htparse.c

elinks下lynx是最重要的二个文本浏览器, 在linux下非常实用, lynx比elinks早的多, 目前好像停止开发, 这是lynx源代码

		    }
		}
	    }
#endif /* CLEAN_URLS */
	}
    }

    /*
     * Trim any blanks from the result so far - there's no excuse for blanks
     * in a hostname.  Also update the tail here.
     */
    tail = LYRemoveBlanks(result);

    /*
    **	If host in given or related was ended directly with a '?' (no
    **  slash), fake the search part into absolute.  This is the only
    **  case search is returned from scan.  A host must have been present.
    **  this restores the '?' at which the host part had been truncated in
    **  scan, we have to do this after host part handling is done. - kw
    */
    if (given.search && *(given.search - 1) == '\0') {
	given.absolute = given.search - 1;
	given.absolute[0] = '?';
    } else if (related.search && !related.absolute &&
	       *(related.search - 1) == '\0') {
	related.absolute = related.search - 1;
	related.absolute[0] = '?';
    }

    /*
    **	If different hosts, inherit no path.
    */
    if (given.host && related.host)
	if (strcmp(given.host, related.host) != 0) {
	    related.absolute = NULL;
	    related.relative = NULL;
	    related.anchor = NULL;
	}

    /*
    **	Handle the path.
    */
    if (wanted & (PARSE_PATH | PARSE_STRICTPATH | PARSE_QUERY)) {
	int want_detail = (wanted & (PARSE_STRICTPATH | PARSE_QUERY));

	if (acc_method && !given.absolute && given.relative) {
	    /*
	     * Treat all given nntp or snews paths, or given paths for news
	     * URLs with a host, as absolute.
	     */
	    switch (*acc_method) {
	    case 'N':
	    case 'n':
		if (!strcasecomp(acc_method, "nntp") ||
		    (!strcasecomp(acc_method, "news") &&
		     !strncasecomp(result, "news://", 7))) {
		    given.absolute = given.relative;
		    given.relative = NULL;
		}
		break;
	    case 'S':
	    case 's':
		if (!strcasecomp(acc_method, "snews")) {
		    given.absolute = given.relative;
		    given.relative = NULL;
		}
		break;
	    }
	}

	if (given.absolute) {			/* All is given */
	    if (wanted & PARSE_PUNCTUATION)
		*tail++ = '/';
	    strcpy(tail, given.absolute);
	    CTRACE((tfp, "HTParse: (ABS)\n"));
	} else if (related.absolute) {		/* Adopt path not name */
	    *tail++ = '/';
	    strcpy(tail, related.absolute);
	    if (given.relative) {
		p = strchr(tail, '?');	/* Search part? */
		if (p == NULL)
		    p = (tail + strlen(tail) - 1);
		for (; *p != '/'; p--)
		    ;				/* last / */
		p[1] = '\0';			/* Remove filename */
		strcat(p, given.relative); /* Add given one */
		HTSimplify (result);
	    }
	    CTRACE((tfp, "HTParse: (Related-ABS)\n"));
	} else if (given.relative) {
	    strcpy(tail, given.relative);		/* what we've got */
	    CTRACE((tfp, "HTParse: (REL)\n"));
	} else if (related.relative) {
	    strcpy(tail, related.relative);
	    CTRACE((tfp, "HTParse: (Related-REL)\n"));
	} else {  /* No inheritance */
	    if (!isLYNXCGI(aName) &&
		!isLYNXEXEC(aName) &&
		!isLYNXPROG(aName)) {
		*tail++ = '/';
		*tail = '\0';
	    }
	    if (!strcmp(result, "news:/"))
		result[5] = '*';
	    CTRACE((tfp, "HTParse: (No inheritance)\n"));
	}
	if (want_detail) {
	    p = strchr(tail, '?');	/* Search part? */
	    if (p) {
		if (PARSE_STRICTPATH) {
		    *p = '\0';
		} else {
		    if (!(wanted & PARSE_PUNCTUATION))
			p++;
		    do {
			*tail++ = *p;
		    } while (*p++);
		}
	    } else {
		if (wanted & PARSE_QUERY)
		    *tail = '\0';
	    }
	}
    }

    /*
    **	Handle the fragment (anchor). Never inherit.
    */
    if (wanted & PARSE_ANCHOR) {
	if (given.anchor && *given.anchor) {
	    tail += strlen(tail);
	    if (wanted & PARSE_PUNCTUATION)
		*tail++ = '#';
	    strcpy(tail, given.anchor);
	}
    }

    /*
     * If there are any blanks remaining in the string, escape them as needed.
     * See the discussion in LYLegitimizeHREF() for example.
     */
    if ((p = strchr(result, ' ')) != 0) {
	switch (is_url(result)) {
	case UNKNOWN_URL_TYPE:
	    CTRACE((tfp, "HTParse:      ignore:`%s'\n", result));
	    break;
	case LYNXEXEC_URL_TYPE:
	case LYNXPROG_URL_TYPE:
	case LYNXCGI_URL_TYPE:
	case LYNXPRINT_URL_TYPE:
	case LYNXHIST_URL_TYPE:
	case LYNXDOWNLOAD_URL_TYPE:
	case LYNXKEYMAP_URL_TYPE:
	case LYNXIMGMAP_URL_TYPE:
	case LYNXCOOKIE_URL_TYPE:
	case LYNXDIRED_URL_TYPE:
	case LYNXOPTIONS_URL_TYPE:
	case LYNXCFG_URL_TYPE:
	case LYNXCOMPILE_OPTS_URL_TYPE:
	case LYNXMESSAGES_URL_TYPE:
	    CTRACE((tfp, "HTParse:      spaces:`%s'\n", result));
	    break;
	case NOT_A_URL_TYPE:
	default:
	    CTRACE((tfp, "HTParse:      encode:`%s'\n", result));
	    do {
		char *q = p + strlen(p) + 2;
		while (q != p + 1) {
		    q[0] = q[-2];
		    --q;
		}
		p[0] = '%';
		p[1] = '2';
		p[2] = '0';
	    } while ((p = strchr(result, ' ')) != 0);
	    break;
	}
    }
    CTRACE((tfp, "HTParse:      result:`%s'\n", result));

    StrAllocCopy(return_value, result);
    LYalloca_free(result);

    /* FIXME: could be optimized using HTParse() internals */
    if (*relatedName &&
	((wanted & PARSE_ALL_WITHOUT_ANCHOR) == PARSE_ALL_WITHOUT_ANCHOR)) {
	/*
	 *  Check whether to fill in localhost. - FM
	 */
	LYFillLocalFileURL(&return_value, relatedName);
	CTRACE((tfp, "pass LYFillLocalFile:`%s'\n", return_value));
    }

    return return_value;		/* exactly the right length */
}

/*	HTParseAnchor(), fast HTParse() specialization
**	----------------------------------------------
**
** On exit,
**	returns		A pointer within input string (probably to its end '\0')
*/
PUBLIC CONST char * HTParseAnchor ARGS1(
	CONST char *,	aName)
{
    CONST char* p = aName;
    for ( ; *p && *p != '#'; p++)
	;
    if (*p == '#') {
	/* the safe way based on HTParse() -
	 * keeping in mind scan() peculiarities on schemes:
	 */
	struct struct_parts given;

	char* name = (char*)LYalloca((p - aName) + strlen(p) + 1);
	if (name == NULL) {
	    outofmem(__FILE__, "HTParseAnchor");
	}
	strcpy(name, aName);
	scan(name, &given);
	LYalloca_free(name);

	p++; /*next to '#'*/
	if (given.anchor == NULL) {
	    for ( ; *p; p++)  /*scroll to end '\0'*/
		;
	}
    }
    return p;
}

/*	Simplify a filename.				HTSimplify()
**	--------------------
**
**  A unix-style file is allowed to contain the sequence xxx/../ which may
**  be replaced by "" , and the sequence "/./" which may be replaced by "/".
**  Simplification helps us recognize duplicate filenames.
**
**	Thus,	/etc/junk/../fred	becomes /etc/fred
**		/etc/junk/./fred	becomes /etc/junk/fred
**
**	but we should NOT change
**		http://fred.xxx.edu/../..
**
**	or	../../albert.html
*/
PUBLIC void HTSimplify ARGS1(
	char *,		filename)
{
    char *p;
    char *q, *q1;

    if (filename == NULL)
	return;

    if (!(filename[0] && filename[1]) ||
	filename[0] == '?' || filename[1] == '?' || filename[2] == '?')
	return;

    if (strchr(filename, '/') != NULL) {
	for (p = (filename + 2); *p; p++) {
	    if (*p == '?') {
		/*
		**  We're still treating a ?searchpart as part of
		**  the path in HTParse() and scan(), but if we
		**  encounter a '?' here, assume it's the delimiter
		**  and break.	We also could check for a parameter
		**  delimiter (';') here, but the current Fielding
		**  draft (wisely or ill-advisedly :) says that it
		**  should be ignored and collapsing be allowed in
		**  it's value).  The only defined parameter at
		**  present is ;type=[A, I, or D] for ftp URLs, so
		**  if there's a "/..", "/../", "/./", or terminal
		**  '.' following the ';', it must be due to the
		**  ';' being an unescaped path character and not
		**  actually a parameter delimiter. - FM
		*/
		break;
	    }
	    if (*p == '/') {
		if ((p[1] == '.') && (p[2] == '.') &&
		    (p[3] == '/' || p[3] == '?' || p[3] == '\0')) {
		    /*
		    **	Handle "../", "..?" or "..".
		    */
		    for (q = (p - 1); (q >= filename) && (*q != '/'); q--)
			/*
			**  Back up to previous slash or beginning of string.
			*/
			;
		    if ((q[0] == '/') &&
			(strncmp(q, "/../", 4) &&
			 strncmp(q, "/..?", 4)) &&
			!((q - 1) > filename && q[-1] == '/')) {
			/*
			**  Not at beginning of string or in a
			**  host field, so remove the "/xxx/..".
			*/
			q1 = (p + 3);
			p = q;
			while (*q1 != '\0')
			    *p++ = *q1++;
			*p = '\0';		/* terminate */
			/*
			**  Start again with previous slash.
			*/
			p = (q - 1);
		    }
		} else if (p[1] == '.' && p[2] == '/') {
		    /*
		    **	Handle "./" by removing both characters.
		    */
		    q = p;
		    q1 = (p + 2);
		    while (*q1 != '\0')
			*q++ = *q1++;
		    *q = '\0';		/* terminate */
		    p--;
		} else if (p[1] == '.' && p[2] == '?') {
		    /*
		    **	Handle ".?" by removing the dot.
		    */
		    q = (p + 1);
		    q1 = (p + 2);
		    while (*q1 != '\0')
			*q++ = *q1++;
		    *q = '\0';		/* terminate */
		    p--;
		} else if (p[1] == '.' && p[2] == '\0') {
		    /*
		    **	Handle terminal "." by removing the character.
		    */
		    p[1] = '\0';
		}
	    }
	}
	if (p >= filename + 2 && *p == '?' && *(p-1)  == '.') {
	    if (*(p-2) == '/') {
		/*
		**  Handle "/.?" by removing the dot.
		*/
		q = p - 1;
		q1 = p;
		while (*q1 != '\0')
		    *q++ = *q1++;
		*q = '\0';
	    } else if (*(p-2) == '.' &&
		       p >= filename + 4 && *(p-3) == '/' &&
		       (*(p-4) != '/' ||
			(p > filename + 4 && *(p-5) != ':'))) {
		    /*
		    **	Handle "xxx/..?"
		    */
		for (q = (p - 4); (q > filename) && (*q != '/'); q--)
			/*
			**  Back up to previous slash or beginning of string.
			*/
		    ;
		if (*q == '/') {
		    if (q > filename && *(q-1) == '/' &&
			!(q > filename + 1 && *(q-1) != ':'))
			return;
		    q++;
		}
		if (strncmp(q, "../", 3) && strncmp(q, "./", 2)) {
			/*
			**  Not after "//" at beginning of string or
			**  after "://", and xxx is not ".." or ".",
			**  so remove the "xxx/..".
			*/
		    q1 = p;
		    p = q;
		    while (*q1 != '\0')
			*p++ = *q1++;
		    *p = '\0';		/* terminate */
		}
	    }
	}