/* url.c */

#include "elvis.h"
#ifdef FEATURE_RCSID
char id_url[] = "$Id: url.c,v 2.24 2003/10/18 17:01:29 steve Exp $";
#endif

static char	url_protocol[10];
static char	url_site_port[50];
static char	url_resource[1000];

#if USE_PROTOTYPES
static ELVBOOL parseurl(char *url);
# if defined(PROTOCOL_HTTP) || defined(PROTOCOL_FTP)
static char *findproxy(ELVBOOL never_direct);
# endif
#endif

#ifdef FEATURE_PROTO
/* This indicates whether we're current running a URL protocol alias.  This is
 * important because URL protocol aliases such as readMAILTO aren't allowed
 * to switch buffers.
 */
ELVBOOL urlprotobusy;
#endif

/* Parse an URL, and return ElvTrue if successful. */
static ELVBOOL parseurl(url)
	char	*url;
{
	int	i;

	/* Copy the protocol into url_protocol[], converting to lowercase */
	for (i = 0; i < QTY(url_protocol) - 1 && elvalpha(url[i]); i++)
	{
		url_protocol[i] = elvtolower(url[i]);
	}
	url_protocol[i] = '\0';

	/* We expect a colon after the protocol.  If we don't have that, then
	 * fail.  Also fail if the protocol is less than two characters long,
	 * so we don't mistake the "C:" drive letter for a protocol.
	 */
	if (url[i++] != ':' || i < 3)
	{
		/* Directories are treated like "dir:" protocol */
		if (dirperm(url) == DIR_DIRECTORY)
		{
			strcpy(url_protocol, "dir");
			*url_site_port = '\0';
			strcpy(url_resource, url);
			return ElvTrue;
		}

		/* else non-file or a normal file */
		return ElvFalse;
	}

	/* Check for a host.  Not all protocols require one. */
	url += i;
	url_site_port[0] = '\0';
	if ((url[0] == '/' || url[0] == '\\') && url[1] == url[0])
	{
		url += 2;
		for (i = 0; i < QTY(url_site_port) - 1 && *url && *url != '/' && *url != '\\'; i++, url++)
		{
			url_site_port[i] = elvtolower(*url);
		}
		url_site_port[i] = '\0';
	}

	/* The rest of the URL is assumed to be a resource name.  If it is a
	 * null string, then assume it should be "/".  Convert backslashes to
	 * forward slashes.
	 */
	if (!*url)
		strcpy(url_resource, "/");
	else
	{
		for (i = 0; i < QTY(url_resource) - 1 && *url; url++, i++)
			if (*url == '\\')
				url_resource[i] = '/';
			else
				url_resource[i] = *url;
		url_resource[i] = '\0';
	}

	/* some debugging code */
	if (o_verbose >= 4)
		msg(MSG_INFO, "[sss]protocol=$1, site_port=$2, resource=$3",
			url_protocol, url_site_port, url_resource);

	return ElvTrue;
}

/* If the URL refers to a local file, this returns the name of the
 * file.  This handles "file:" protocol and names which have no
 * protocol; for all others it returns NULL.  It is also smart enough
 * to realize that "C:" is a drive letter rather than a protocol.
 */
char *urllocal(url)
	char	*url;
{
	if (!parseurl(url))
		return url;
	else if (!strcmp(url_protocol, "file"))
		return url + 5;
	return NULL;
}

/* Check the permissions on a URL. */
DIRPERM urlperm(url)
	char	*url;
{
	/* "-" is used for stdio -- not an actual filename */
	if (!strcmp(url, "-"))
		return DIR_READWRITE;

#if defined(PROTOCOL_HTTP) || defined(PROTOCOL_FTP)
	/* Verify that this is a remote URL, other than mailto:.  This also
	 * has the side-effect of parsing the URL.  For local URLs try using
	 * the dirperm() function.
	 */
	if (!urlremote(url))
		return dirperm(url);

# ifdef PROTOCOL_FTP
	/* If the protocol is anything other than "ftp" then assume the URL
	 * exists and is readonly.
	 */
	if (strcmp(url_protocol, "ftp"))
		return DIR_READONLY;

	/* Else use FTP to inspect the file (not easy!) and then return
	 * whatever permissions it found.
	 */
	(void)ftpopen(url_site_port, url_resource, ElvFalse, 'p');
	return ftpperms;
# else /* PROTOCOL_HTTP but not PROTOCOL_FTP */
	/* it must be HTTP, which is readonly */
	return DIR_READONLY;
# endif

#else /* no network protocols */

	return dirperm(url);
#endif
}


#ifdef FEATURE_PROTO

/* If this is a URL and its protocol is the same as a readXXX or writeXXX
 * alias, then run the alias and return RESULT_COMPLETE or RESULT_ERROR
 * to indicate the result.  Otherwise return RESULT_MORE.
 */
RESULT urlalias(from, to, url)
	MARK	from;	/* starting mark */
	MARK	to;	/* ending mark when writing, or NULL when reading */
	char	*url;	/* URL to read */
{
	char	name[20];/* name of a possible alias */
	EXINFO	xinf;	/* a constructed ex command line */
	RESULT	result;
	int	i, j;

	/* try to extract the protocol */
	if (!parseurl(url))
		return RESULT_MORE;

	/* is there an alias with this name? */
	strcpy(name, to ? "write" : "read");
	for (i=0, j=strlen(name); (name[j++] = elvtoupper(url_protocol[i++])); )
	{
	}
	memset(&xinf, 0, sizeof xinf);
	xinf.cmdname = exisalias(name, ElvFalse);
	if (!xinf.cmdname)
		return RESULT_MORE;

	/* turn on the buffer's userprotocol option */
	o_userprotocol(markbuffer(from)) = ElvTrue;

	/* construct an ex command, and execute it */
	xinf.command = EX_DOPROTO;
	xinf.window = windefault;
	xinf.defaddr = *from;
	xinf.fromaddr = from;
	xinf.toaddr = to ? to : from;
	xinf.from = markline(from);
	xinf.to = markline(xinf.toaddr) - 1;
	xinf.anyaddr = (ELVBOOL)(markoffset(xinf.toaddr) > 0);
	(void)buildstr(&xinf.rhs, url);
	urlprotobusy = ElvTrue;
	result = ex_doalias(&xinf);
	urlprotobusy = ElvFalse;
	safefree(xinf.rhs);

	/* return the result */
	return result;
}

#endif /* FEATURE_PROTO */


#if defined(PROTOCOL_HTTP) || defined(PROTOCOL_FTP)

/* Return ElvTrue if the url uses a protocol other than "mailto:" or
 * "file:".
 */
ELVBOOL urlremote(url)
	char	*url;
{
	/* Parse the url.  If parsing fails, then it isn't remote.  Also,
	 * check for a few specific protocols which aren't supported.
	 */
	if (!parseurl(url)
	 || !strcmp(url_protocol, "mailto")
	 || !strcmp(url_protocol, "file"))
		return ElvFalse;
	return ElvTrue;
}

static long	totalbytes;
static long	expectbytes;
static long	oldpollfreq;
#ifdef PROTOCOL_FTP
static ELVBOOL	useftp;
#endif

/* Scan through the elvis.net file for this URL's domain.  Return the proxy
 * site's name if found, or NULL if the URL should be accessed directly.
 * The proxy name will be in a dynamically-allocated buffer, which must be
 * freed by the calling function.
 *
 * This assumes that parseurl() was recently called on the URL, so the site
 * name can be found in url_site_port[].
 */
static char *findproxy(never_direct)
	ELVBOOL	never_direct;	/* protocol isn't supported directly by elvis */
{
	CHAR	*word, *proxy;	/* words from file */
	int	sitelen, len;	/* lengths of site name and current word */
	FILE	*fp;		/* stream used for reading from file */
	int	ch;		/* character from file */
	ELVBOOL	expect_proxy;	/* is next word expected to be proxy name? */
	char	*tmp;

	/* compute the length of the site name, excluding the port number */
	tmp = strchr(url_site_port, ':');
	if (tmp)
		sitelen = (int)(tmp - url_site_port);
	else
		sitelen = strlen(url_site_port);

	/* find the "elvis.net" file (actually the buffer containing it) */
	tmp = iopath(tochar8(o_elvispath), NET_FILE, ElvFalse);
	if (!tmp)
		return NULL;

	/* scan for the domain in the "elvis.net" file */
	fp = fopen(tmp, "r");
	proxy = word = NULL;
	expect_proxy = ElvFalse;
	if (fp != NULL)
	{
		while ((ch = getc(fp)) != EOF)
		{
			/* If '#' is encountered, then skip ahead to the next
			 * newline.  Later code will then see the newline and
			 * hence treat the entire comment like whitespace.
			 */
			if (ch == '#')
			{
				do
				{
					if ((ch = getc(fp)) == EOF)
						goto End;
				} while (ch != '\n');
			}

			/* non-whitespace just adds the character to the current word */
			if (!elvspace(ch))
			{
				buildCHAR(&word, ch);
				continue;
			}

			/* if we have no word (multiple spaces) then continue */
			if (!word)
				continue;

			/* if this word is supposed to be a proxy name, then use it */
			if (expect_proxy)
			{
				expect_proxy = ElvFalse;
				assert(!proxy);
				proxy = word;
				word = NULL;
				continue;
			}

			/* Handle the "direct" keyword */
			if (!strcmp(tochar8(word), "direct"))
			{
				/* forget the current proxy */
				if (proxy && !never_direct)
				{
					safefree(proxy);
					proxy = NULL;
				}
				goto NextWord;
			}

			/* handle the "proxy" keyword */
			if (!strcmp(tochar8(word), "proxy"))
			{
				if (proxy)
				{
					safefree(proxy);
					proxy = NULL;
				}
				expect_proxy = ElvTrue;
				goto NextWord;
			}

			/* Then it must be a domain name.  Is the the URL's? */
			len = CHARlen(word);
			if ((len < sitelen && *word == '.' && !strncmp(&url_site_port[len - sitelen], tochar8(word), len))
			 || (len == sitelen && !strncmp(url_site_port, tochar8(word), len)))
			{
				/* Found it! Stop looking! */
				break;
			}

			/* otherwise we can forget the current word */
	NextWord:
			safefree(word);
			word = NULL;
		}
		fclose(fp);
	}

	/* if there was a last word, free it */
End:
	if (word)
		safefree(word);

	/* return the proxy (which may be NULL) */
	return tochar8(proxy);
}


/* Open a remote URL for reading or writing.  Returns ElvTrue if successful.  If
 * error, it gives an error message and returns ElvFalse.
 */
ELVBOOL urlopen(url, force, rwa)
	char	*url;
	ELVBOOL	force;
	_char_	rwa;
{
	char	*proxy;
	ELVBOOL	unsupported;
	ELVBOOL	reading;
	ELVBOOL	retval;

	/* reset the expectbytes/totalbytes values to an impossible value */
	expectbytes = -1L;
	totalbytes = 0;

	/* Verify that this is a remote URL, other than mailto:.  This also
	 * has the side-effect of parsing the URL.
	 */
	if (!urlremote(url))
	{
		msg(MSG_ERROR, "[s]$1 not a remote url", url);
		return ElvFalse;
	}

	/* set "unsupported" if this isn't a built-in protocol */
	unsupported = (ELVBOOL)!(
#ifdef PROTOCOL_FTP
		!strcmp(url_protocol, "ftp") ||
#endif
		!strcmp(url_protocol, "http"));

	/* If a proxy is specified in "elvis.net", then use it.  Otherwise
	 * we'll access the site/port directly.
	 */
	proxy = findproxy(unsupported);
	if (proxy)
	{
		/* Use the proxy.  Assume it uses the HTTP protocol */
		strcpy(url_protocol, "http");
		strcpy(url_site_port, proxy);
		strcpy(url_resource, url);

		/* Free the proxy name, but don't zero the proxy pointer.
		 * Later, we can still compare proxt against NULL, to detect
		 * whether we're using a proxy.
		 */
		safefree(proxy);
	}
	else
	{
		/* We'll access it directly... unless we don't know how! */
		if (unsupported)
		{
			msg(MSG_ERROR, "[s]unsupported protocol $1", url_protocol);
			return ElvFalse;
		}
	}

	/* Will we be reading or writing? */
	reading = (ELVBOOL)(rwa == 'r');

	/* arrange for frequent polling during the network operation */
	oldpollfreq = o_pollfrequency;
	o_pollfrequency = 1L;
	(void)guipoll(ElvTrue);

	/* At this point, we know we have a supported protocol.  Which one? */
#ifdef PROTOCOL_FTP
	useftp = (ELVBOOL)!strcmp(url_protocol, "ftp");
	if (useftp)
		retval = ftpopen(url_site_port, url_resource, force, rwa);
	else
#endif
	{
		if (!reading)
		{
			msg(MSG_ERROR, "[s]can only READ from http $1", proxy ? "proxy" : "server");
			retval = ElvFalse;
		}
		else
			retval = httpopen(url_site_port, url_resource);
	}

	/* restore pollfrequency */
	o_pollfrequency = oldpollfreq;
	return retval;
}

/* During a "write" operation, this value will be displayed to the
 * user as the total number of bytes to be transferred.  You should
 * call urlbytes() after calling urlopen() with 'w' or 'a' as the
 * rwa parameter.  Alternatively, you may omit calling this function
 * in which case elvis simply won't display the total number of bytes
 * to the user.  When rwa is 'r', this function does nothing.
 *
 * Is is also harmless to call urlbytes() after other flavors of ioopen()
 * for writing; i.e., you can safely call urlbytes() after any ioopen() which
 * has an rwa argument of anything except 'r'.  The httpopen() and ftpopen()
 * functions call urlbytes() for 'r' operations.
 */
void urlbytes(totbytes)
	long	totbytes;	/* number of bytes we expect to write */
{
	expectbytes = totbytes;
}

/* Read data from an URL.  Returns the number of bytes read (0 at EOF);
 * if error or user-abort, it given an error message and returns -1.
 */
int urlread(buf, bytes)
	CHAR	*buf;
	int	bytes;
{
#ifdef PROTOCOL_FTP
	if (useftp)
		bytes = ftpread(buf, bytes);
	else
#endif
		bytes = httpread(buf, bytes);
	if (bytes > 0)
	{
		totalbytes += bytes;
		if (windefault)
		{
			if (totalbytes <= expectbytes)
				msg(MSG_STATUS, "[dd]receiving... $1 of $2 bytes", totalbytes, expectbytes);
			else
				msg(MSG_STATUS, "[d]receiving... $1 bytes", totalbytes);
		}
	}
	return bytes;
}

/* Write data to an URL.  Returns the number of bytes written;
 * if error or user-abort, it given an error message and returns -1.
 */
int urlwrite(buf, bytes)
	CHAR	*buf;
	int	bytes;
{
#ifdef PROTOCOL_FTP
	if (useftp)
		bytes = ftpwrite(buf, bytes);
	/* there is no httpwrite() */

	if (bytes > 0)
	{
		totalbytes += bytes;
		if (windefault)
		{
			if (totalbytes <= expectbytes)
				msg(MSG_STATUS, "[dd]sending... $1 of $2 bytes", totalbytes, expectbytes);
			else
				msg(MSG_STATUS, "[d]sending... $1 bytes", totalbytes);
		}
	}
#endif
	return bytes;
}

/* Close an URL. */
void urlclose()
{
#ifdef PROTOCOL_FTP
	if (useftp)
		ftpclose();
	else
#endif
		httpclose();

	/* restore the old poll frequency */
	o_pollfrequency = oldpollfreq;
}

#endif /* PROTOCOL_HTTP || PROTOCOL_FTP */