/* $Id$ */
/* 
 * Copyright (C) 2008-2011 Teluu Inc. (http://www.teluu.com)
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA 
 */
#include <pjsip/sip_multipart.h>
#include <pjsip/sip_parser.h>
#include <pjlib-util/scanner.h>
#include <pj/assert.h>
#include <pj/ctype.h>
#include <pj/errno.h>
#include <pj/except.h>
#include <pj/guid.h>
#include <pj/log.h>
#include <pj/pool.h>
#include <pj/string.h>

#define THIS_FILE		"sip_multipart.c"

#define IS_SPACE(c)	((c)==' ' || (c)=='\t')

#if 0
#   define TRACE_(x)	PJ_LOG(4,x)
#else
#   define TRACE_(x)
#endif

extern pj_bool_t pjsip_use_compact_form;

/* Type of "data" in multipart pjsip_msg_body */
struct multipart_data
{
    pj_str_t	    	  boundary;
    pjsip_multipart_part  part_head;
};


static int multipart_print_body(struct pjsip_msg_body *msg_body,
			        char *buf, pj_size_t size)
{
    const struct multipart_data *m_data;
    pj_str_t clen_hdr =  { "Content-Length: ", 16};
    pjsip_multipart_part *part;
    char *p = buf, *end = buf+size;

#define SIZE_LEFT()	(end-p)

    m_data = (const struct multipart_data*)msg_body->data;

    PJ_ASSERT_RETURN(m_data && !pj_list_empty(&m_data->part_head), PJ_EINVAL);

    part = m_data->part_head.next;
    while (part != &m_data->part_head) {
	enum { CLEN_SPACE = 5 };
	char *clen_pos;
	const pjsip_hdr *hdr;

	clen_pos = NULL;

	/* Print delimiter */
	if (SIZE_LEFT() <= (m_data->boundary.slen+8) << 1)
	    return -1;
	*p++ = 13; *p++ = 10; *p++ = '-'; *p++ = '-';
	pj_memcpy(p, m_data->boundary.ptr, m_data->boundary.slen);
	p += m_data->boundary.slen;
	*p++ = 13; *p++ = 10;

	/* Print optional headers */
	hdr = part->hdr.next;
	while (hdr != &part->hdr) {
	    int printed = pjsip_hdr_print_on((pjsip_hdr*)hdr, p,
	                                     SIZE_LEFT()-2);
	    if (printed < 0)
		return -1;
	    p += printed;
	    *p++ = '\r';
	    *p++ = '\n';
	    hdr = hdr->next;
	}

	/* Automaticly adds Content-Type and Content-Length headers, only
	 * if content_type is set in the message body.
	 */
	if (part->body && part->body->content_type.type.slen) {
	    pj_str_t ctype_hdr = { "Content-Type: ", 14};
	    const pjsip_media_type *media = &part->body->content_type;

	    if (pjsip_use_compact_form) {
		ctype_hdr.ptr = "c: ";
		ctype_hdr.slen = 3;
	    }

	    /* Add Content-Type header. */
	    if ( (end-p) < 24 + media->type.slen + media->subtype.slen) {
		return -1;
	    }
	    pj_memcpy(p, ctype_hdr.ptr, ctype_hdr.slen);
	    p += ctype_hdr.slen;
	    p += pjsip_media_type_print(p, (unsigned)(end-p), media);
	    *p++ = '\r';
	    *p++ = '\n';

	    /* Add Content-Length header. */
	    if ((end-p) < clen_hdr.slen + 12 + 2) {
		return -1;
	    }
	    pj_memcpy(p, clen_hdr.ptr, clen_hdr.slen);
	    p += clen_hdr.slen;

	    /* Print blanks after "Content-Length:", this is where we'll put
	     * the content length value after we know the length of the
	     * body.
	     */
	    pj_memset(p, ' ', CLEN_SPACE);
	    clen_pos = p;
	    p += CLEN_SPACE;
	    *p++ = '\r';
	    *p++ = '\n';
	}

	/* Empty newline */
	*p++ = 13; *p++ = 10;

	/* Print the body */
	pj_assert(part->body != NULL);
	if (part->body) {
	    int printed = part->body->print_body(part->body, p, SIZE_LEFT());
	    if (printed < 0)
		return -1;
	    p += printed;

	    /* Now that we have the length of the body, print this to the
	     * Content-Length header.
	     */
	    if (clen_pos) {
		char tmp[16];
		int len;

		len = pj_utoa(printed, tmp);
		if (len > CLEN_SPACE) len = CLEN_SPACE;
		pj_memcpy(clen_pos+CLEN_SPACE-len, tmp, len);
	    }
	}

	part = part->next;
    }

    /* Print closing delimiter */
    if (SIZE_LEFT() < m_data->boundary.slen+8)
	return -1;
    *p++ = 13; *p++ = 10; *p++ = '-'; *p++ = '-';
    pj_memcpy(p, m_data->boundary.ptr, m_data->boundary.slen);
    p += m_data->boundary.slen;
    *p++ = '-'; *p++ = '-'; *p++ = 13; *p++ = 10;

#undef SIZE_LEFT

    return (int)(p - buf);
}

static void* multipart_clone_data(pj_pool_t *pool, const void *data,
				  unsigned len)
{
    const struct multipart_data *src;
    struct multipart_data *dst;
    const pjsip_multipart_part *src_part;

    PJ_UNUSED_ARG(len);

    src = (const struct multipart_data*) data;
    dst = PJ_POOL_ALLOC_T(pool, struct multipart_data);

    pj_strdup(pool, &dst->boundary, &src->boundary);

    src_part = src->part_head.next;
    while (src_part != &src->part_head) {
	pjsip_multipart_part *dst_part;
	const pjsip_hdr *src_hdr;

	dst_part = pjsip_multipart_create_part(pool);

	src_hdr = src_part->hdr.next;
	while (src_hdr != &src_part->hdr) {
	    pjsip_hdr *dst_hdr = (pjsip_hdr*)pjsip_hdr_clone(pool, src_hdr);
	    pj_list_push_back(&dst_part->hdr, dst_hdr);
	    src_hdr = src_hdr->next;
	}

	dst_part->body = pjsip_msg_body_clone(pool, src_part->body);

	pj_list_push_back(&dst->part_head, dst_part);

	src_part = src_part->next;
    }

    return (void*)dst;
}

/*
 * Create an empty multipart body.
 */
PJ_DEF(pjsip_msg_body*) pjsip_multipart_create( pj_pool_t *pool,
						const pjsip_media_type *ctype,
						const pj_str_t *boundary)
{
    pjsip_msg_body *body;
    pjsip_param *ctype_param;
    struct multipart_data *mp_data;
    pj_str_t STR_BOUNDARY = { "boundary", 8 };

    PJ_ASSERT_RETURN(pool, NULL);

    body = PJ_POOL_ZALLOC_T(pool, pjsip_msg_body);

    /* content-type */
    if (ctype && ctype->type.slen) {
	pjsip_media_type_cp(pool, &body->content_type, ctype);
    } else {
	pj_str_t STR_MULTIPART = {"multipart", 9};
	pj_str_t STR_MIXED = { "mixed", 5 };

        pjsip_media_type_init(&body->content_type,
                              &STR_MULTIPART, &STR_MIXED);
    }

    /* multipart data */
    mp_data = PJ_POOL_ZALLOC_T(pool, struct multipart_data);
    pj_list_init(&mp_data->part_head);
    if (boundary) {
	pj_strdup(pool, &mp_data->boundary, boundary);
    } else {
	pj_create_unique_string(pool, &mp_data->boundary);
    }
    body->data = mp_data;

    /* Add ";boundary" parameter to content_type parameter. */
    ctype_param = pjsip_param_find(&body->content_type.param, &STR_BOUNDARY);
    if (!ctype_param) {
	ctype_param = PJ_POOL_ALLOC_T(pool, pjsip_param);
	ctype_param->name = STR_BOUNDARY;
	pj_list_push_back(&body->content_type.param, ctype_param);
    }
    ctype_param->value = mp_data->boundary;

    /* function pointers */
    body->print_body = &multipart_print_body;
    body->clone_data = &multipart_clone_data;

    return body;
}

/*
 * Create an empty multipart part.
 */
PJ_DEF(pjsip_multipart_part*) pjsip_multipart_create_part(pj_pool_t *pool)
{
    pjsip_multipart_part *mp;

    mp = PJ_POOL_ZALLOC_T(pool, pjsip_multipart_part);
    pj_list_init(&mp->hdr);

    return mp;
}


/*
 * Deep clone.
 */
PJ_DEF(pjsip_multipart_part*)
pjsip_multipart_clone_part(pj_pool_t *pool,
			   const pjsip_multipart_part *src)
{
    pjsip_multipart_part *dst;
    const pjsip_hdr *hdr;

    dst = pjsip_multipart_create_part(pool);

    hdr = src->hdr.next;
    while (hdr != &src->hdr) {
	pj_list_push_back(&dst->hdr, pjsip_hdr_clone(pool, hdr));
	hdr = hdr->next;
    }

    dst->body = pjsip_msg_body_clone(pool, src->body);

    return dst;
}


/*
 * Add a part into multipart bodies.
 */
PJ_DEF(pj_status_t) pjsip_multipart_add_part( pj_pool_t *pool,
					      pjsip_msg_body *mp,
					      pjsip_multipart_part *part)
{
    struct multipart_data *m_data;

    /* All params must be specified */
    PJ_ASSERT_RETURN(pool && mp && part, PJ_EINVAL);

    /* mp must really point to an actual multipart msg body */
    PJ_ASSERT_RETURN(mp->print_body==&multipart_print_body, PJ_EINVAL);

    /* The multipart part must contain a valid message body */
    PJ_ASSERT_RETURN(part->body && part->body->print_body, PJ_EINVAL);

    m_data = (struct multipart_data*)mp->data;
    pj_list_push_back(&m_data->part_head, part);

    PJ_UNUSED_ARG(pool);

    return PJ_SUCCESS;
}

/*
 * Get the first part of multipart bodies.
 */
PJ_DEF(pjsip_multipart_part*)
pjsip_multipart_get_first_part(const pjsip_msg_body *mp)
{
    struct multipart_data *m_data;

    /* Must specify mandatory params */
    PJ_ASSERT_RETURN(mp, NULL);

    /* mp must really point to an actual multipart msg body */
    PJ_ASSERT_RETURN(mp->print_body==&multipart_print_body, NULL);

    m_data = (struct multipart_data*)mp->data;
    if (pj_list_empty(&m_data->part_head))
	return NULL;

    return m_data->part_head.next;
}

/*
 * Get the next part after the specified part.
 */
PJ_DEF(pjsip_multipart_part*)
pjsip_multipart_get_next_part(const pjsip_msg_body *mp,
			      pjsip_multipart_part *part)
{
    struct multipart_data *m_data;

    /* Must specify mandatory params */
    PJ_ASSERT_RETURN(mp && part, NULL);

    /* mp must really point to an actual multipart msg body */
    PJ_ASSERT_RETURN(mp->print_body==&multipart_print_body, NULL);

    m_data = (struct multipart_data*)mp->data;

    /* the part parameter must be really member of the list */
    PJ_ASSERT_RETURN(pj_list_find_node(&m_data->part_head, part) != NULL,
		     NULL);

    if (part->next == &m_data->part_head)
	return NULL;

    return part->next;
}

/*
 * Find a body inside multipart bodies which has the specified content type.
 */
PJ_DEF(pjsip_multipart_part*)
pjsip_multipart_find_part( const pjsip_msg_body *mp,
			   const pjsip_media_type *content_type,
			   const pjsip_multipart_part *start)
{
    struct multipart_data *m_data;
    pjsip_multipart_part *part;

    /* Must specify mandatory params */
    PJ_ASSERT_RETURN(mp && content_type, NULL);

    /* mp must really point to an actual multipart msg body */
    PJ_ASSERT_RETURN(mp->print_body==&multipart_print_body, NULL);

    m_data = (struct multipart_data*)mp->data;

    if (start)
	part = start->next;
    else
	part = m_data->part_head.next;

    while (part != &m_data->part_head) {
	if (pjsip_media_type_cmp(&part->body->content_type,
				 content_type, 0)==0)
	{
	    return part;
	}
	part = part->next;
    }

    return NULL;
}

/* Parse a multipart part. "pct" is parent content-type  */
static pjsip_multipart_part *parse_multipart_part(pj_pool_t *pool,
						  char *start,
						  pj_size_t len,
						  const pjsip_media_type *pct)
{
    pjsip_multipart_part *part = pjsip_multipart_create_part(pool);
    char *p = start, *end = start+len, *end_hdr = NULL, *start_body = NULL;
    pjsip_ctype_hdr *ctype_hdr = NULL;

    TRACE_((THIS_FILE, "Parsing part: begin--\n%.*s\n--end",
	    (int)len, start));

    /* Find the end of header area, by looking at an empty line */
    for (;;) {
	while (p!=end && *p!='\n') ++p;
	if (p==end) {
	    start_body = end;
	    break;
	}
	if ((p==start) || (p==start+1 && *(p-1)=='\r')) {
	    /* Empty header section */
	    end_hdr = start;
	    start_body = ++p;
	    break;
	} else if (p==end-1) {
	    /* Empty body section */
	    end_hdr = end;
	    start_body = ++p;
	} else if ((p>=start+1 && *(p-1)=='\n') ||
	           (p>=start+2 && *(p-1)=='\r' && *(p-2)=='\n'))
	{
	    /* Found it */
	    end_hdr = (*(p-1)=='\r') ? (p-1) : p;
	    start_body = ++p;
	    break;
	} else {
	    ++p;
	}
    }

    /* Parse the headers */
    if (end_hdr-start > 0) {
	pjsip_hdr *hdr;
	pj_status_t status;

	status = pjsip_parse_headers(pool, start, end_hdr-start, 
				     &part->hdr, 0);
	if (status != PJ_SUCCESS) {
	    PJ_PERROR(2,(THIS_FILE, status, "Warning: error parsing multipart"
					    " header"));
	}

	/* Find Content-Type header */
	hdr = part->hdr.next;
	while (hdr != &part->hdr) {
	    TRACE_((THIS_FILE, "Header parsed: %.*s", (int)hdr->name.slen,
		    hdr->name.ptr));
	    if (hdr->type == PJSIP_H_CONTENT_TYPE) {
		ctype_hdr = (pjsip_ctype_hdr*)hdr;
	    }
	    hdr = hdr->next;
	}
    }

    /* Assign the body */
    part->body = PJ_POOL_ZALLOC_T(pool, pjsip_msg_body);
    if (ctype_hdr) {
	pjsip_media_type_cp(pool, &part->body->content_type, &ctype_hdr->media);
    } else if (pct && pj_stricmp2(&pct->subtype, "digest")==0) {
	part->body->content_type.type = pj_str("message");
	part->body->content_type.subtype = pj_str("rfc822");
    } else {
	part->body->content_type.type = pj_str("text");
	part->body->content_type.subtype = pj_str("plain");
    }

    if (start_body < end) {
	part->body->data = start_body;
	part->body->len = (unsigned)(end - start_body);
    } else {
	part->body->data = (void*)"";
	part->body->len = 0;
    }
    TRACE_((THIS_FILE, "Body parsed: \"%.*s\"", (int)part->body->len,
	    part->body->data));
    part->body->print_body = &pjsip_print_text_body;
    part->body->clone_data = &pjsip_clone_text_data;

    return part;
}

/* Public function to parse multipart message bodies into its parts */
PJ_DEF(pjsip_msg_body*) pjsip_multipart_parse(pj_pool_t *pool,
					      char *buf, pj_size_t len,
					      const pjsip_media_type *ctype,
					      unsigned options)
{
    pj_str_t boundary, delim;
    char *curptr, *endptr;
    const pjsip_param *ctype_param;
    const pj_str_t STR_BOUNDARY = { "boundary", 8 };
    pjsip_msg_body *body = NULL;

    PJ_ASSERT_RETURN(pool && buf && len && ctype && !options, NULL);

    TRACE_((THIS_FILE, "Started parsing multipart body"));

    /* Get the boundary value in the ctype */
    boundary.ptr = NULL;
    boundary.slen = 0;
    ctype_param = pjsip_param_find(&ctype->param, &STR_BOUNDARY);
    if (ctype_param) {
	boundary = ctype_param->value;
	if (boundary.slen>2 && *boundary.ptr=='"') {
	    /* Remove quote */
	    boundary.ptr++;
	    boundary.slen -= 2;
	}
	TRACE_((THIS_FILE, "Boundary is specified: '%.*s'", (int)boundary.slen,
		boundary.ptr));
    }

    if (!boundary.slen) {
	/* Boundary not found or not specified. Try to be clever, get
	 * the boundary from the body.
	 */
	char *p=buf, *end=buf+len;

	PJ_LOG(4,(THIS_FILE, "Warning: boundary parameter not found or "
			     "not specified when parsing multipart body"));

	/* Find the first "--". This "--" must be right after a CRLF, unless
	 * it really appears at the start of the buffer.
	 */
	for (;;) {
	    while (p!=end && *p!='-') ++p;
	    if (p!=end && *(p+1)=='-' &&
		((p>buf && *(p-1)=='\n') || (p==buf)))
	    {
		p+=2;
		break;
	    } else {
		++p;
	    }
	}

	if (p==end) {
	    /* Unable to determine boundary. Maybe this is not a multipart
	     * message?
	     */
	    PJ_LOG(4,(THIS_FILE, "Error: multipart boundary not specified and"
				 " unable to calculate from the body"));
	    return NULL;
	}

	boundary.ptr = p;
	while (p!=end && !pj_isspace(*p)) ++p;
	boundary.slen = p - boundary.ptr;

	TRACE_((THIS_FILE, "Boundary is calculated: '%.*s'",
		(int)boundary.slen, boundary.ptr));
    }

    /* Build the delimiter:
     *   delimiter = "--" boundary
     */
    delim.slen = boundary.slen+2;
    delim.ptr = (char*)pj_pool_alloc(pool, (int)delim.slen);
    delim.ptr[0] = '-';
    delim.ptr[1] = '-';
    pj_memcpy(delim.ptr+2, boundary.ptr, boundary.slen);

    /* Start parsing the body, skip until the first delimiter. */
    curptr = buf;
    endptr = buf + len;
    {
	pj_str_t body;

	body.ptr = buf; body.slen = len;
	curptr = pj_strstr(&body, &delim);
	if (!curptr)
	    return NULL;
    }

    body = pjsip_multipart_create(pool, ctype, &boundary);

    for (;;) {
	char *start_body, *end_body;
	pjsip_multipart_part *part;

	/* Eat the boundary */
	curptr += delim.slen;
	if (*curptr=='-' && curptr<endptr-1 && *(curptr+1)=='-') {
	    /* Found the closing delimiter */
	    curptr += 2;
	    break;
	}
	/* Optional whitespace after delimiter */
	while (curptr!=endptr && IS_SPACE(*curptr)) ++curptr;
	/* Mandatory CRLF */
	if (*curptr=='\r') ++curptr;
	if (*curptr!='\n') {
	    /* Expecting a newline here */
	    return NULL;
	}
	++curptr;

	/* We now in the start of the body */
	start_body = curptr;

	/* Find the next delimiter */
	{
	    pj_str_t subbody;

	    subbody.ptr = curptr; subbody.slen = endptr - curptr;
	    curptr = pj_strstr(&subbody, &delim);
	    if (!curptr) {
		/* We're really expecting end delimiter to be found. */
		return NULL;
	    }
	}

	end_body = curptr;

	/* The newline preceeding the delimiter is conceptually part of
	 * the delimiter, so trim it from the body.
	 */
	if (*(end_body-1) == '\n')
	    --end_body;
	if (*(end_body-1) == '\r')
	    --end_body;

	/* Now that we have determined the part's boundary, parse it
	 * to get the header and body part of the part.
	 */
	part = parse_multipart_part(pool, start_body, end_body - start_body,
				    ctype);
	if (part) {
	    pjsip_multipart_add_part(pool, body, part);
	}
    }

    return body;
}