1

I'm using cJson for C, into an embedded platform, to build a JSON like this:

{"ProtocolVersion":1,"ID":"123","Zone":"xyz","MessageType":42,"Message":"Json payload MSG #6"}

To do this I use this function provided by cJson:

  jsonObject = cJSON_CreateObject();

  /* Add Protocol Version */
  cJSON_AddNumberToObject( jsonObject, "ProtocolVersion", 1);
  /* Add Machine Id */
  cJSON_AddStringToObject( jsonObject, "ID", idStr );
  /* Add Market */
  cJSON_AddStringToObject( jsonObject, "Zone", xyzStr);
  /* Add Message Type */
  cJSON_AddNumberToObject( jsonObject, "MessageType", typeNum);
  /* Add Message */
  cJSON_AddStringToObject( jsonObject, "Message", msgStr);

To create the json I used the function:

cJSON_PrintPreallocated( jsonObject, *jsonMessage, *jsonMessageLen, 0 );

I prefer to pass to cJson a buffer preallocated by my application and I compute the buffer length basically by sum the length of each Key and Object.

For example: strlen("Zone") + strlen(xyzStr) + ... + Number of "" + Number of {} + Number of , + Number of :

In this way I obtain the exact length of my JSON.

Unfortunately, the function "cJSON_PrintPreallocated" fails due to a incorrect buffer length (it seems to short).

If I add an extra 30 bytes to my "jsonMessage" everything works.

Where I'm wrong?

Which is the best way to compute the buffer length required by cJson?

Thanks!

too honest for this site
  • 12,050
  • 4
  • 30
  • 52
Federico
  • 1,117
  • 6
  • 20
  • 37
  • Think carefully about how JSON is structured. It's not _just_ the keys and values, and the various separators. Open up an example JSON document and look at what other characters, besides those in the keys and values, can be there. – Nic Mar 17 '17 at 16:16
  • Also, why is `jsonMessageLen` a pointer? If it's not, you need to turn on warnings, because you're trying to take the value at a non-pointer. – Nic Mar 17 '17 at 16:17
  • Use use the pointer to jsonMessageLen because I allocate the jsonMessage dynamically. – Federico Mar 20 '17 at 09:46
  • I use cJson to print the json unformatted, in this way I can avoid the print of tabs and spaces. – Federico Mar 20 '17 at 13:57

1 Answers1

1

When printing numbers, cJSON conservatively reserves 64 bytes in the output buffer, because it can't know for sure how long a sprintf will print a number. In case of the number 1 this can mean that cJSON requests 63 bytes more than it actually needs.

This will probably change in a future version of cJSON, but for now there is now way around that.

To be safe you should always make your buffer 63 bytes longer than what you expect it to be.

UPDATE:

This has been updated. Since version 1.4.6 only requires 5 additional bytes.

FSMaxB
  • 2,280
  • 3
  • 22
  • 41