php-src/ext/standard/html.c
2001-08-11 17:03:37 +00:00

503 lines
14 KiB
C

/*
+----------------------------------------------------------------------+
| PHP version 4.0 |
+----------------------------------------------------------------------+
| Copyright (c) 1997-2001 The PHP Group |
+----------------------------------------------------------------------+
| This source file is subject to version 2.02 of the PHP license, |
| that is bundled with this package in the file LICENSE, and is |
| available at through the world-wide-web at |
| http://www.php.net/license/2_02.txt. |
| If you did not receive a copy of the PHP license and are unable to |
| obtain it through the world-wide-web, please send a note to |
| license@php.net so we can mail you a copy immediately. |
+----------------------------------------------------------------------+
| Authors: Rasmus Lerdorf <rasmus@lerdorf.on.ca> |
| Jaakko Hyvätti <jaakko.hyvatti@iki.fi> |
| Wez Furlong <wez@thebrainroom.com> |
+----------------------------------------------------------------------+
*/
/* $Id$ */
#include "php.h"
#include "reg.h"
#include "html.h"
#if HAVE_LOCALE_H
#include <locale.h>
#endif
#if HAVE_LANGINFO_H
#include <langinfo.h>
#endif
/* This must be fixed to handle the input string according to LC_CTYPE.
Defaults to ISO-8859-1 for now. */
enum entity_charset { cs_terminator, cs_8859_1, cs_cp1252,
cs_8859_15, cs_utf_8 };
typedef const char * entity_table_t;
/* codepage 1252 is a Windows extension to iso-8859-1. */
static entity_table_t ent_cp_1252[] = {
NULL, NULL, "sbquo", "fnof", "bdquo", "hellip", "dagger",
"Dagger", "circ", "permil", "Scaron", "lsaquo", "OElig",
NULL, NULL, NULL, NULL, "lsquo", "rsquo", "ldquo", "rdquo",
"bull", "ndash", "mdash", "tilde", "trade", "scaron", "rsaquo",
"oelig", NULL, NULL, "Yuml"
};
static entity_table_t ent_iso_8859_1[] = {
"nbsp", "iexcl", "cent", "pound", "curren", "yen", "brvbar",
"sect", "uml", "copy", "ordf", "laquo", "not", "shy", "reg",
"macr", "deg", "plusmn", "sup2", "sup3", "acute", "micro",
"para", "middot", "cedil", "sup1", "ordm", "raquo", "frac14",
"frac12", "frac34", "iquest", "Agrave", "Aacute", "Acirc",
"Atilde", "Auml", "Aring", "AElig", "Ccedil", "Egrave",
"Eacute", "Ecirc", "Euml", "Igrave", "Iacute", "Icirc",
"Iuml", "ETH", "Ntilde", "Ograve", "Oacute", "Ocirc", "Otilde",
"Ouml", "times", "Oslash", "Ugrave", "Uacute", "Ucirc", "Uuml",
"Yacute", "THORN", "szlig", "agrave", "aacute", "acirc",
"atilde", "auml", "aring", "aelig", "ccedil", "egrave",
"eacute", "ecirc", "euml", "igrave", "iacute", "icirc",
"iuml", "eth", "ntilde", "ograve", "oacute", "ocirc", "otilde",
"ouml", "divide", "oslash", "ugrave", "uacute", "ucirc",
"uuml", "yacute", "thorn", "yuml"
};
static entity_table_t ent_iso_8859_15[] = {
"nbsp", "iexcl", "cent", "pound", "euro", "yen", "Scaron",
"sect", "scaron", "copy", "ordf", "laquo", "not", "shy", "reg",
"macr", "deg", "plusmn", "sup2", "sup3", NULL, /* Zcaron */
"micro", "para", "middot", NULL, /* zcaron */ "sup1", "ordm",
"raquo", "OElig", "oelig", "Yuml", "iquest", "Agrave", "Aacute",
"Acirc", "Atilde", "Auml", "Aring", "AElig", "Ccedil", "Egrave",
"Eacute", "Ecirc", "Euml", "Igrave", "Iacute", "Icirc",
"Iuml", "ETH", "Ntilde", "Ograve", "Oacute", "Ocirc", "Otilde",
"Ouml", "times", "Oslash", "Ugrave", "Uacute", "Ucirc", "Uuml",
"Yacute", "THORN", "szlig", "agrave", "aacute", "acirc",
"atilde", "auml", "aring", "aelig", "ccedil", "egrave",
"eacute", "ecirc", "euml", "igrave", "iacute", "icirc",
"iuml", "eth", "ntilde", "ograve", "oacute", "ocirc", "otilde",
"ouml", "divide", "oslash", "ugrave", "uacute", "ucirc",
"uuml", "yacute", "thorn", "yuml"
};
struct html_entity_map {
enum entity_charset charset; /* charset identifier */
unsigned short basechar; /* char code at start of table */
unsigned short endchar; /* last char code in the table */
entity_table_t * table; /* the table of mappings */
};
static const struct html_entity_map entity_map[] = {
{ cs_cp1252, 0x80, 0x9f, ent_cp_1252 },
{ cs_cp1252, 0xa0, 0xff, ent_iso_8859_1 },
{ cs_8859_1, 0xa0, 0xff, ent_iso_8859_1 },
{ cs_8859_15, 0xa0, 0xff, ent_iso_8859_15 },
{ cs_utf_8, 0xa0, 0xff, ent_iso_8859_1 },
{ cs_terminator }
};
static const struct {
const char * codeset;
enum entity_charset charset;
} charset_map[] = {
{ "ISO-8859-1", cs_8859_1 },
{ "ISO-8859-15", cs_8859_15 },
{ "utf-8", cs_utf_8 },
{ "cp1252", cs_cp1252 },
{ NULL }
};
/* {{{ get_next_char
*/
inline static unsigned short get_next_char(enum entity_charset charset,
unsigned char * str,
int * newpos,
unsigned char * mbseq,
int * mbseqlen
)
{
int pos = *newpos;
int mbpos = 0;
unsigned short this_char = str[pos++];
mbseq[mbpos++] = (unsigned char)this_char;
if (charset == cs_utf_8) {
unsigned long utf = 0;
int stat = 0;
int more = 1;
/* unpack utf-8 encoding into a wide char.
* Code stolen from the mbstring extension */
do {
if (this_char < 0x80) {
more = 0;
break;
}
else if (this_char < 0xc0) {
switch(stat) {
case 0x10: /* 2, 2nd */
case 0x21: /* 3, 3rd */
case 0x32: /* 4, 4th */
case 0x43: /* 5, 5th */
case 0x54: /* 6, 6th */
/* last byte in sequence */
more = 0;
utf |= (this_char & 0x3f);
this_char = (unsigned short)utf;
break;
case 0x20: /* 3, 2nd */
case 0x31: /* 4, 3rd */
case 0x42: /* 5, 4th */
case 0x53: /* 6, 5th */
/* penultimate char */
utf |= ((this_char & 0x3f) << 6);
stat++;
break;
case 0x30: /* 4, 2nd */
case 0x41: /* 5, 3rd */
case 0x52: /* 6, 4th */
utf |= ((this_char & 0x3f) << 12);
stat++;
break;
case 0x40: /* 5, 2nd */
case 0x51:
utf |= ((this_char & 0x3f) << 18);
stat++;
break;
case 0x50: /* 6, 2nd */
utf |= ((this_char & 0x3f) << 24);
stat++;
default:
/* invalid */
more = 0;
}
}
/* lead byte */
else if (this_char < 0xe0) {
stat = 0x10; /* 2 byte */
utf = (this_char & 0x1f) << 6;
} else if (this_char < 0xf0) {
stat = 0x20; /* 3 byte */
utf = (this_char & 0xf) << 12;
} else if (this_char < 0xf8) {
stat = 0x30; /* 4 byte */
utf = (this_char & 0x7) << 18;
} else if (this_char < 0xfc) {
stat = 0x40; /* 5 byte */
utf = (this_char & 0x3) << 24;
} else if (this_char < 0xfe) {
stat = 0x50; /* 6 byte */
utf = (this_char & 0x1) << 30;
}
else {
/* invalid; bail */
more = 0;
break;
}
if (more)
{
this_char = str[pos++];
mbseq[mbpos++] = (unsigned char)this_char;
}
} while(more);
}
*newpos = pos;
mbseq[mbpos] = '\0';
*mbseqlen = mbpos;
return this_char;
}
/* }}} */
/* {{{ entity_charset determine_charset
* returns the charset identifier based on current locale or a hint.
* defaults to iso-8859-1 */
static enum entity_charset determine_charset(char * charset_hint)
{
int i;
enum entity_charset charset = cs_8859_1;
int len;
/* Guarantee default behaviour */
if (charset_hint == NULL)
return cs_8859_1;
if (strlen(charset_hint) == 0) {
/* try to detect the charset for the locale */
#if HAVE_NL_LANGINFO && HAVE_LOCALE_H && defined(CODESET)
charset_hint = nl_langinfo(CODESET);
#endif
#if HAVE_LOCALE_H
if (charset_hint == NULL)
{
/* try to figure out the charset from the locale */
char * localename;
char * dot, * at;
/* lang[_territory][.codeset][@modifier] */
localename = setlocale(LC_CTYPE, NULL);
dot = strchr(localename, '.');
if (dot) {
dot++;
/* locale specifies a codeset */
at = strchr(dot, '@');
if (at)
len = at - dot;
else
len = strlen(dot);
charset_hint = dot;
}
else {
/* no explicit name; see if the name itself
* is the charset */
charset_hint = localename;
len = strlen(charset_hint);
}
}
else
len = strlen(charset_hint);
#else
if (charset_hint)
len = strlen(charset_hint);
#endif
}
if (charset_hint) {
/* now walk the charset map and look for the codeset */
for (i = 0; charset_map[i].codeset; i++) {
if (strncasecmp(charset_hint, charset_map[i].codeset, len) == 0) {
charset = charset_map[i].charset;
break;
}
}
}
return charset;
}
/* }}} */
/* {{{ php_escape_html_entities
*/
PHPAPI char *php_escape_html_entities(unsigned char *old, int oldlen, int *newlen, int all, int quote_style, char * hint_charset)
{
int i, maxlen, len;
char *new;
enum entity_charset charset = determine_charset(hint_charset);
maxlen = 2 * oldlen;
if (maxlen < 128)
maxlen = 128;
new = emalloc (maxlen);
len = 0;
i = 0;
while (i < oldlen) {
int mbseqlen;
unsigned char mbsequence[16]; /* allow up to 15 characters
in a multibyte sequence
it should be more than enough.. */
unsigned short this_char = get_next_char(charset, old, &i, mbsequence, &mbseqlen);
int matches_map = 0;
if (len + 9 > maxlen)
new = erealloc (new, maxlen += 128);
if (all) {
/* look for a match in the maps for this charset */
int j;
unsigned char * rep;
for (j=0; entity_map[j].charset != cs_terminator; j++) {
if (entity_map[j].charset == charset
&& this_char >= entity_map[j].basechar
&& this_char <= entity_map[j].endchar)
{
rep = (unsigned char*)entity_map[j].table[this_char - entity_map[j].basechar];
if (rep == NULL) {
/* there is no entity for this position; fall through and
* just output the character itself */
break;
}
matches_map = 1;
break;
}
}
if (matches_map) {
new[len++] = '&';
strcpy(new + len, rep);
len += strlen(rep);
new[len++] = ';';
}
}
if (!matches_map) {
if (38 == this_char) {
memcpy (new + len, "&amp;", 5);
len += 5;
} else if (34 == this_char && !(quote_style&ENT_NOQUOTES)) {
memcpy (new + len, "&quot;", 6);
len += 6;
} else if (39 == this_char && (quote_style&ENT_QUOTES)) {
memcpy (new + len, "&#039;", 6);
len += 6;
} else if (60 == this_char) {
memcpy (new + len, "&lt;", 4);
len += 4;
} else if (62 == this_char) {
memcpy (new + len, "&gt;", 4);
len += 4;
} else if (this_char > 0xff) {
/* a wide char without a named entity; pass through the original sequence */
memcpy(new + len, mbsequence, mbseqlen);
len += mbseqlen;
} else {
new [len++] = (unsigned char)this_char;
}
}
}
new [len] = '\0';
*newlen = len;
return new;
}
/* }}} */
/* {{{ php_html_entities
*/
static void php_html_entities(INTERNAL_FUNCTION_PARAMETERS, int all)
{
zval **arg, **quotes, **charset;
int len, quote_style = ENT_COMPAT;
int ac = ZEND_NUM_ARGS();
char *hint_charset = NULL;
char *new;
if (ac < 1 || ac > 3 || zend_get_parameters_ex(ac, &arg, &quotes, &charset) == FAILURE) {
WRONG_PARAM_COUNT;
}
convert_to_string_ex(arg);
if(ac==2) {
convert_to_long_ex(quotes);
quote_style = (*quotes)->value.lval;
}
if (ac == 3) {
convert_to_string_ex(charset);
hint_charset = Z_STRVAL_PP(charset);
}
new = php_escape_html_entities((*arg)->value.str.val, (*arg)->value.str.len, &len, all, quote_style, hint_charset);
RETVAL_STRINGL(new, len, 0);
}
/* }}} */
#define HTML_SPECIALCHARS 0
#define HTML_ENTITIES 1
/* {{{ register_html_constants
*/
void register_html_constants(INIT_FUNC_ARGS)
{
REGISTER_LONG_CONSTANT("HTML_SPECIALCHARS", HTML_SPECIALCHARS, CONST_PERSISTENT|CONST_CS);
REGISTER_LONG_CONSTANT("HTML_ENTITIES", HTML_ENTITIES, CONST_PERSISTENT|CONST_CS);
REGISTER_LONG_CONSTANT("ENT_COMPAT", ENT_COMPAT, CONST_PERSISTENT|CONST_CS);
REGISTER_LONG_CONSTANT("ENT_QUOTES", ENT_QUOTES, CONST_PERSISTENT|CONST_CS);
REGISTER_LONG_CONSTANT("ENT_NOQUOTES", ENT_NOQUOTES, CONST_PERSISTENT|CONST_CS);
}
/* }}} */
/* {{{ proto string htmlspecialchars(string string [, int quote_style][, string charset])
Convert special characters to HTML entities */
PHP_FUNCTION(htmlspecialchars)
{
php_html_entities(INTERNAL_FUNCTION_PARAM_PASSTHRU, 0);
}
/* }}} */
/* {{{ proto string htmlentities(string string [, int quote_style][, string charset])
Convert all applicable characters to HTML entities */
PHP_FUNCTION(htmlentities)
{
php_html_entities(INTERNAL_FUNCTION_PARAM_PASSTHRU, 1);
}
/* }}} */
/* {{{ proto array get_html_translation_table([int table [, int quote_style][, string charset]])
Returns the internal translation table used by htmlspecialchars and htmlentities */
PHP_FUNCTION(get_html_translation_table)
{
zval **whichone, **quotes;
int which = HTML_SPECIALCHARS, quote_style = ENT_COMPAT;
int ac = ZEND_NUM_ARGS();
int i, j;
char ind[ 2 ];
enum entity_charset charset = determine_charset(NULL);
if (ac < 0 || ac > 2 || zend_get_parameters_ex(ac, &whichone, &quotes) == FAILURE) {
WRONG_PARAM_COUNT;
}
if (ac > 0) {
convert_to_long_ex(whichone);
which = (*whichone)->value.lval;
}
if (ac == 2) {
convert_to_long_ex(quotes);
quote_style = (*quotes)->value.lval;
}
array_init(return_value);
ind[1] = 0;
switch (which) {
case HTML_ENTITIES:
for (j=0; entity_map[j].charset != cs_terminator; j++) {
if (entity_map[j].charset != charset)
continue;
for (i = 0; i < entity_map[j].endchar - entity_map[j].basechar; i++)
{
char buffer[16];
if (entity_map[j].table[i] == NULL)
continue;
/* what about wide chars here ?? */
ind[0] = i + entity_map[j].basechar;
sprintf(buffer, "&%s;", entity_map[j].table[i]);
add_assoc_string(return_value, ind, buffer, 1);
}
}
/* break thru */
case HTML_SPECIALCHARS:
ind[0]=38; add_assoc_string(return_value, ind, "&amp;", 1);
if(quote_style&ENT_QUOTES) {
ind[0]=39; add_assoc_string(return_value, ind, "&#039;", 1);
}
if(!(quote_style&ENT_NOQUOTES)) {
ind[0]=34; add_assoc_string(return_value, ind, "&quot;", 1);
}
ind[0]=60; add_assoc_string(return_value, ind, "&lt;", 1);
ind[0]=62; add_assoc_string(return_value, ind, "&gt;", 1);
break;
}
}
/* }}} */
/*
* Local variables:
* tab-width: 4
* c-basic-offset: 4
* End:
* vim600: sw=4 ts=4 tw=78 fdm=marker
* vim<600: sw=4 ts=4 tw=78
*/