#! /usr/local/bin/pike

import Protocols.XMLRPC;

#if constant(Roxen)
#define PMOD
#endif

#ifdef PMOD
mixed parse(object id)
#else
int main(int argc, array(string) argv, array(string) env)
#endif
{
	string response;

	mixed err = catch {

#ifdef PMOD
		if (!sizeof(id->data))
			error("No data");
	
		mixed c = decode_call(id->data);
#else
		string req = Stdio.stdin->read();

		if (!sizeof(req))
			error("No data");

		mixed c = decode_call(req);
#endif
		mixed retval;

		switch (c->method_name)
		{
		case "learnspam":
			retval = learnspam(c);
			break;
		case "checkspam":
			retval = checkspam(c);
			break;
		default:
			break;
		}
	
		if (!retval)
			error("No data to return");

		if (!arrayp(retval))
			retval = ({ retval });

		response = encode_response( retval );
	
	};
	
	if (err)
	{
		response = encode_response_fault(666, describe_error(err));
	}

#ifdef PMOD
    return Roxen.http_string_answer(response, "text/xml");
#else
	write("Content-type: text/xml\r\n\r\n%s", response );
#endif
}

mixed learnspam(mixed c)
{
	if (sizeof(c->params) != 2)
		error("learnspam takes two parameters: an array of messageids, and an options struct");

	mixed retval = "";
	mapping options = ([ ]);

	/* accept a string for arg 1, and convert it to a 1-elem array */
	if (stringp(c->params[0]))
	{
		c->params[0] = ({ c->params[0] });
	}
	
	/* Accept a string as arg 2, and convert it to the mapping
	   ([ "flag":$2 ]).	 This means you want defaults for any other options.
	*/
	   
	if (stringp(c->params[1]))
	{
		c->params[1] = ([ "flag":c->params[1] ]);
	}
	options = c->params[1];

	if (!options->user)
		options->user = "nouser";

	if (options->flag != "HAM" && options->flag != "SPAM")
		error("Invalid spam indicator " + options->flag);

	object s = Sql.Sql("mysql://address@hidden/mail");

	/* remove empty elements */
	c->params[0] = c->params[0] - ({ "" });

	/* fetch the messages */
	array r = s->query("SELECT * FROM messages WHERE messageid in "+
			"('" + map(c->params[0], s->quote)*"', '" + "')");

	if (sizeof(r) == 0)
	{
		if (sizeof(c->params[0]) == 1)
		{
			retval += sprintf("%=-40s\n", 
				"Your message was not found in the database, and cannot be analyzed.");
		} else
		{
			retval += sprintf("%=-40s\n", 
				"None of the selected messages were found in the database.  Nothing to do.");
		}
		return retval;
	}
	
	/* If we're the primary server, submit to razor, and submit to the slave
	   server via XML-RPC */
	if (!options->mirror)
	{
		retval += sprintf("%d message%s found in the database.\n", sizeof(r), sizeof(r) == 1?" was":"s were");

		/* Fetch a list of messages already submitted.  Only match ones
		   flagged the same as the current request, so you can correct bad
		   submissions. 
		*/
		array dups = s->query("SELECT * FROM submitted WHERE messageid in " +
			"('" + map(r->messageid, s->quote)*"', '" + "') and flag = '" + options->flag + "'");

		// retval += sprintf("%d dups for %O\n", sizeof(dups), r->messageid);

		if (sizeof(dups) == sizeof(r))
		{
			if (sizeof(r) == 1)
			{
				retval += sprintf("%=-40s\n", 
					"This message has already been submitted.");
			} else
			{				
				retval += sprintf("%=-40s\n", 
					"All of the selected messages have already been submitted.");
			}
			return retval;
		}
		
		if (sizeof(dups))
		{
			retval += sprintf("%d message%s previously submitted.\n",
				sizeof(dups), sizeof(dups) == 1?" was":"s were");

			r = filter(r, lambda(mapping element) { return search(dups->messageid, element->messageid) == -1;});
		}
	}

	/* Build a fake mbox containing the messages */
	string mbox = "";
	string mboxname = sprintf("/tmp/mbox.%u.%u", getpid(), Thread.id_number());
	foreach (r, mapping message)
	{
		mbox += sprintf("From %s %s\n%s\n\n", message->env_from, 
			ctime(time())-"\n", message->message);
	}
	Stdio.FILE(mboxname, "ctw")->write(mbox);

	/* pipe the message to sa-learn */
	retval += "Spamassassin results:\n";
	retval += prepend(Process.popen(sprintf("/usr/local/bin/sa-learn %s --mbox %s 2>&1",
		(options->flag == "SPAM")?"--spam":"--ham", mboxname)), " ");
	
	/* pipe it to razor-report and also submit to our neighbor if we're the
	   master server
	*/
	if (!options->mirror)
	{
		/* find the name of the slave */
		string host = (System.gethostname()/".")[0];

		switch (host) {
		  case "email1":
			host = "email2"; break;
		  case "email2":
			host = "email1"; break;
		  default:
			error("Not running on a known host");
		}

		retval += "Razor results:\n";
		retval += prepend(Process.popen(sprintf("/usr/local/bin/%s -home=/usr/tmp/razor -f %s 2>&1 && echo Done.", 
			(options->flag == "SPAM")?"razor-report":"razor-revoke", mboxname)), " ");

		retval += "Neighbor results:\n";
		
		string nretval;
		
		/* This may fail if the slave server is down */
		mixed e = catch {
			nretval = Protocols.XMLRPC.Client("http://"+host+"/spam/submit.pike")
					["learnspam"](r->messageid,
								  options + ([ "mirror":"1" ]))[0];
		};
		if (e)
			nretval = describe_error(e);

		retval += prepend(nretval, " ");

		/* This should not fail, but it's okay if it does. */
		e = catch {
			s->query("INSERT IGNORE INTO submitted (messageid, env_from, env_rcpt, message, messagelen, user, flag) " +
				"SELECT messageid, env_from, env_rcpt, message, messagelen, '"+
					s->quote(options->user)+"', '"+
					s->quote(options->flag)+"' "+
				"FROM messages WHERE messageid in "+
				"('" + map(r->messageid, s->quote)*"', '" + "')");
		};
		if (e)
			retval += describe_error(e);
	}
	return retval;
}

mixed checkspam(mixed c)
{
	string retval = "";
	if (sizeof(c->params) != 1)
		error("checkspam takes one parameter: a string containing the messageid to check");
	
	object s = Sql.Sql("mysql://address@hidden/mail");

	/* fetch the message */
	array r = s->query("SELECT * FROM messages WHERE messageid = %s", c->params[0]);

	if (sizeof(r) == 0)
	{
		retval += "Your message was not found in the database, and cannot be analyzed.";
		return retval;
	}

	/* Build a fake mbox containing the message */
	string mbox;
	string mboxname = sprintf("/tmp/mbox.%u.%u", getpid(), Thread.id_number());
	mbox = sprintf("From %s %s\n%s\n\n", r[0]->env_from, 
			ctime(time())-"\n", r[0]->message);
	Stdio.FILE(mboxname, "ctw")->write(mbox);

	/* pipe the message to spamassassin */
	retval += "Spamassassin results:\n";
	mixed result = Process.popen(sprintf("/usr/local/bin/spamassassin -x -t < %s 2>&1", mboxname));
	result = result / "\n";

	/* find the text description at the bottom of the message */
	int i;
	for (i=sizeof(result)-1; i>=0 ; i--)
	{
		if (search(result[i], "Content analysis details:")==0)
			break;
	}
	if (i < 0)
		error(result);
	return result[i..]*"\n";
}

/* Takes a multiline string and prepends another string to each line */
string prepend(string text, string indent)
{
	array ret = ({ });
	foreach (text/"\n", string line)
		ret += ({ indent+line });
	/* don't mess with the null string after a trailing eol */
	if (ret[-1] == indent)
		ret[-1] = "";
	return ret*"\n";
}