parse.c

elinks下lynx是最重要的二个文本浏览器, 在linux下非常实用, elinks也是gentoo安装过程中默认使用的浏览器, 这是elinks源代码

/* HTML core parser routines */

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#include <errno.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#include "elinks.h"

#include "document/css/apply.h"
#include "document/css/parser.h"
#include "document/html/parser/forms.h"
#include "document/html/parser/general.h"
#include "document/html/parser/link.h"
#include "document/html/parser/parse.h"
#include "document/html/parser/stack.h"
#include "document/html/parser.h"
#include "document/options.h"
#include "intl/charsets.h"
#include "util/conv.h"
#include "util/error.h"
#include "util/fastfind.h"
#include "util/memdebug.h"
#include "util/memory.h"
#include "util/string.h"

/* Unsafe macros */
#include "document/html/internal.h"


#define end_of_tag(c) ((c) == '>' || (c) == '<')

static inline int
atchr(register unsigned char c)
{
	return (c < 127 && (c > '>' || (c > ' ' && c != '=' && !end_of_tag(c))));
}

/* This function eats one html element. */
/* - e is pointer to the begining of the element (*e must be '<')
 * - eof is pointer to the end of scanned area
 * - parsed element name is stored in name, it's length is namelen
 * - first attribute is stored in attr
 * - end points to first character behind the html element */
/* It returns -1 when it failed (returned values in pointers are invalid) and
 * 0 for success. */
int
parse_element(register unsigned char *e, unsigned char *eof,
	      unsigned char **name, int *namelen,
	      unsigned char **attr, unsigned char **end)
{
#define next_char() if (++e == eof) return -1;

	assert(e && eof);
	if (e >= eof || *e != '<') return -1;

	next_char();
	if (name) *name = e;

	if (*e == '/') next_char();
	if (!isident(*e)) return -1;

	while (isident(*e)) next_char();

	if (!isspace(*e) && !end_of_tag(*e) && *e != '/' && *e != ':' && *e != '=')
		return -1;

	if (name && namelen) *namelen = e - *name;

	while (isspace(*e) || *e == '/' || *e == ':') next_char();

	/* Skip bad attribute */
	while (!atchr(*e) && !end_of_tag(*e) && !isspace(*e)) next_char();

	if (attr) *attr = e;

next_attr:
	while (isspace(*e)) next_char();

	/* Skip bad attribute */
	while (!atchr(*e) && !end_of_tag(*e) && !isspace(*e)) next_char();

	if (end_of_tag(*e)) goto end;

	while (atchr(*e)) next_char();
	while (isspace(*e)) next_char();

	if (*e != '=') {
		if (end_of_tag(*e)) goto end;
		goto next_attr;
	}
	next_char();

	while (isspace(*e)) next_char();

	if (isquote(*e)) {
		unsigned char quote = *e;

/* quoted_value: */
		next_char();
		while (*e != quote) next_char();
		next_char();
		/* The following apparently handles the case of <foo
		 * id="a""b">, however that is very rare and probably not
		 * conforming. More frequent (and mishandling it more fatal) is
		 * probably the typo of <foo id="a""> - we can handle it as
		 * long as this is commented out. --pasky */
		/* if (*e == quote) goto quoted_value; */
	} else {
		while (!isspace(*e) && !end_of_tag(*e)) next_char();
	}

	while (isspace(*e)) next_char();

	if (!end_of_tag(*e)) goto next_attr;

end:
	if (end) *end = e + (*e == '>');

	return 0;
}


#define realloc_chrs(x, l) mem_align_alloc(x, l, (l) + 1, unsigned char, 0xFF)

#define add_chr(s, l, c)						\
	do {								\
		if (!realloc_chrs(&(s), l)) return NULL;		\
		(s)[(l)++] = (c);					\
	} while (0)

unsigned char *
get_attr_value(register unsigned char *e, unsigned char *name,
	       struct document_options *options, enum html_attr_flags flags)
{
	unsigned char *n;
	unsigned char *name_start;
	unsigned char *attr = NULL;
	int attrlen = 0;
	int found;

next_attr:
	skip_space(e);
	if (end_of_tag(*e) || !atchr(*e)) goto parse_error;
	n = name;
	name_start = e;

	while (atchr(*n) && atchr(*e) && toupper(*e) == toupper(*n)) e++, n++;
	found = !*n && !atchr(*e);

	if (found && (flags & HTML_ATTR_TEST)) return name_start;

	while (atchr(*e)) e++;
	skip_space(e);
	if (*e != '=') {
		if (found) goto found_endattr;
		goto next_attr;
	}
	e++;
	skip_space(e);

	if (found) {
		if (!isquote(*e)) {
			while (!isspace(*e) && !end_of_tag(*e)) {
				if (!*e) goto parse_error;
				add_chr(attr, attrlen, *e);
				e++;
			}
		} else {
			unsigned char quote = *e;

/* parse_quoted_value: */
			while (*(++e) != quote) {
				if (*e == ASCII_CR) continue;
				if (!*e) goto parse_error;
				if (*e != ASCII_TAB && *e != ASCII_LF)
					add_chr(attr, attrlen, *e);
				else if (!(flags & HTML_ATTR_EAT_NL))
					add_chr(attr, attrlen, ' ');
			}
			e++;
			/* The following apparently handles the case of <foo
			 * id="a""b">, however that is very rare and probably
			 * not conforming. More frequent (and mishandling it
			 * more fatal) is probably the typo of <foo id="a""> -
			 * we can handle it as long as this is commented out.
			 * --pasky */
#if 0
			if (*e == quote) {
				add_chr(attr, attrlen, *e);
				goto parse_quoted_value;
			}
#endif
		}

found_endattr:
		add_chr(attr, attrlen, '\0');
		attrlen--;

		if (/* Unused: !(flags & HTML_ATTR_NO_CONV) && */
		    memchr(attr, '&', attrlen)) {
			unsigned char *saved_attr = attr;

			attr = convert_string(NULL, saved_attr, attrlen,
			                      options->cp, CSM_QUERY,
			                      NULL, NULL, NULL);
			mem_free(saved_attr);
		}

		set_mem_comment(trim_chars(attr, ' ', NULL), name, strlen(name));
		return attr;

	} else {
		if (!isquote(*e)) {
			while (!isspace(*e) && !end_of_tag(*e)) {
				if (!*e) goto parse_error;
				e++;
			}
		} else {
			unsigned char quote = *e;

			do {
				while (*(++e) != quote)
					if (!*e) goto parse_error;
				e++;
			} while (/* See above. *e == quote */ 0);
		}
	}

	goto next_attr;

parse_error:
	mem_free_if(attr);
	return NULL;
}

#undef add_chr


/* Extract numerical value of attribute @name.
 * It will return a positive integer value on success,
 * or -1 on error. */
int
get_num(unsigned char *a, unsigned char *name, struct document_options *options)
{
	unsigned char *al = get_attr_val(a, name, options);
	int result = -1;

	if (al) {
		unsigned char *end;
		long num;

		errno = 0;
		num = strtol(al, (char **) &end, 10);
		if (!errno && *al && !*end && num >= 0 && num <= INT_MAX)
			result = (int) num;

		mem_free(al);
	}

	return result;
}

/* Parse 'width[%],....'-like attribute @name of element @a.  If @limited is
 * set, it will limit the width value to the current usable width. Note that
 * @limited must be set to be able to parse percentage widths. */
/* The function returns width in characters or -1 in case of error. */
int
get_width(unsigned char *a, unsigned char *name, int limited,
          struct html_context *html_context)
{
	unsigned char *value = get_attr_val(a, name, html_context->options);
	unsigned char *str = value;
	unsigned char *end;
	int percentage = 0;
	int len;
	long width;

	if (!value) return -1;

	/* Skip spaces at start of string if any. */
	skip_space(str);

	/* Search for end of string or ',' character (ie. in "100,200") */
	for (len = 0; str[len] && str[len] != ','; len++);

	/* Go back, and skip spaces after width if any. */
	while (len && isspace(str[len - 1])) len--;
	if (!len) { mem_free(value); return -1; } /* Nothing to parse. */

	/* Is this a percentage ? */
	if (str[len - 1] == '%') len--, percentage = 1;

	/* Skip spaces between width number and percentage if any. */
	while (len && isspace(str[len - 1])) len--;
	if (!len) { mem_free(value); return -1; } /* Nothing to parse. */

	/* Shorten the string a bit, so strtoul() will work on useful
	 * part of it. */
	str[len] = '\0';

	/* Convert to number if possible. */
	errno = 0;
	width = strtoul((char *) str, (char **) &end, 10);

	/* @end points into the @value string so check @end position
	 * before freeing @value. */
	if (errno || *end || width >= INT_MAX) {
		/* Not a valid number. */
		mem_free(value);
		return -1;
	}

	mem_free(value);

#define WIDTH_PIXELS2CHARS(width) ((width) + (HTML_CHAR_WIDTH - 1) / 2) / HTML_CHAR_WIDTH;

	if (limited) {
		int maxwidth = get_html_max_width();

		if (percentage) {
			/* Value is a percentage. */
			width = width * maxwidth / 100;
		} else {
			/* Value is a number of pixels, makes an approximation. */
			width = WIDTH_PIXELS2CHARS(width);
		}

		if (width > maxwidth)
			width = maxwidth;

	} else {
		if (percentage) {
			/* No sense, we need @limited and @maxwidth for percentage. */
			return -1;
		} else {
			/* Value is a number of pixels, makes an approximation,
			 * no limit here */
			width = WIDTH_PIXELS2CHARS(width);
		}
	}

#undef WIDTH_PIXELS2CHARS

	if (width < 0)
		width = 0;

	return width;
}