[FIXED] Pandas converts string-typed JSON value to INT

Issue

I have list of objects as JSON. Each object has two properties: id(string) and arg(number).

When I use pandas.read_json(...), the resulting DataFrame has the id interpreted as number as well, which causes problems, since information is lost.

import pandas as pd

json = '[{ "id" : "1", "arg": 1 },{ "id" : "1_1", "arg": 2}, { "id" : "11", "arg": 2}]'

df = pd.read_json(json)

I'd expect to have a DataFrame like this:

     id     arg  
0   "1"      1
1   "1_1"    2
2   "11"     2

I get

     id     arg  
0     1      1
1    11      2
2    11      2

and suddenly, the once unique id is not so unique anymore.

How can I tell pandas to stop doing that?

My search so far only yielded results, where people where trying to achive the opposite - having columns of string beeing interpreted as numbers. I totally don't want to achive that in this case!

Solution

Use dtype parameter for preventing cast id to numbers:

df = pd.read_json(json, dtype={'id':str})

print (df)
    id  arg
0    1    1
1  1_1    2
2   11    2


print (df.dtypes)
id     object
arg     int64
dtype: object

Answered By - jezrael