feat: implement get_by_schema

[CPunisher]

Continuation of #108

• array index needs specification Not support yet
• test_failed_4 is not pass because I just do one step of skip_one in this case
• object.get_mut(&key.deref()) could be a slow path
• document

[liuq19]

1. we can add the jsonPointer into the field as the index, and make it easy to write the nested schema, and also support the array index.

The example can as follows:

use sonic_rs::json;

fn main() {
    let schema = json!({
        "a": null, // default value is `null`
        "b": {
            "b1": {},
            "b2": "default string" // default value is string
        },
        "c": "/1", // jsonpointer is "/1"
    });

    let data = r#"
    {
        "a": {},
        "b": {
            "b1": 123
        },
        "c": [1, 2, 3],
        "d": "balabala..."
    }"#;

    // parse json data by schem, we can parse into the schema value inplace
    let got = sonic_rs::get_by_schema(data, schema).unwrap();
    assert_eq!(
        got,
        json!({
            "a": {},
            "b": {
                "b1": 123,
                "b2": "default string"
            },
            "c": 2 // The 1 index from the array
        })
    );
}

2. the skip one will not validate the trailing chars, so its result is expected.

[CPunisher]

"c": "/1", // jsonpointer is "/1"

I think we can use "/c/1": "default value" instead to keep the consistance of the semantic of member values.
That means:
"c": "default value" == "/c": "default value"
"/c/1": "default value" == array index query or object member query

Here "/" is like ./ in file system. It could be somehow misleading and contradict the RFC:

Evaluation of a JSON Pointer begins with a reference to the root value of a JSON document and completes with a reference to some value within the document. Each reference token in the JSON Pointer is evaluated sequentially.

But I think it's ok because It is allowed to start with any (nested) Value and query with pointer.

Another question is that should we expand the results or keep the pointers. For example, "/c/d": 1 => "c": { "d": 2 } or "/c/d": 2, but pay attention that the query with any array index is not expandable, like "/c/1": 1 => "c": [??, 2]. I prefer the latter.

[liuq19]

For the second question, if we do not expand it is a JSON pointer. The problem is how to name the key, the key is still the JSON pointer, such as "/c/d": 1 or the lastest name ``"d": 1` ? If the jsonpointer is only an index, the key name is?

Does the problem become complicated? Maybe we will not need to support the array index.

[liuq19]

I think we do not need to support the array index in the schema now.

the object.get_mut(&key.deref()) is slow

are there more details?

[CPunisher]

I think we do not need to support the array index in the schema now.

the object.get_mut(&key.deref()) is slow

are there more details?

Since the time complexity of object.get_mut is O(n), I use a HashMap to save the key value results.