json: json_obj_parse will modify the input string

[codingspirit]

Describe the bug
The end " of string items will be changed to \0 after json_obj_parse, i.e. we have a char array like this:

{
    "method": "hello",
    "id": 1
}

if using json_obj_parse to proceed it, then it will be changed to

{
    "method": "hello�, //here " symbol changed to "\0"
    "id": 1
}

To Reproduce
Steps to reproduce the behavior:

1. create a json char array with string items
2. decode it with json_obj_parse
3. print out the json char array
4. see modifications

Expected behavior
From my point of view, json_obj_parse should be

int json_obj_parse(const char *payload, size_t len,
		   const struct json_obj_descr *descr, size_t descr_len,
		   void *val)

which means the input json string should not be modified.

Impact
Sometimes we'd like to proceed a json string twice:

1. the first time only proceed the first filed to know the info of this json
2. according the info we parse again with specific json_obj_descr
   If json_obj_parse will modify string, user have to create a copy of string .

Environment (please complete the following information):

• OS: Linux
• Toolchain: Zephyr SDK
• Commit SHA or Version used: f05cbb4

[mbolivar]

Is this really a bug? I think this is the library working as designed. Otherwise it would have to duplicate the string.

[dcpleung]

(Not sure why I was assigned)

However, I agree with @mbolivar that this is not exactly a bug. Parsing JSON and returning an hierarchy of JSON objects is memory intensive (not to mention the manual memory allocation handling). Since Zephyr is targeting embedded devices with limited memory, parsing in-place with modifications is an acceptable compromise.

If you need to parse the string twice, could you try duplicating the string first to extract information for your first step? You can release the memory for the duplicated string, and then proceed to parse the original string.

[mbolivar]

@jhedberg I think the labels need a -bug +question and this issue can be closed.

[jhedberg]

(Not sure why I was assigned)

@dcpleung unfortunately I can't remember, but seems like a mistake - sorry about that.

@jhedberg I think the labels need a -bug +question and this issue can be closed.

@mbolivar agreed.