Python - 从数据框创建嵌套 JSON

Question

I have a dataframe which looks like this:

                      key         text
0                    title  Lorem ipsum
1                   header  Lorem ipsum
2              description  Lorem ipsum
.
.
.
.
10            pyramid.male  Lorem ipsum
11    pyramid.male_surplus  Lorem ipsum
12          pyramid.female  Lorem ipsum
13  pyramid.female_surplus  Lorem ipsum
.
.
.
.
29    jitterplot.title1          Lorem ipsum
30    jitterplot.metric_1.label  Lorem ipsum
31  jitterplot.metric_1.tooltip  Lorem ipsum
32    jitterplot.metric_2.label  Lorem ipsum
33  jitterplot.metric_2.tooltip  Lorem ipsum

The keys represent keys in a JSON file. The JSON structure should look like the following:

{
  "title": "Lorem ipsum",
  "header": "Lorem ipsum",
  "description": "Lorem ipsum",

  "pyramid": {
    "male": "Lorem ipsum",
    "male_surplus": "Lorem ipsum",
    "female": "Lorem ipsum",
    "female_surplus": "Lorem ipsum"
  },

  "jitterplot": {
    "title1": "Lorem ipsum",
    "metric_1": {
      "label": "Lorem ipsum",
      "tooltip": "Lorem ipsum"
    },
    "metric_2": {
      "label": "Lorem ipsum",
      "tooltip": "Lorem ipsum"
    }
  }
}

Meaning, a . in the key column represents a nested level.

Is there a 'Pythonic' way to achieve this? Currently, I'm just hacking it by manually writing each row to a text file with a custom parser I wrote. But obviously this is not very scalable.

I've prepared a sample CSV which you can read, and added some additional columns if they help. Use the following code:

import pandas as pd

url = 'https://raw.githubusercontent.com/Thevesh/Display/master/i18n_sample.csv'
df = pd.read_csv(url)

df['n_levels'] = df['key'].str.count('\.') # column with number of levels
max_levels = df.n_levels.max() # 
df = df.join(df['key'].str.split('.',expand=True))
df.columns = list(df.columns)[:-max_levels-1] + ['key_' + str(x) for x in range(max_levels+1)] 

Answer 1

def autonesting_dict():
    return collections.defaultdict(autonesting_dict)

json_dict = autonesting_dict()

key, value = 'jitterplot.metric_2.tooltip', "Lorem ipsum"
subkeys = key.split('.')

nested_dict = functools.reduce(lambda d, key: d[key], subkeys[:-1], json_dict)
nested_dict[subkeys[-1]] = value

The above will make it so that:

json_dict['jitterplot']['metric_2']['tooltip']  # 'Lorem ipsum'

Just repeat for all rows.

---

Sidenote regarding:

I've prepared a sample CSV which you can read, and added some additional columns if they help. Use the following code:

Maybe it's just me, but that sounds like something that might be given on an assignment or quiz, not like someone asking for assistance.

Answer 2

This seems like a good fit for a recursive function:

# Dataframe with columns key and value:
df = ...
json_data = {}

def set_value(nested_dict, keys, value):
    if len(keys) == 1:
        nested_dict[keys[0]] = value
        return
    if keys[0] not in nested_dict:
        nested_dict[keys[0]] = {}
    set_value(nested_dict[keys[0]], keys[1:], value)

for full_key, value in zip(df.key, df.text):
    keys = full_key.split('.')
    set_value(json_data, keys, value)

print(json_data)

Answer 3

Similarly but a bit simpler than the other answers:

def set_nested_value(d, keys, value):
    for key in keys[:-1]:
        d = d.setdefault(key, {})
    d[keys[-1]] = value
    
result = {}
for _, row in df.iterrows():
    set_nested_value(result, row["key"].split("."), row["text"])