/* Copyright (C) 1994, 1995, 1997, 1998, 1999 Aladdin Enterprises.  All rights reserved.
  
  This software is provided AS-IS with no warranty, either express or
  implied.
  
  This software is distributed under license and may not be copied,
  modified or distributed except as expressly authorized under the terms
  of the license contained in the file LICENSE in this distribution.
  
  For more information about licensing, please refer to
  http://www.ghostscript.com/licensing/. For information on
  commercial licensing, go to http://www.artifex.com/licensing/ or
  contact Artifex Software, Inc., 101 Lucas Valley Road #110,
  San Rafael, CA  94903, U.S.A., +1(415)492-9861.
*/

/* $Id: iscannum.c,v 1.10 2004/09/15 19:41:01 ray Exp $ */
/* Number scanner for Ghostscript interpreter */
#include "math_.h"
#include "ghost.h"
#include "ierrors.h"
#include "scommon.h"
#include "iscannum.h"		/* defines interface */
#include "scanchar.h"
#include "store.h"

/*
 * Warning: this file has a "spaghetti" control structure.  But since this
 * code accounts for over 10% of the execution time of some PostScript
 * files, this is one of the few places we feel this is justified.
 */

/*
 * Scan a number.  If the number consumes the entire string, return 0;
 * if not, set *psp to the first character beyond the number and return 1.
 */
int
scan_number(const byte * str, const byte * end, int sign,
	    ref * pref, const byte ** psp, const bool PDFScanInvNum)
{
    const byte *sp = str;
#define GET_NEXT(cvar, sp, end_action)\
  if (sp >= end) { end_action; } else cvar = *sp++

    /*
     * Powers of 10 up to 6 can be represented accurately as
     * a single-precision float.
     */
#define NUM_POWERS_10 6
    static const float powers_10[NUM_POWERS_10 + 1] = {
	1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6
    };
    static const double neg_powers_10[NUM_POWERS_10 + 1] = {
	1e0, 1e-1, 1e-2, 1e-3, 1e-4, 1e-5, 1e-6
    };

    int ival;
    long lval;
    double dval;
    int exp10;
    int code = 0;
    int c, d;
    const byte *const decoder = scan_char_decoder;
#define IS_DIGIT(d, c)\
  ((d = decoder[c]) < 10)
#define WOULD_OVERFLOW(val, d, maxv)\
  (val >= maxv / 10 && (val > maxv / 10 || d > (int)(maxv % 10)))

    GET_NEXT(c, sp, return_error(e_syntaxerror));
    if (!IS_DIGIT(d, c)) {
	if (c != '.')
	    return_error(e_syntaxerror);
	/* Might be a number starting with '.'. */
	GET_NEXT(c, sp, return_error(e_syntaxerror));
	if (!IS_DIGIT(d, c))
	    return_error(e_syntaxerror);
	ival = 0;
	goto i2r;
    }
    /* Accumulate an integer in ival. */
    /* Do up to 4 digits without a loop, */
    /* since we know this can't overflow and since */
    /* most numbers have 4 (integer) digits or fewer. */
    ival = d;
    if (end - sp >= 3) {	/* just check once */
	if (!IS_DIGIT(d, (c = *sp))) {
	    sp++;
	    goto ind;
	}
	ival = ival * 10 + d;
	if (!IS_DIGIT(d, (c = sp[1]))) {
	    sp += 2;
	    goto ind;
	}
	ival = ival * 10 + d;
	sp += 3;
	if (!IS_DIGIT(d, (c = sp[-1])))
	    goto ind;
	ival = ival * 10 + d;
    }
    for (;; ival = ival * 10 + d) {
	GET_NEXT(c, sp, goto iret);
	if (!IS_DIGIT(d, c))
	    break;
	if (WOULD_OVERFLOW(ival, d, max_int))
	    goto i2l;
    }
  ind:				/* We saw a non-digit while accumulating an integer in ival. */
    switch (c) {
	case '.':
	    GET_NEXT(c, sp, c = EOFC);
	    goto i2r;
	default:
	    *psp = sp;
	    code = 1;
	    break;
	case 'e':
	case 'E':
	    if (sign < 0)
		ival = -ival;
	    dval = ival;
	    exp10 = 0;
	    goto fe;
	case '#':
	    {
		const uint radix = (uint)ival;
		ulong uval = 0, lmax;

		if (sign || radix < min_radix || radix > max_radix)
		    return_error(e_syntaxerror);
		/* Avoid multiplies for power-of-2 radix. */
		if (!(radix & (radix - 1))) {
		    int shift;

		    switch (radix) {
			case 2:
			    shift = 1, lmax = max_ulong >> 1;
			    break;
			case 4:
			    shift = 2, lmax = max_ulong >> 2;
			    break;
			case 8:
			    shift = 3, lmax = max_ulong >> 3;
			    break;
			case 16:
			    shift = 4, lmax = max_ulong >> 4;
			    break;
			case 32:
			    shift = 5, lmax = max_ulong >> 5;
			    break;
			default:	/* can't happen */
			    return_error(e_rangecheck);
		    }
		    for (;; uval = (uval << shift) + d) {
			GET_NEXT(c, sp, break);
			d = decoder[c];
			if (d >= radix) {
			    *psp = sp;
			    code = 1;
			    break;
			}
			if (uval > lmax)
			    return_error(e_limitcheck);
		    }
		} else {
		    int lrem = max_ulong % radix;

		    lmax = max_ulong / radix;
		    for (;; uval = uval * radix + d) {
			GET_NEXT(c, sp, break);
			d = decoder[c];
			if (d >= radix) {
			    *psp = sp;
			    code = 1;
			    break;
			}
			if (uval >= lmax &&
			    (uval > lmax || d > lrem)
			    )
			    return_error(e_limitcheck);
		    }
		}
		make_int(pref, uval);
		return code;
	    }
    }
iret:
    make_int(pref, (sign < 0 ? -ival : ival));
    return code;

    /* Accumulate a long in lval. */
i2l:
    for (lval = ival;;) {
	if (WOULD_OVERFLOW(lval, d, max_long)) {
	    /* Make a special check for entering the smallest */
	    /* (most negative) integer. */
	    if (lval == max_long / 10 &&
		d == (int)(max_long % 10) + 1 && sign < 0
		) {
		GET_NEXT(c, sp, c = EOFC);
		dval = -(double)min_long;
		if (c == 'e' || c == 'E') {
		    exp10 = 0;
		    goto fs;
		} else if (c == '.') {
                    GET_NEXT(c, sp, c = EOFC);
		    exp10 = 0;
		    goto fd;
                } else if (!IS_DIGIT(d, c)) {
		    lval = min_long;
		    break;
		}
	    } else
		dval = lval;
	    goto l2d;
	}
	lval = lval * 10 + d;
	GET_NEXT(c, sp, goto lret);
	if (!IS_DIGIT(d, c))
	    break;
    }
    switch (c) {
	case '.':
	    GET_NEXT(c, sp, c = EOFC);
	    exp10 = 0;
	    goto l2r;
	case EOFC:
	    break;
	default:
	    *psp = sp;
	    code = 1;
	    break;
	case 'e':
	case 'E':
	    exp10 = 0;
	    goto le;
	case '#':
	    return_error(e_syntaxerror);
    }
lret:
    make_int(pref, (sign < 0 ? -lval : lval));
    return code;

    /* Accumulate a double in dval. */
l2d:
    exp10 = 0;
    for (;;) {
	dval = dval * 10 + d;
	GET_NEXT(c, sp, c = EOFC);
	if (!IS_DIGIT(d, c))
	    break;
    }
    switch (c) {
	case '.':
	    GET_NEXT(c, sp, c = EOFC);
	    exp10 = 0;
	    goto fd;
	default:
	    *psp = sp;
	    code = 1;
	    /* falls through */
	case EOFC:
	    if (sign < 0)
		dval = -dval;
	    goto rret;
	case 'e':
	case 'E':
	    exp10 = 0;
	    goto fs;
	case '#':
	    return_error(e_syntaxerror);
    }

    /* We saw a '.' while accumulating an integer in ival. */
i2r:
    exp10 = 0;
    while (IS_DIGIT(d, c) || c == '-') {
	/*
	 * PostScript gives an error on numbers with a '-' following a '.'
	 * Adobe Acrobat Reader (PDF) apparently doesn't treat this as an
	 * error. Experiments show that the numbers following the '-' are
	 * ignored, so we swallow the fractional part. PDFScanInvNum enables
	 * this compatibility kloodge.
	 */
	if (c == '-') {
	    if (!PDFScanInvNum)
		break;
	    do {
		GET_NEXT(c, sp, c = EOFC);
	    } while (IS_DIGIT(d, c));
	    break;
	}
	if (WOULD_OVERFLOW(ival, d, max_int)) {
	    lval = ival;
	    goto l2r;
	}
	ival = ival * 10 + d;
	exp10--;
	GET_NEXT(c, sp, c = EOFC);
    }
    if (sign < 0)
	ival = -ival;
    /* Take a shortcut for the common case */
    if (!(c == 'e' || c == 'E' || exp10 < -NUM_POWERS_10)) {	/* Check for trailing garbage */
	if (c != EOFC)
	    *psp = sp, code = 1;
	make_real(pref, ival * neg_powers_10[-exp10]);
	return code;
    }
    dval = ival;
    goto fe;

    /* We saw a '.' while accumulating a long in lval. */
l2r:
    while (IS_DIGIT(d, c) || c == '-') {
	/* Handle bogus '-' following '.' as in i2r above.	*/
	if (c == '-') {
	    if (!PDFScanInvNum)
		break;
	    do {
		GET_NEXT(c, sp, c = EOFC);
	    } while (IS_DIGIT(d, c));
	    break;
	}
	if (WOULD_OVERFLOW(lval, d, max_long)) {
	    dval = lval;
	    goto fd;
	}
	lval = lval * 10 + d;
	exp10--;
	GET_NEXT(c, sp, c = EOFC);
    }
le:
    if (sign < 0)
	lval = -lval;
    dval = lval;
    goto fe;

    /* Now we are accumulating a double in dval. */
fd:
    while (IS_DIGIT(d, c)) {
	dval = dval * 10 + d;
	exp10--;
	GET_NEXT(c, sp, c = EOFC);
    }
fs:
    if (sign < 0)
	dval = -dval;
fe:
    /* Now dval contains the value, negated if necessary. */
    switch (c) {
	case 'e':
	case 'E':
	    {			/* Check for a following exponent. */
		int esign = 0;
		int iexp;

		GET_NEXT(c, sp, return_error(e_syntaxerror));
		switch (c) {
		    case '-':
			esign = 1;
		    case '+':
			GET_NEXT(c, sp, return_error(e_syntaxerror));
		}
		/* Scan the exponent.  We limit it arbitrarily to 999. */
		if (!IS_DIGIT(d, c))
		    return_error(e_syntaxerror);
		iexp = d;
		for (;; iexp = iexp * 10 + d) {
		    GET_NEXT(c, sp, break);
		    if (!IS_DIGIT(d, c)) {
			*psp = sp;
			code = 1;
			break;
		    }
		    if (iexp > 99)
			return_error(e_limitcheck);
		}
		if (esign)
		    exp10 -= iexp;
		else
		    exp10 += iexp;
		break;
	    }
	default:
	    *psp = sp;
	    code = 1;
	case EOFC:
	    ;
    }
    /* Compute dval * 10^exp10. */
    if (exp10 > 0) {
	while (exp10 > NUM_POWERS_10)
	    dval *= powers_10[NUM_POWERS_10],
		exp10 -= NUM_POWERS_10;
	if (exp10 > 0)
	    dval *= powers_10[exp10];
    } else if (exp10 < 0) {
	while (exp10 < -NUM_POWERS_10)
	    dval /= powers_10[NUM_POWERS_10],
		exp10 += NUM_POWERS_10;
	if (exp10 < 0)
	    dval /= powers_10[-exp10];
    }
    /*
     * Check for an out-of-range result.  Currently we don't check for
     * absurdly large numbers of digits in the accumulation loops,
     * but we should.
     */
    if (dval >= 0) {
	if (dval > MAX_FLOAT)
	    return_error(e_limitcheck);
    } else {
	if (dval < -MAX_FLOAT)
	    return_error(e_limitcheck);
    }
rret:
    make_real(pref, dval);
    return code;
}