NDIGITS|EXPOK)) == EXPOK) {
(flags & ~(EXPOK|DPTOK)) |
* If no digits, might be missing exponent digits
* (just give back the exponent) or might be missing
* regular digits, but had sign and/or decimal point.
sputbackc(*(u_char *)--p);
/* just a bad exponent (e and maybe sign) */
if (c != 'e' && c != 'E') {
(void)sputbackc(c);/* sign */
if ((flags & SUPPRESS) == 0) {
*va_arg(ap, double *) = res;
*va_arg(ap, float *) = res;
#endif /* FLOATING_POINT */
stream->set(ios::failbit);
stream->set(ios::eofbit);
* Fill in the given table from the scanset at the given format
* (just after `['). Return a pointer to the character past the
* closing `]'. The table has a 1 wherever characters should be
* considered part of the scanset.
static const u_char *__sccl(register char *tab, register const u_char *fmt)
/* first `clear' the whole table */
c = *fmt++; /* first char hat => negated scanset */
v = 1; /* default => accept */
c = *fmt++; /* get new first char */
v = 0; /* default => reject */
/* should probably use memset here */
for (n = 0; n < 256; n++)
return (fmt - 1);/* format ended before closing ] */
* Now set the entries corresponding to the actual scanset
* to the opposite of the above.
* The first character may be ']' (or '-') without being special;
* the last character may be '-'.
tab[c] = v; /* take character c */
n = *fmt++; /* and examine the next */
case 0: /* format ended too soon */
* is defined as `the digit 0, the digit 1,
* the character +, the character -', but
* the effect of a scanset such as
* is implementation defined. The V7 Unix
* scanf treats `a-z' as `the letters a through
* z', but treats `a-a' as `the letter a, the
* character -, and the letter a'.
* For compatibility, the `-' is not considerd
* to define a range if the character following
* it is either a close bracket (required by ANSI)
* or is not numerically greater than the character
* we just stored in the table (c).
break; /* resume the for(;;) */
do { /* fill in the range */
#if 1 /* XXX another disgusting compatibility hack */
* Alas, the V7 Unix scanf also treats formats
* such as [a-c-e] as `the letters a through e'.
* This too is permitted by the standard....
case ']': /* end of scanset */
default: /* just another character */
int streambuf::scan(char const *format ...)
int count = vscan(format, ap);