使用包含包含字典的列表的嵌套字典从字典中提取数据

Question

我有一个来自 api 的响应，其中包含来自加热系统的数据集，其结构为具有嵌套字典的字典，其中包含包含字典的列表。

例如

    sample = {"zoneType": "HEATING",
              "interval": {"from": "2020-10-23T22:45:00.000Z", "to": "2020-10-24T23:15:00.000Z"},
              "hoursInDay": 24,
              "measuredData": {
                  "measuringDeviceConnected": {
                      "timeSeriesType": "dataIntervals",
                      "valueType": "boolean",
                      "dataIntervals": [{
                          "from": "2020-10-23T22:45:00.000Z", "to": "2020-10-24T23:15:00.000Z", "value": True}]
                          },
                  "insideTemperature": {
                      "timeSeriesType": "dataPoints",
                      "valueType": "temperature",
                      "min": {
                          "celsius": 19.34,
                          "fahrenheit": 66.81},
                      "max": {
                          "celsius": 20.6,
                          "fahrenheit": 69.08},
                      "dataPoints": [
                          {"timestamp": "2020-10-23T22:45:00.000Z", "value": {"celsius": 20.6, "fahrenheit": 69.08}},
                          {"timestamp": "2020-10-23T23:00:00.000Z", "value": {"celsius": 20.55, "fahrenheit": 68.99}},
                          {"timestamp": "2020-10-23T23:15:00.000Z", "value": {"celsius": 20.53, "fahrenheit": 68.95}},
                          {"timestamp": "2020-10-23T23:30:00.000Z", "value": {"celsius": 20.51, "fahrenheit": 68.92}},
                          {"timestamp": "2020-10-23T23:45:00.000Z", "value": {"celsius": 20.48, "fahrenheit": 68.86}},
                          {"timestamp": "2020-10-24T00:00:00.000Z", "value": {"celsius": 20.48, "fahrenheit": 68.86}},
                          {"timestamp": "2020-10-24T00:15:00.000Z", "value": {"celsius": 20.44, "fahrenheit": 68.79}}]
                  },
                  "humidity": {
                      "timeSeriesType": "dataPoints",
                      "valueType": "percentage",
                      "percentageUnit": "UNIT_INTERVAL",
                      "min": 0.615,
                      "max": 0.627,
                      "dataPoints": [
                          {"timestamp": "2020-10-23T22:45:00.000Z", "value": 0.615},
                          {"timestamp": "2020-10-23T23:00:00.000Z", "value": 0.615},
                          {"timestamp": "2020-10-23T23:15:00.000Z", "value": 0.619},
                          {"timestamp": "2020-10-23T23:30:00.000Z", "value": 0.620},
                          {"timestamp": "2020-10-23T23:45:00.000Z", "value": 0.621},
                          {"timestamp": "2020-10-24T00:00:00.000Z", "value": 0.623},
                          {"timestamp": "2020-10-24T00:15:00.000Z", "value": 0.627}]
                  }
              }}

我想从上面提取 ['insideTemperature']['datapoints'] 时间戳和摄氏度值（实际数据跨越更多时间段）并将它们作为列放在新的 pd.DataFrame 中以及来自 '湿度'的关键。在适当的时候，我想将其与来自具有类似结构的单独 API 调用的数据合并，尽管可能没有一致的时间戳值。

许多顶级字典包含摘要数据（例如最小值和最大值），因此可以忽略。同样，如果需要，从摄氏度到 f 等的转换很容易，所以我不想提取这些数据。

什么是干净地创建一个数据文件的最佳方法，该文件列出了来自该查询的时间戳、摄氏温度和湿度，然后我可以将其与其他查询输出连接？

到目前为止，我一直在使用以下内容：

import pandas as pd
df = pd.DataFrame(sample['measuredData']['insideTemperature']['dataPoints'])

## remove column that contains dictionary data, leaving time data
df.drop(labels='value', axis=1, inplace=True)

## get temp data into new column
input_data_point = sample['measuredData']['insideTemperature']['dataPoints']

temps = []

for i in input_data_point:
    temps.append(i['value']['celsius'])

df['inside_temp_c'] = pd.DataFrame(temps)

## repeat for humidity
input_data_point = sample['measuredData']['humidity']['dataPoints']

temps = []

for i in input_data_point:
    temps.append(i['value'])

df['humidity_pct'] = pd.DataFrame(temps)

作为 python 编码的新手，我想知道是否有更快的方法从原始下载数据中提取数据，直接进入干净的 Pandas DataFrame？感谢您的任何建议。

Answer 1

您可以使用json_normalize获取数据：

df1 = pd.json_normalize(sample,
                       record_path=['measuredData', 'insideTemperature', 'dataPoints'],
                       meta=['zoneType'])
print(df1)
df2 = pd.json_normalize(sample,
                       record_path=['measuredData', 'humidity', 'dataPoints'],
                       meta=['zoneType'])
print(df2)

df1:

                 timestamp  value.celsius  value.fahrenheit zoneType
0  2020-10-23T22:45:00.000Z          20.60             69.08  HEATING
1  2020-10-23T23:00:00.000Z          20.55             68.99  HEATING
2  2020-10-23T23:15:00.000Z          20.53             68.95  HEATING
3  2020-10-23T23:30:00.000Z          20.51             68.92  HEATING
4  2020-10-23T23:45:00.000Z          20.48             68.86  HEATING
5  2020-10-24T00:00:00.000Z          20.48             68.86  HEATING
6  2020-10-24T00:15:00.000Z          20.44             68.79  HEATING

df2:

                  timestamp  value zoneType
0  2020-10-23T22:45:00.000Z  0.615  HEATING
1  2020-10-23T23:00:00.000Z  0.615  HEATING
2  2020-10-23T23:15:00.000Z  0.619  HEATING