curl/lib/formdata.c

/*****************************************************************************
 *                                  _   _ ____  _     
 *  Project                     ___| | | |  _ \| |    
 *                             / __| | | | |_) | |    
 *                            | (__| |_| |  _ <| |___ 
 *                             \___|\___/|_| \_\_____|
 *
 * Copyright (C) 2000, Daniel Stenberg, <daniel@haxx.se>, et al.
 *
 * In order to be useful for every potential user, curl and libcurl are
 * dual-licensed under the MPL and the MIT/X-derivate licenses.
 *
 * You may opt to use, copy, modify, merge, publish, distribute and/or sell
 * copies of the Software, and permit persons to whom the Software is
 * furnished to do so, under the terms of the MPL or the MIT/X-derivate
 * licenses. You may pick one of these licenses.
 *
 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
 * KIND, either express or implied.
 *
 * $Id$
 *****************************************************************************/

/*
  Debug the form generator stand-alone by compiling this source file with:

  gcc -DHAVE_CONFIG_H -I../ -g -D_FORM_DEBUG -o formdata -I../include formdata.c

  run the 'formdata' executable and make sure the output is ok!

  try './formdata "name=Daniel" "poo=noo" "foo=bar"' and similarly

 */

#include "setup.h"

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

#include <time.h>

#include <curl/curl.h>
#include "formdata.h"

#include "strequal.h"

/* The last #include file should be: */
#ifdef MALLOCDEBUG
#include "memdebug.h"
#endif

/* Length of the random boundary string. The risk of this being used
   in binary data is very close to zero, 64^32 makes
   6277101735386680763835789423207666416102355444464034512896
   combinations... */
#define BOUNDARY_LENGTH 32

/* What kind of Content-Type to use on un-specified files with unrecognized
   extensions. */
#define HTTPPOST_CONTENTTYPE_DEFAULT "text/plain"

/* This is a silly duplicate of the function in main.c to enable this source
   to compile stand-alone for better debugging */
static void GetStr(char **string,
		   char *value)
{
  if(*string)
    free(*string);
  *string = strdup(value);
}

/***************************************************************************
 *
 * FormParse()
 *	
 * Reads a 'name=value' paramter and builds the appropriate linked list.
 *
 * Specify files to upload with 'name=@filename'. Supports specified
 * given Content-Type of the files. Such as ';type=<content-type>'.
 *
 * You may specify more than one file for a single name (field). Specify
 * multiple files by writing it like:
 *
 * 'name=@filename,filename2,filename3'
 *
 * If you want content-types specified for each too, write them like:
 *
 * 'name=@filename;type=image/gif,filename2,filename3'
 *
 ***************************************************************************/

#define FORM_FILE_SEPARATOR ','
#define FORM_TYPE_SEPARATOR ';'

static
int FormParse(char *input,
	      struct HttpPost **httppost,
	      struct HttpPost **last_post)
{
  /* nextarg MUST be a string in the format 'name=contents' and we'll
     build a linked list with the info */
  char name[256];
  char contents[4096]="";
  char major[128];
  char minor[128];
  long flags = 0;
  char *contp;
  char *type = NULL;
  char *prevtype = NULL;
  char *sep;
  char *sep2;
  struct HttpPost *post;
  struct HttpPost *subpost; /* a sub-node */
  unsigned int i;

  if(1 <= sscanf(input, "%255[^ =] = %4095[^\n]", name, contents)) {
    /* the input was using the correct format */
    contp = contents;

    if('@' == contp[0]) {
      /* we use the @-letter to indicate file name(s) */
      
      flags = HTTPPOST_FILENAME;
      contp++;

      post=NULL;

      do {
	/* since this was a file, it may have a content-type specifier
	   at the end too */

	sep=strchr(contp, FORM_TYPE_SEPARATOR);
	sep2=strchr(contp, FORM_FILE_SEPARATOR);

	/* pick the closest */
	if(sep2 && (sep2 < sep)) {
	  sep = sep2;

	  /* no type was specified! */
	}
	if(sep) {

	  /* if we got here on a comma, don't do much */
	  if(FORM_FILE_SEPARATOR != *sep)
	    type = strstr(sep+1, "type=");
	  else
	    type=NULL;

	  *sep=0; /* terminate file name at separator */

	  if(type) {
	    type += strlen("type=");
	    
	    if(2 != sscanf(type, "%127[^/]/%127[^,\n]",
			   major, minor)) {
	      fprintf(stderr, "Illegally formatted content-type field!\n");
	      return 2; /* illegal content-type syntax! */
	    }
	    /* now point beyond the content-type specifier */
	    sep = type + strlen(major)+strlen(minor)+1;

	    /* find the following comma */
	    sep=strchr(sep, FORM_FILE_SEPARATOR);
	  }
	}
	else {
	  type=NULL;
	  sep=strchr(contp, FORM_FILE_SEPARATOR);
	}
	if(sep) {
	  /* the next file name starts here */
	  *sep =0;
	  sep++;
	}
	if(!type) {
	  /*
	   * No type was specified, we scan through a few well-known
	   * extensions and pick the first we match!
	   */
	  struct ContentType {
	    char *extension;
	    char *type;
	  };
	  static struct ContentType ctts[]={
	    {".gif",  "image/gif"},
	    {".jpg",  "image/jpeg"},
	    {".jpeg", "image/jpeg"},
	    {".txt",  "text/plain"},
	    {".html", "text/plain"}
	  };

	  if(prevtype)
	    /* default to the previously set/used! */
	    type = prevtype;
	  else
	    /* It seems RFC1867 defines no Content-Type to default to
	       text/plain so we don't actually need to set this: */
	    type = HTTPPOST_CONTENTTYPE_DEFAULT;

	  for(i=0; i<sizeof(ctts)/sizeof(ctts[0]); i++) {
	    if(strlen(contp) >= strlen(ctts[i].extension)) {
	      if(strequal(contp +
			  strlen(contp) - strlen(ctts[i].extension),
			  ctts[i].extension)) {
		type = ctts[i].type;
		break;
	      }	      
	    }
	  }
	  /* we have a type by now */
	}

	if(NULL == post) {
	  /* For the first file name, we allocate and initiate the main list
	     node */

	  post = (struct HttpPost *)malloc(sizeof(struct HttpPost));
	  if(post) {
	    memset(post, 0, sizeof(struct HttpPost));
	    GetStr(&post->name, name);      /* get the name */
	    GetStr(&post->contents, contp); /* get the contents */
	    post->flags = flags;
	    if(type) {
	      GetStr(&post->contenttype, type); /* get type */
	      prevtype=post->contenttype; /* point to the allocated string! */
	    }
	    /* make the previous point to this */
	    if(*last_post)
	      (*last_post)->next = post;
	    else
	      (*httppost) = post;

	    (*last_post) = post;	  
	  }

	}
	else {
	  /* we add a file name to the previously allocated node, known as
             'post' now */
	  subpost =(struct HttpPost *)malloc(sizeof(struct HttpPost));
	  if(subpost) {
	     memset(subpost, 0, sizeof(struct HttpPost));
	     GetStr(&subpost->name, name);      /* get the name */
	     GetStr(&subpost->contents, contp); /* get the contents */
	     subpost->flags = flags;
	     if(type) {
	       GetStr(&subpost->contenttype, type); /* get type */
	       prevtype=subpost->contenttype; /* point to the allocated string! */
	     }
	     /* now, point our 'more' to the original 'more' */
	     subpost->more = post->more;

	     /* then move the original 'more' to point to ourselves */
	     post->more = subpost;	     
	  }
	}
	contp = sep; /* move the contents pointer to after the separator */
      } while(sep && *sep); /* loop if there's another file name */
    }
    else {
      post = (struct HttpPost *)malloc(sizeof(struct HttpPost));
      if(post) {
	memset(post, 0, sizeof(struct HttpPost));
	GetStr(&post->name, name);      /* get the name */
	if( contp[0]=='<' ) {
	  GetStr(&post->contents, contp+1); /* get the contents */
	  post->flags = HTTPPOST_READFILE;
	}
	else {
	  GetStr(&post->contents, contp); /* get the contents */
	  post->flags = 0;
	}

	/* make the previous point to this */
	if(*last_post)
	  (*last_post)->next = post;
	else
	  (*httppost) = post;

	(*last_post) = post;	  
      }

    }

  }
  else {
    fprintf(stderr, "Illegally formatted input field!\n");
    return 1;
  }
  return 0;
}

int curl_formparse(char *input,
                   struct HttpPost **httppost,
                   struct HttpPost **last_post)
{
  return FormParse(input, httppost, last_post);
}

static int AddFormData(struct FormData **formp,
			void *line,
			long length)
{
  struct FormData *newform = (struct FormData *)
    malloc(sizeof(struct FormData));
  newform->next = NULL;

  /* we make it easier for plain strings: */
  if(!length)
    length = strlen((char *)line);

  newform->line = (char *)malloc(length+1);
  memcpy(newform->line, line, length+1);
  newform->length = length;
  newform->line[length]=0; /* zero terminate for easier debugging */
  
  if(*formp) {
    (*formp)->next = newform;
    *formp = newform;
  }
  else
    *formp = newform;

  return length;
}


static int AddFormDataf(struct FormData **formp,
			 char *fmt, ...)
{
  char s[4096];
  va_list ap;
  va_start(ap, fmt);
  vsprintf(s, fmt, ap);
  va_end(ap);

  return AddFormData(formp, s, 0);
}


char *Curl_FormBoundary(void)
{
  char *retstring;
  static int randomizer=0; /* this is just so that two boundaries within
			      the same form won't be identical */
  int i;

  static char table64[]=
    "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";

  retstring = (char *)malloc(BOUNDARY_LENGTH);

  if(!retstring)
    return NULL; /* failed */

  srand(time(NULL)+randomizer++); /* seed */

  strcpy(retstring, "curl"); /* bonus commercials 8*) */

  for(i=4; i<(BOUNDARY_LENGTH-1); i++) {
    retstring[i] = table64[rand()%64];
  }
  retstring[BOUNDARY_LENGTH-1]=0; /* zero terminate */

  return retstring;
}

/* Used from http.c */ 
void Curl_FormFree(struct FormData *form)
{
  struct FormData *next;
  do {
    next=form->next;  /* the following form line */
    free(form->line); /* free the line */
    free(form);       /* free the struct */
  
  } while((form=next)); /* continue */
}

/* external function to free up a whole form post chain */
void curl_formfree(struct HttpPost *form)
{
  struct HttpPost *next;
  do {
    next=form->next;  /* the following form line */

    /* recurse to sub-contents */
    if(form->more)
      curl_formfree(form->more);

    if(form->name)
      free(form->name); /* free the name */
    if(form->contents)
      free(form->contents); /* free the contents */
    if(form->contenttype)
      free(form->contenttype); /* free the content type */
    free(form);       /* free the struct */

  } while((form=next)); /* continue */
}

struct FormData *Curl_getFormData(struct HttpPost *post,
                                  int *sizep)
{
  struct FormData *form = NULL;
  struct FormData *firstform;

  struct HttpPost *file;

  int size =0;
  char *boundary;
  char *fileboundary=NULL;

  if(!post)
    return NULL; /* no input => no output! */

  boundary = Curl_FormBoundary();
  
  /* Make the first line of the output */
  AddFormDataf(&form,
               "Content-Type: multipart/form-data;"
               " boundary=%s\r\n",
               boundary);
  /* we DO NOT count that line since that'll be part of the header! */

  firstform = form;
  
  do {

    /* boundary */
    size += AddFormDataf(&form, "\r\n--%s\r\n", boundary);

    size += AddFormDataf(&form,
			 "Content-Disposition: form-data; name=\"%s\"",
			 post->name);

    if(post->more) {
      /* If used, this is a link to more file names, we must then do
         the magic to include several files with the same field name */

      fileboundary = Curl_FormBoundary();

      size += AddFormDataf(&form,
			   "\r\nContent-Type: multipart/mixed,"
			   " boundary=%s\r\n",
			   fileboundary);
    }

    file = post;

    do {
      if(post->more) {
	/* if multiple-file */
	size += AddFormDataf(&form,
			     "\r\n--%s\r\nContent-Disposition: attachment; filename=\"%s\"",
			     fileboundary, file->contents);
      }
      else if(post->flags & HTTPPOST_FILENAME) {
	size += AddFormDataf(&form,
			     "; filename=\"%s\"",
			     post->contents);
      }
      
      if(file->contenttype) {
	/* we have a specified type */
	size += AddFormDataf(&form,
			     "\r\nContent-Type: %s",
			     file->contenttype);
      }

#if 0
      /* The header Content-Transfer-Encoding: seems to confuse some receivers
       * (like the built-in PHP engine). While I can't see any reason why it
       * should, I can just as well skip this to the benefit of the users who
       * are using such confused receivers.
       */
      
      if(file->contenttype &&
	 !strnequal("text/", file->contenttype, 5)) {
	/* this is not a text content, mention our binary encoding */
	size += AddFormData(&form, "\r\nContent-Transfer-Encoding: binary", 0);
      }
#endif

      size += AddFormData(&form, "\r\n\r\n", 0);

      if((post->flags & HTTPPOST_FILENAME) ||
	 (post->flags & HTTPPOST_READFILE)) {
	/* we should include the contents from the specified file */
	FILE *fileread;
	char buffer[1024];
	int nread;

	fileread = strequal("-", file->contents)?stdin:
          /* binary read for win32 crap */
          fopen(file->contents, "rb");
	if(fileread) {
	  while((nread = fread(buffer, 1, 1024, fileread))) {
	    size += AddFormData(&form,
				buffer,
				nread);
	  }
          if(fileread != stdin)
            fclose(fileread);
	} else {
	  size += AddFormData(&form, "[File wasn't found by client]", 0);
	}
      } else {
	/* include the contents we got */
	size += AddFormData(&form, post->contents, 0);
      }
    } while((file = file->more)); /* for each specified file for this field */

    if(post->more) {
      /* this was a multiple-file inclusion, make a termination file
         boundary: */
      size += AddFormDataf(&form,
			   "\r\n--%s--",
			   fileboundary);     
      free(fileboundary);
    }

  } while((post=post->next)); /* for each field */

  /* end-boundary for everything */
  size += AddFormDataf(&form,
		       "\r\n--%s--\r\n",
		       boundary);

  *sizep = size;

  free(boundary);

  return firstform;
}

int Curl_FormInit(struct Form *form, struct FormData *formdata )
{
  if(!formdata)
    return 1; /* error */

  form->data = formdata;
  form->sent = 0;

  return 0;
}

/* fread() emulation */
int Curl_FormReader(char *buffer,
                    size_t size,
                    size_t nitems,
                    FILE *mydata)
{
  struct Form *form;
  int wantedsize;
  int gotsize;

  form=(struct Form *)mydata;

  wantedsize = size * nitems;

  if(!form->data)
    return -1; /* nothing, error, empty */

  do {
  
    if( (form->data->length - form->sent ) > wantedsize ) {

      memcpy(buffer, form->data->line + form->sent, wantedsize);

      form->sent += wantedsize;

      return wantedsize;
    }

    memcpy(buffer,
           form->data->line + form->sent,
           gotsize = (form->data->length - form->sent) );

    form->sent = 0;

    form->data = form->data->next; /* advance */

  } while(!gotsize && form->data);
  /* If we got an empty line and we have more data, we proceed to the next
     line immediately to avoid returning zero before we've reached the end.
     This is the bug reported November 22 1999 on curl 6.3. (Daniel) */

  return gotsize;
}


#ifdef _FORM_DEBUG

int main(int argc, char **argv)
{
#if 0
  char *testargs[]={
    "name1 = data in number one",
    "name2 = number two data",
    "test = @upload"
  };
#endif
  int i;
  char *nextarg;
  struct HttpPost *httppost=NULL;
  struct HttpPost *last_post=NULL;
  struct HttpPost *post;
  int size;
  int nread;
  char buffer[4096];

  struct FormData *form;
  struct Form formread;

  for(i=1; i<argc; i++) {

    if( FormParse( argv[i],
		   &httppost,
		   &last_post)) {
      fprintf(stderr, "Illegally formatted input field: '%s'!\n",
	      argv[i]);
      return 1;
    }
  }

  form=Curl_getFormData(httppost, &size);

  FormInit(&formread, form);

  while(nread = FormReader(buffer, 1, sizeof(buffer), (FILE *)&formread)) {
    fwrite(buffer, nread, 1, stderr);
  }

  fprintf(stderr, "size: %d\n", size);

  return 0;
}

#endif