gecko-dev/xpcom/ds/nsTextFormater.cpp

1413 строки
30 KiB
C++

/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/*
* The contents of this file are subject to the Netscape Public License
* Version 1.1 (the "NPL"); you may not use this file except in
* compliance with the NPL. You may obtain a copy of the NPL at
* http://www.mozilla.org/NPL/
*
* Software distributed under the NPL is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the NPL
* for the specific language governing rights and limitations under the
* NPL.
*
* The Initial Developer of this code under the NPL is Netscape
* Communications Corporation. Portions created by Netscape are
* Copyright (C) 1998 Netscape Communications Corporation. All Rights
* Reserved.
*/
/* The code are copy from rev 3.7 mozilla/nsprpub/src/io/prprf.c */
/*
** Port from prprf.c by : Frank Yung-Fong Tang
*/
/*
** Portable safe sprintf code.
**
** Author: Kipp E.B. Hickman
*/
#include <stdarg.h>
#include <stddef.h>
#include <stdio.h>
#include <string.h>
#include "prlong.h"
#include "prlog.h"
#include "prmem.h"
#include "nsCRT.h"
#include "nsTextFormater.h"
#include "nsString.h"
/*
** Note: on some platforms va_list is defined as an array,
** and requires array notation.
*/
#ifdef HAVE_VA_LIST_AS_ARRAY
#define VARARGS_ASSIGN(foo, bar) foo[0] = bar[0]
#else
#define VARARGS_ASSIGN(foo, bar) (foo) = (bar)
#endif
/*
** WARNING: This code may *NOT* call PR_LOG (because PR_LOG calls it)
*/
/*
** XXX This needs to be internationalized!
*/
typedef struct SprintfStateStr SprintfState;
struct SprintfStateStr {
int (*stuff)(SprintfState *ss, const PRUnichar *sp, PRUint32 len);
PRUnichar *base;
PRUnichar *cur;
PRUint32 maxlen;
int (*func)(void *arg, const PRUnichar *sp, PRUint32 len);
void *arg;
};
/*
** Numbered Arguement State
*/
struct NumArgState{
int type; /* type of the current ap */
va_list ap; /* point to the corresponding position on ap */
};
static PRBool l10n_debug_init = PR_FALSE;
static PRBool l10n_debug = PR_FALSE;
#define NAS_DEFAULT_NUM 20 /* default number of NumberedArgumentState array */
#define TYPE_INT16 0
#define TYPE_UINT16 1
#define TYPE_INTN 2
#define TYPE_UINTN 3
#define TYPE_INT32 4
#define TYPE_UINT32 5
#define TYPE_INT64 6
#define TYPE_UINT64 7
#define TYPE_STRING 8
#define TYPE_DOUBLE 9
#define TYPE_INTSTR 10
#define TYPE_UNISTRING 11
#define TYPE_UNKNOWN 20
#define _LEFT 0x1
#define _SIGNED 0x2
#define _SPACED 0x4
#define _ZEROS 0x8
#define _NEG 0x10
/*
** Fill into the buffer using the data in src
*/
static int fill2(SprintfState *ss, const PRUnichar *src, int srclen, int width,
int flags)
{
PRUnichar space = ' ';
int rv;
width -= srclen;
if ((width > 0) && ((flags & _LEFT) == 0)) { /* Right adjusting */
if (flags & _ZEROS) {
space = '0';
}
while (--width >= 0) {
rv = (*ss->stuff)(ss, &space, 1);
if (rv < 0) {
return rv;
}
}
}
/* Copy out the source data */
rv = (*ss->stuff)(ss, src, srclen);
if (rv < 0) {
return rv;
}
if ((width > 0) && ((flags & _LEFT) != 0)) { /* Left adjusting */
while (--width >= 0) {
rv = (*ss->stuff)(ss, &space, 1);
if (rv < 0) {
return rv;
}
}
}
return 0;
}
/*
** Fill a number. The order is: optional-sign zero-filling conversion-digits
*/
static int fill_n(SprintfState *ss, const PRUnichar *src, int srclen, int width,
int prec, int type, int flags)
{
int zerowidth = 0;
int precwidth = 0;
int signwidth = 0;
int leftspaces = 0;
int rightspaces = 0;
int cvtwidth;
int rv;
PRUnichar sign;
PRUnichar space = ' ';
PRUnichar zero = '0';
if ((type & 1) == 0) {
if (flags & _NEG) {
sign = '-';
signwidth = 1;
} else if (flags & _SIGNED) {
sign = '+';
signwidth = 1;
} else if (flags & _SPACED) {
sign = ' ';
signwidth = 1;
}
}
cvtwidth = signwidth + srclen;
if (prec > 0) {
if (prec > srclen) {
precwidth = prec - srclen; /* Need zero filling */
cvtwidth += precwidth;
}
}
if ((flags & _ZEROS) && (prec < 0)) {
if (width > cvtwidth) {
zerowidth = width - cvtwidth; /* Zero filling */
cvtwidth += zerowidth;
}
}
if (flags & _LEFT) {
if (width > cvtwidth) {
/* Space filling on the right (i.e. left adjusting) */
rightspaces = width - cvtwidth;
}
} else {
if (width > cvtwidth) {
/* Space filling on the left (i.e. right adjusting) */
leftspaces = width - cvtwidth;
}
}
while (--leftspaces >= 0) {
rv = (*ss->stuff)(ss, &space, 1);
if (rv < 0) {
return rv;
}
}
if (signwidth) {
rv = (*ss->stuff)(ss, &sign, 1);
if (rv < 0) {
return rv;
}
}
while (--precwidth >= 0) {
rv = (*ss->stuff)(ss, &space, 1);
if (rv < 0) {
return rv;
}
}
while (--zerowidth >= 0) {
rv = (*ss->stuff)(ss, &zero, 1);
if (rv < 0) {
return rv;
}
}
rv = (*ss->stuff)(ss, src, srclen);
if (rv < 0) {
return rv;
}
while (--rightspaces >= 0) {
rv = (*ss->stuff)(ss, &space, 1);
if (rv < 0) {
return rv;
}
}
return 0;
}
/*
** Convert a long into its printable form
*/
static int cvt_l(SprintfState *ss, long num, int width, int prec, int radix,
int type, int flags, const PRUnichar *hexp)
{
PRUnichar cvtbuf[100];
PRUnichar *cvt;
int digits;
/* according to the man page this needs to happen */
if ((prec == 0) && (num == 0)) {
return 0;
}
/*
** Converting decimal is a little tricky. In the unsigned case we
** need to stop when we hit 10 digits. In the signed case, we can
** stop when the number is zero.
*/
cvt = &cvtbuf[0] + sizeof(cvtbuf);
digits = 0;
while (num) {
int digit = (((unsigned long)num) % radix) & 0xF;
*--cvt = hexp[digit];
digits++;
num = (long)(((unsigned long)num) / radix);
}
if (digits == 0) {
*--cvt = '0';
digits++;
}
/*
** Now that we have the number converted without its sign, deal with
** the sign and zero padding.
*/
return fill_n(ss, cvt, digits, width, prec, type, flags);
}
/*
** Convert a 64-bit integer into its printable form
*/
static int cvt_ll(SprintfState *ss, PRInt64 num, int width, int prec, int radix,
int type, int flags, const PRUnichar *hexp)
{
PRUnichar cvtbuf[100];
PRUnichar *cvt;
int digits;
PRInt64 rad;
/* according to the man page this needs to happen */
if ((prec == 0) && (LL_IS_ZERO(num))) {
return 0;
}
/*
** Converting decimal is a little tricky. In the unsigned case we
** need to stop when we hit 10 digits. In the signed case, we can
** stop when the number is zero.
*/
LL_I2L(rad, radix);
cvt = &cvtbuf[0] + sizeof(cvtbuf);
digits = 0;
while (!LL_IS_ZERO(num)) {
PRInt32 digit;
PRInt64 quot, rem;
LL_UDIVMOD(&quot, &rem, num, rad);
LL_L2I(digit, rem);
*--cvt = hexp[digit & 0xf];
digits++;
num = quot;
}
if (digits == 0) {
*--cvt = '0';
digits++;
}
/*
** Now that we have the number converted without its sign, deal with
** the sign and zero padding.
*/
return fill_n(ss, cvt, digits, width, prec, type, flags);
}
/*
** Convert a double precision floating point number into its printable
** form.
**
** XXX stop using sprintf to convert floating point
*/
static int cvt_f(SprintfState *ss, double d, const PRUnichar *fmt0, const PRUnichar *fmt1)
{
char fin[20];
char fout[300];
PRUnichar fout2[300];
int amount = fmt1 - fmt0;
int i;
PR_ASSERT((amount > 0) && (amount < (int)sizeof(fin)));
if (amount >= (int)sizeof(fin)) {
/* Totally bogus % command to sprintf. Just ignore it */
return 0;
}
for(i=0;i<amount;i++)
fin[i] = (char) fmt0[i]; // cast down here
fin[amount] = 0;
/* Convert floating point using the native sprintf code */
#ifdef DEBUG
{
const char *p = fin;
while (*p) {
PR_ASSERT(*p != 'L');
p++;
}
}
#endif
sprintf(fout, fin, d);
/*
** This assert will catch overflow's of fout, when building with
** debugging on. At least this way we can track down the evil piece
** of calling code and fix it!
*/
PR_ASSERT((nsCRT::strlen(fout)*2) < sizeof(fout));
for(i=0; fout[i]; i++)
fout2[i]=fout[i];
fout2[i] = 0;
return (*ss->stuff)(ss, fout2, nsCRT::strlen(fout2));
}
/*
** Convert a string into its printable form. "width" is the output
** width. "prec" is the maximum number of characters of "s" to output,
** where -1 means until NUL.
*/
static int cvt_S(SprintfState *ss, const PRUnichar *s, int width, int prec,
int flags)
{
int slen;
if (prec == 0)
return 0;
/* Limit string length by precision value */
slen = s ? nsCRT::strlen(s) : 6;
if (prec > 0) {
if (prec < slen) {
slen = prec;
}
}
/* and away we go */
nsAutoString nullstr("(null)");
return fill2(ss, s ? s : nullstr.GetUnicode(), slen, width, flags);
}
static PRUnichar* UTF8ToUCS2(const char *aSrc, PRUint32 aSrcLen, PRUnichar* aDest, PRUint32 aDestLen)
{
const char *in, *inend;
inend = aSrc + aSrcLen;
PRUnichar *out;
PRUint32 state;
PRUint32 ucs4;
// decide the length of the UCS2 first.
PRUint32 needLen = 0;
for(in=aSrc,state=0,ucs4=0;in < inend; in++)
{
if(0 == state) {
if( 0 == (0x80 & (*in))) {
needLen++;
} else if( 0xC0 == (0xE0 & (*in))) {
needLen++;
state=1;
} else if( 0xE0 == (0xF0 & (*in))) {
needLen++;
state=2;
} else if( 0xF0 == (0xF8 & (*in))) {
needLen+=2;
state=3;
} else if( 0xF8 == (0xFC & (*in))) {
needLen+=2;
state=4;
} else if( 0xFC == (0xFE & (*in))) {
needLen+=2;
state=5;
} else {
needLen++;
state=0;
}
} else {
NS_ASSERTION( (0x80 == (0xC0 & (*in))) , "The input string is not in utf8");
if(0x80 == (0xC0 & (*in)))
{
state--;
} else {
state=0;
}
}
}
needLen++; // add null termination.
if(needLen >= aDestLen)
aDest = (PRUnichar*)PR_MALLOC(sizeof(PRUnichar) * needLen);
if(nsnull == aDest)
return nsnull;
out= aDest;
for(in=aSrc,state=0,ucs4=0;in < inend; in++)
{
if(0 == state) {
if( 0 == (0x80 & (*in))) {
// ASCII
*out++ = (PRUnichar)*in;
} else if( 0xC0 == (0xE0 & (*in))) {
// 2 bytes UTF8
ucs4 = (PRUint32)(*in);
ucs4 = (ucs4 << 6) & 0x000007C0L;
state=1;
} else if( 0xE0 == (0xF0 & (*in))) {
ucs4 = (PRUint32)(*in);
ucs4 = (ucs4 << 12) & 0x0000F000L;
state=2;
} else if( 0xF0 == (0xF8 & (*in))) {
ucs4 = (PRUint32)(*in);
ucs4 = (ucs4 << 18) & 0x001F0000L;
state=3;
} else if( 0xF8 == (0xFC & (*in))) {
ucs4 = (PRUint32)(*in);
ucs4 = (ucs4 << 24) & 0x03000000L;
state=4;
} else if( 0xFC == (0xFE & (*in))) {
ucs4 = (PRUint32)(*in);
ucs4 = (ucs4 << 30) & 0x40000000L;
state=5;
} else {
NS_ASSERTION(0, "The input string is not in utf8");
state=0;
ucs4=0;
}
} else {
NS_ASSERTION( (0x80 == (0xC0 & (*in))) , "The input string is not in utf8");
if(0x80 == (0xC0 & (*in)))
{
PRUint32 tmp = (*in);
int shift = (state-1) * 6;
tmp = (tmp << shift ) & ( 0x0000003FL << shift);
ucs4 |= tmp;
if(0 == --state)
{
if(ucs4 >= 0x00010000) {
if(ucs4 >= 0x001F0000) {
*out++ = 0xFFFD;
} else {
ucs4 -= 0x00010000;
*out++ = 0xD800 | (0x000003FF & (ucs4 >> 10));
*out++ = 0xDC00 | (0x000003FF & ucs4);
}
} else {
*out++ = ucs4;
}
ucs4=0;
}
} else {
state=0;
ucs4=0;
}
}
}
*out = 0x0000;
return aDest;
}
/*
** Convert a string into its printable form. "width" is the output
** width. "prec" is the maximum number of characters of "s" to output,
** where -1 means until NUL.
*/
static int cvt_s(SprintfState *ss, const char *s, int width, int prec,
int flags)
{
// convert s from UTF8 to PRUnichar*
// Fix me !!!
PRUnichar buf[256];
PRUnichar *retbuf = nsnull;
retbuf = UTF8ToUCS2(s, nsCRT::strlen(s), buf, 256);
if(nsnull == retbuf)
return -1;
int ret = cvt_S(ss, retbuf, width, prec, flags);
if(retbuf != buf)
PR_DELETE(retbuf);
return ret;
}
/*
** BiuldArgArray stands for Numbered Argument list Sprintf
** for example,
** fmp = "%4$i, %2$d, %3s, %1d";
** the number must start from 1, and no gap among them
*/
static struct NumArgState* BuildArgArray( const PRUnichar *fmt, va_list ap, int* rv, struct NumArgState* nasArray )
{
int number = 0, cn = 0, i;
const PRUnichar* p;
PRUnichar c;
struct NumArgState* nas;
/*
** set the l10n_debug flag
** this routine should be executed only once
** 'cause getenv does take time
*/
if( !l10n_debug_init ){
l10n_debug_init = PR_TRUE;
const char *env;
env = getenv( "NETSCAPE_LOCALIZATION_DEBUG" );
if( ( env != NULL ) && ( *env == '1' ) ){
l10n_debug = PR_TRUE;
}
}
/*
** first pass:
** detemine how many legal % I have got, then allocate space
*/
p = fmt;
*rv = 0;
i = 0;
while( ( c = *p++ ) != 0 ){
if( c != '%' )
continue;
if( ( c = *p++ ) == '%' ) /* skip %% case */
continue;
while( c != 0 ){
if( c > '9' || c < '0' ){
if( c == '$' ){ /* numbered argument csae */
if( i > 0 ){
*rv = -1;
return NULL;
}
number++;
break;
} else{ /* non-numbered argument case */
if( number > 0 ){
*rv = -1;
return NULL;
}
i = 1;
break;
}
}
c = *p++;
}
}
if( number == 0 ){
return NULL;
}
if( number > NAS_DEFAULT_NUM ){
nas = (struct NumArgState*)PR_MALLOC( number * sizeof( struct NumArgState ) );
if( !nas ){
*rv = -1;
return NULL;
}
} else {
nas = nasArray;
}
for( i = 0; i < number; i++ ){
nas[i].type = TYPE_UNKNOWN;
}
/*
** second pass:
** set nas[].type
*/
p = fmt;
while( ( c = *p++ ) != 0 ){
if( c != '%' ) continue;
c = *p++;
if( c == '%' ) continue;
cn = 0;
while( c && c != '$' ){ /* should imporve error check later */
cn = cn*10 + c - '0';
c = *p++;
}
if( !c || cn < 1 || cn > number ){
*rv = -1;
break;
}
/* nas[cn] starts from 0, and make sure nas[cn].type is not assigned */
cn--;
if( nas[cn].type != TYPE_UNKNOWN )
continue;
c = *p++;
/* width */
if (c == '*') {
/* not supported feature, for the argument is not numbered */
*rv = -1;
break;
} else {
while ((c >= '0') && (c <= '9')) {
c = *p++;
}
}
/* precision */
if (c == '.') {
c = *p++;
if (c == '*') {
/* not supported feature, for the argument is not numbered */
*rv = -1;
break;
} else {
while ((c >= '0') && (c <= '9')) {
c = *p++;
}
}
}
/* size */
nas[cn].type = TYPE_INTN;
if (c == 'h') {
nas[cn].type = TYPE_INT16;
c = *p++;
} else if (c == 'L') {
/* XXX not quite sure here */
nas[cn].type = TYPE_INT64;
c = *p++;
} else if (c == 'l') {
nas[cn].type = TYPE_INT32;
c = *p++;
if (c == 'l') {
nas[cn].type = TYPE_INT64;
c = *p++;
}
}
/* format */
switch (c) {
case 'd':
case 'c':
case 'i':
case 'o':
case 'u':
case 'x':
case 'X':
break;
case 'e':
case 'f':
case 'g':
nas[ cn ].type = TYPE_DOUBLE;
break;
case 'p':
/* XXX should use cpp */
if (sizeof(void *) == sizeof(PRInt32)) {
nas[ cn ].type = TYPE_UINT32;
} else if (sizeof(void *) == sizeof(PRInt64)) {
nas[ cn ].type = TYPE_UINT64;
} else if (sizeof(void *) == sizeof(PRIntn)) {
nas[ cn ].type = TYPE_UINTN;
} else {
nas[ cn ].type = TYPE_UNKNOWN;
}
break;
case 'C':
//case 'S':
case 'E':
case 'G':
/* XXX not supported I suppose */
PR_ASSERT(0);
nas[ cn ].type = TYPE_UNKNOWN;
break;
case 'S':
nas[ cn ].type = TYPE_UNISTRING;
break;
case 's':
nas[ cn ].type = TYPE_STRING;
break;
case 'n':
nas[ cn ].type = TYPE_INTSTR;
break;
default:
PR_ASSERT(0);
nas[ cn ].type = TYPE_UNKNOWN;
break;
}
/* get a legal para. */
if( nas[ cn ].type == TYPE_UNKNOWN ){
*rv = -1;
break;
}
}
/*
** third pass
** fill the nas[cn].ap
*/
if( *rv < 0 ){
if( nas != nasArray )
PR_DELETE( nas );
return NULL;
}
cn = 0;
while( cn < number ){
if( nas[cn].type == TYPE_UNKNOWN ){
cn++;
continue;
}
VARARGS_ASSIGN(nas[cn].ap, ap);
switch( nas[cn].type ){
case TYPE_INT16:
case TYPE_UINT16:
case TYPE_INTN:
case TYPE_UINTN: (void)va_arg( ap, PRIntn ); break;
case TYPE_INT32: (void)va_arg( ap, PRInt32 ); break;
case TYPE_UINT32: (void)va_arg( ap, PRUint32 ); break;
case TYPE_INT64: (void)va_arg( ap, PRInt64 ); break;
case TYPE_UINT64: (void)va_arg( ap, PRUint64 ); break;
case TYPE_STRING: (void)va_arg( ap, char* ); break;
case TYPE_INTSTR: (void)va_arg( ap, PRIntn* ); break;
case TYPE_DOUBLE: (void)va_arg( ap, double ); break;
case TYPE_UNISTRING: (void)va_arg( ap, PRUnichar* ); break;
default:
if( nas != nasArray )
PR_DELETE( nas );
*rv = -1;
return NULL;
}
cn++;
}
return nas;
}
/*
** The workhorse sprintf code.
*/
static int dosprintf(SprintfState *ss, const PRUnichar *fmt, va_list ap)
{
PRUnichar c;
int flags, width, prec, radix, type;
union {
PRUnichar ch;
int i;
long l;
PRInt64 ll;
double d;
const char *s;
const PRUnichar *S;
int *ip;
} u;
PRUnichar space = ' ';
const PRUnichar *fmt0;
nsAutoString hex("0123456789abcdef");
nsAutoString HEX("0123456789ABCDEF");
const PRUnichar *hexp;
int rv, i;
struct NumArgState* nas = NULL;
struct NumArgState nasArray[ NAS_DEFAULT_NUM ];
PRUnichar pattern[20];
const PRUnichar* dolPt = NULL; /* in "%4$.2f", dolPt will poiont to . */
/*
** build an argument array, IF the fmt is numbered argument
** list style, to contain the Numbered Argument list pointers
*/
nas = BuildArgArray( fmt, ap, &rv, nasArray );
if( rv < 0 ){
/* the fmt contains error Numbered Argument format, jliu@netscape.com */
PR_ASSERT(0);
return rv;
}
while ((c = *fmt++) != 0) {
if (c != '%') {
rv = (*ss->stuff)(ss, fmt - 1, 1);
if (rv < 0) {
return rv;
}
continue;
}
fmt0 = fmt - 1;
/*
** Gobble up the % format string. Hopefully we have handled all
** of the strange cases!
*/
flags = 0;
c = *fmt++;
if (c == '%') {
/* quoting a % with %% */
rv = (*ss->stuff)(ss, fmt - 1, 1);
if (rv < 0) {
return rv;
}
continue;
}
if( nas != NULL ){
/* the fmt contains the Numbered Arguments feature */
i = 0;
while( c && c != '$' ){ /* should imporve error check later */
i = ( i * 10 ) + ( c - '0' );
c = *fmt++;
}
if( nas[i-1].type == TYPE_UNKNOWN ){
if( nas && ( nas != nasArray ) )
PR_DELETE( nas );
return -1;
}
ap = nas[i-1].ap;
dolPt = fmt;
c = *fmt++;
}
/*
* Examine optional flags. Note that we do not implement the
* '#' flag of sprintf(). The ANSI C spec. of the '#' flag is
* somewhat ambiguous and not ideal, which is perhaps why
* the various sprintf() implementations are inconsistent
* on this feature.
*/
while ((c == '-') || (c == '+') || (c == ' ') || (c == '0')) {
if (c == '-') flags |= _LEFT;
if (c == '+') flags |= _SIGNED;
if (c == ' ') flags |= _SPACED;
if (c == '0') flags |= _ZEROS;
c = *fmt++;
}
if (flags & _SIGNED) flags &= ~_SPACED;
if (flags & _LEFT) flags &= ~_ZEROS;
/* width */
if (c == '*') {
c = *fmt++;
width = va_arg(ap, int);
} else {
width = 0;
while ((c >= '0') && (c <= '9')) {
width = (width * 10) + (c - '0');
c = *fmt++;
}
}
/* precision */
prec = -1;
if (c == '.') {
c = *fmt++;
if (c == '*') {
c = *fmt++;
prec = va_arg(ap, int);
} else {
prec = 0;
while ((c >= '0') && (c <= '9')) {
prec = (prec * 10) + (c - '0');
c = *fmt++;
}
}
}
/* size */
type = TYPE_INTN;
if (c == 'h') {
type = TYPE_INT16;
c = *fmt++;
} else if (c == 'L') {
/* XXX not quite sure here */
type = TYPE_INT64;
c = *fmt++;
} else if (c == 'l') {
type = TYPE_INT32;
c = *fmt++;
if (c == 'l') {
type = TYPE_INT64;
c = *fmt++;
}
}
/* format */
hexp = hex.GetUnicode();
switch (c) {
case 'd': case 'i': /* decimal/integer */
radix = 10;
goto fetch_and_convert;
case 'o': /* octal */
radix = 8;
type |= 1;
goto fetch_and_convert;
case 'u': /* unsigned decimal */
radix = 10;
type |= 1;
goto fetch_and_convert;
case 'x': /* unsigned hex */
radix = 16;
type |= 1;
goto fetch_and_convert;
case 'X': /* unsigned HEX */
radix = 16;
hexp = HEX.GetUnicode();
type |= 1;
goto fetch_and_convert;
fetch_and_convert:
switch (type) {
case TYPE_INT16:
u.l = va_arg(ap, int);
if (u.l < 0) {
u.l = -u.l;
flags |= _NEG;
}
goto do_long;
case TYPE_UINT16:
u.l = va_arg(ap, int) & 0xffff;
goto do_long;
case TYPE_INTN:
u.l = va_arg(ap, int);
if (u.l < 0) {
u.l = -u.l;
flags |= _NEG;
}
goto do_long;
case TYPE_UINTN:
u.l = (long)va_arg(ap, unsigned int);
goto do_long;
case TYPE_INT32:
u.l = va_arg(ap, PRInt32);
if (u.l < 0) {
u.l = -u.l;
flags |= _NEG;
}
goto do_long;
case TYPE_UINT32:
u.l = (long)va_arg(ap, PRUint32);
do_long:
rv = cvt_l(ss, u.l, width, prec, radix, type, flags, hexp);
if (rv < 0) {
return rv;
}
break;
case TYPE_INT64:
u.ll = va_arg(ap, PRInt64);
if (!LL_GE_ZERO(u.ll)) {
LL_NEG(u.ll, u.ll);
flags |= _NEG;
}
goto do_longlong;
case TYPE_UINT64:
u.ll = va_arg(ap, PRUint64);
do_longlong:
rv = cvt_ll(ss, u.ll, width, prec, radix, type, flags, hexp);
if (rv < 0) {
return rv;
}
break;
}
break;
case 'e':
case 'E':
case 'f':
case 'g':
u.d = va_arg(ap, double);
if( nas != NULL ){
i = fmt - dolPt;
if( i < (int)sizeof( pattern ) ){
pattern[0] = '%';
memcpy( &pattern[1], dolPt, i*sizeof(PRUnichar) );
rv = cvt_f(ss, u.d, pattern, &pattern[i+1] );
}
} else
rv = cvt_f(ss, u.d, fmt0, fmt);
if (rv < 0) {
return rv;
}
break;
case 'c':
u.ch = va_arg(ap, int);
if ((flags & _LEFT) == 0) {
while (width-- > 1) {
rv = (*ss->stuff)(ss, &space, 1);
if (rv < 0) {
return rv;
}
}
}
rv = (*ss->stuff)(ss, &u.ch, 1);
if (rv < 0) {
return rv;
}
if (flags & _LEFT) {
while (width-- > 1) {
rv = (*ss->stuff)(ss, &space, 1);
if (rv < 0) {
return rv;
}
}
}
break;
case 'p':
if (sizeof(void *) == sizeof(PRInt32)) {
type = TYPE_UINT32;
} else if (sizeof(void *) == sizeof(PRInt64)) {
type = TYPE_UINT64;
} else if (sizeof(void *) == sizeof(int)) {
type = TYPE_UINTN;
} else {
PR_ASSERT(0);
break;
}
radix = 16;
goto fetch_and_convert;
#if 0
case 'C':
//case 'S':
case 'E':
case 'G':
/* XXX not supported I suppose */
PR_ASSERT(0);
break;
#endif
case 'S':
u.S = va_arg(ap, const PRUnichar*);
rv = cvt_S(ss, u.S, width, prec, flags);
if (rv < 0) {
return rv;
}
break;
case 's':
u.s = va_arg(ap, const char*);
rv = cvt_s(ss, u.s, width, prec, flags);
if (rv < 0) {
return rv;
}
break;
case 'n':
u.ip = va_arg(ap, int*);
if (u.ip) {
*u.ip = ss->cur - ss->base;
}
break;
default:
/* Not a % token after all... skip it */
#if 0
PR_ASSERT(0);
#endif
PRUnichar perct = '%';
rv = (*ss->stuff)(ss, &perct, 1);
if (rv < 0) {
return rv;
}
rv = (*ss->stuff)(ss, fmt - 1, 1);
if (rv < 0) {
return rv;
}
}
}
/* Stuff trailing NUL */
PRUnichar null = '\0';
rv = (*ss->stuff)(ss, &null, 1);
if( nas && ( nas != nasArray ) ){
PR_DELETE( nas );
}
return rv;
}
/************************************************************************/
#if 0
static int FuncStuff(SprintfState *ss, const PRUnichar *sp, PRUint32 len)
{
int rv;
rv = (*ss->func)(ss->arg, sp, len);
if (rv < 0) {
return rv;
}
ss->maxlen += len;
return 0;
}
PRUint32 nsTextFormater::sxprintf(PRStuffFunc func, void *arg,
const PRUnichar *fmt, ...)
{
va_list ap;
int rv;
va_start(ap, fmt);
rv = nsTextFormater::vsxprintf(func, arg, fmt, ap);
va_end(ap);
return rv;
}
PRUint32) vsxprintf(PRStuffFunc func, void *arg,
const PRUnichar *fmt, va_list ap)
{
SprintfState ss;
int rv;
ss.stuff = FuncStuff;
ss.func = func;
ss.arg = arg;
ss.maxlen = 0;
rv = dosprintf(&ss, fmt, ap);
return (rv < 0) ? (PRUint32)-1 : ss.maxlen;
}
#endif
/*
** Stuff routine that automatically grows the malloc'd output buffer
** before it overflows.
*/
static int GrowStuff(SprintfState *ss, const PRUnichar *sp, PRUint32 len)
{
ptrdiff_t off;
PRUnichar *newbase;
PRUint32 newlen;
off = ss->cur - ss->base;
if (off + len >= ss->maxlen) {
/* Grow the buffer */
newlen = ss->maxlen + ((len > 32) ? len : 32);
if (ss->base) {
newbase = (PRUnichar*) PR_REALLOC(ss->base, newlen*sizeof(PRUnichar));
} else {
newbase = (PRUnichar*) PR_MALLOC(newlen*sizeof(PRUnichar));
}
if (!newbase) {
/* Ran out of memory */
return -1;
}
ss->base = newbase;
ss->maxlen = newlen;
ss->cur = ss->base + off;
}
/* Copy data */
while (len) {
--len;
*ss->cur++ = *sp++;
}
PR_ASSERT((PRUint32)(ss->cur - ss->base) <= ss->maxlen);
return 0;
}
/*
** sprintf into a malloc'd buffer
*/
PRUnichar * nsTextFormater::smprintf(const PRUnichar *fmt, ...)
{
va_list ap;
PRUnichar *rv;
va_start(ap, fmt);
rv = nsTextFormater::vsmprintf(fmt, ap);
va_end(ap);
return rv;
}
/*
** Free memory allocated, for the caller, by smprintf
*/
void nsTextFormater::smprintf_free(PRUnichar *mem)
{
PR_DELETE(mem);
}
PRUnichar * nsTextFormater::vsmprintf(const PRUnichar *fmt, va_list ap)
{
SprintfState ss;
int rv;
ss.stuff = GrowStuff;
ss.base = 0;
ss.cur = 0;
ss.maxlen = 0;
rv = dosprintf(&ss, fmt, ap);
if (rv < 0) {
if (ss.base) {
PR_DELETE(ss.base);
}
return 0;
}
return ss.base;
}
/*
** Stuff routine that discards overflow data
*/
static int LimitStuff(SprintfState *ss, const PRUnichar *sp, PRUint32 len)
{
PRUint32 limit = ss->maxlen - (ss->cur - ss->base);
if (len > limit) {
len = limit;
}
while (len) {
--len;
*ss->cur++ = *sp++;
}
return 0;
}
/*
** sprintf into a fixed size buffer. Make sure there is a NUL at the end
** when finished.
*/
PRUint32 nsTextFormater::snprintf(PRUnichar *out, PRUint32 outlen, const PRUnichar *fmt, ...)
{
va_list ap;
int rv;
PR_ASSERT((PRInt32)outlen > 0);
if ((PRInt32)outlen <= 0) {
return 0;
}
va_start(ap, fmt);
rv = nsTextFormater::vsnprintf(out, outlen, fmt, ap);
va_end(ap);
return rv;
}
PRUint32 nsTextFormater::vsnprintf(PRUnichar *out, PRUint32 outlen,const PRUnichar *fmt,
va_list ap)
{
SprintfState ss;
PRUint32 n;
PR_ASSERT((PRInt32)outlen > 0);
if ((PRInt32)outlen <= 0) {
return 0;
}
ss.stuff = LimitStuff;
ss.base = out;
ss.cur = out;
ss.maxlen = outlen;
(void) dosprintf(&ss, fmt, ap);
/* If we added chars, and we didn't append a null, do it now. */
if( (ss.cur != ss.base) && (*(ss.cur - 1) != '\0') )
*(--ss.cur) = '\0';
n = ss.cur - ss.base;
return n ? n - 1 : n;
}
PRUnichar * nsTextFormater::sprintf_append(PRUnichar *last, const PRUnichar *fmt, ...)
{
va_list ap;
PRUnichar *rv;
va_start(ap, fmt);
rv = nsTextFormater::vsprintf_append(last, fmt, ap);
va_end(ap);
return rv;
}
PRUnichar * nsTextFormater::vsprintf_append(PRUnichar *last, const PRUnichar *fmt, va_list ap)
{
SprintfState ss;
int rv;
ss.stuff = GrowStuff;
if (last) {
int lastlen = nsCRT::strlen(last);
ss.base = last;
ss.cur = last + lastlen;
ss.maxlen = lastlen;
} else {
ss.base = 0;
ss.cur = 0;
ss.maxlen = 0;
}
rv = dosprintf(&ss, fmt, ap);
if (rv < 0) {
if (ss.base) {
PR_DELETE(ss.base);
}
return 0;
}
return ss.base;
}
#ifdef DEBUG
PRBool nsTextFormater::SelfTest()
{
PRBool passed = PR_TRUE ;
nsAutoString fmt("%3$s %4$S %1$d %2$d");
char utf8[] = "Hello";
PRUnichar ucs2[]={'W', 'o', 'r', 'l', 'd', 0x4e00, 0xAc00, 0xFF45, 0x0103};
int d=3;
PRUnichar buf[256];
int ret;
ret = nsTextFormater::snprintf(buf, 256, fmt.GetUnicode(), d, 333, utf8, ucs2);
printf("ret = %d\n", ret);
nsAutoString out(buf);
printf("%s \n",out.ToNewCString());
const PRUnichar *uout = out.GetUnicode();
for(int i=0;i<out.Length();i++)
printf("%2X ", uout[i]);
return passed;
}
#endif