锌JSON

打包将一些常见的文件格式编码/解码为 json

可通过pip install znjson

与pickle此相比，它允许将可读的 json 文件与一些序列化数据结合起来。

例子

import numpy as np
import json
import znjson

data = json.dumps(
    obj={"data_np": np.arange(2), "data": [x for x in range(10)]},
    cls=znjson.ZnEncoder,
    indent=4
)
_ = json.loads(data, cls=znjson.ZnDecoder)

生成的*.json文件是部分可读的，如下所示：

{
    "data_np": {
        "_type": "np.ndarray_small",
        "value": [
            0,
            1
        ]
    },
    "data": [
        0,
        1,
        2,
        3,
        4
    ]
}

自定义转换器

ZnJSON 允许您轻松添加自定义转换器。让我们为datetime.datetime.

from znjson import ConverterBase
from datetime import datetime

class DatetimeConverter(ConverterBase):
    """Encode/Decode datetime objects

    Attributes
    ----------
    level: int
        Priority of this converter over others.
        A higher level will be used first, if there
        are multiple converters available
    representation: str
        An unique identifier for this converter.
    instance:
        Used to select the correct converter.
        This should fulfill isinstance(other, self.instance)
        or __eq__ should be overwritten.
    """
    level = 100
    representation = "datetime"
    instance = datetime

    def _encode(self, obj: datetime) -> str:
        """Convert the datetime object to str / isoformat"""
        return obj.isoformat()
    def _decode(self, value: str) -> datetime:
        """Create datetime object from str / isoformat"""
        return datetime.fromisoformat(value)

这允许我们使用这个新的序列化程序：

znjson.register(DatetimeConverter) # we need to register the new converter first
json_string = json.dumps(dt, cls=znjson.ZnEncoder, indent=4)
json.loads(json_string, cls=znjson.ZnDecoder)

并会导致

{
    "_type": "datetime",
    "value": "2022-03-11T09:47:35.280331"
}