Remove index from dataframe before converting to json with split orientation

Question

I am outputting a pandas dataframe to a json object using the following:

df_as_json = df.to_json(orient='split')

In the json object superfluous indexes are stored. I do no want to include these.

To remove them I tried

df_no_index = df.to_json(orient='records')
df_as_json = df_no_index.to_json(orient='split')

However I get a

AttributeError: 'str' object has no attribute 'to_json'

Is there a fast way to reorganize the dataframe so that is does not contain a separate index column during or prior to the .to_json(orient='split') call?

Answer 1

• import json module
• Convert to json with to_json(orient='split')
• Use the json module to load that string to a dictionary
• Delete the index key with del json_dict['index']
• Convert the dictionary back to json with json.dump or json.dumps

---

Demo

import json

df = pd.DataFrame([[1, 2], [3, 4]], ['x', 'y'], ['a', 'b'])

json_dict = json.loads(df.to_json(orient='split'))
del json_dict['index']
json.dumps(json_dict)

'{"columns": ["a", "b"], "data": [[1, 2], [3, 4]]}'

Answer 2

Since two years back pandas (>= v0.23.0) offers an index argument (only valid for orient='split' and orient='table'):

df = pd.DataFrame([[1, 2], [3, 4]], ['x', 'y'], ['a', 'b'])
df.to_json(orient='split', index=True)
# '{"columns":["a","b"],"index":["x","y"],"data":[[1,2],[3,4]]}'
df.to_json(orient='split', index=False)
# '{"columns":["a","b"],"data":[[1,2],[3,4]]}'

https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.to_json.html