Convert json to pandas DataFrame

Question

I have a JSON file that has several objects, such as:

{"reviewerID": "bc19970fff3383b2fe947cf9a3a5d7b13b6e57ef2cd53abc52bb2dfedf5fb1cd", "asin": "a6ed402934e3c1138111dce09256538afb04c566edf37c16b9ba099d23afb764", "overall": 2.0, "helpful": {"nHelpful": 1, "outOf": 1}, "reviewText": "This remote, for whatever reason, was chosen by Time Warner to replace their previous silver remote, the Time Warner Synergy V RC-U62CP-1.12S. The actual function of this CLIKR-5 is OK, but the ergonomic design sets back remotes by 20 years. The buttons are all the same, there no separation of the number buttons, the volume and channel buttons are the same shape as the other buttons on the remote, and it all adds up to a crappy user experience. Why would TWC accept this as a replacement? I'm skipping this and paying double for a refurbished Synergy V.", "summary": "Ergonomic nightmare", "unixReviewTime": 1397433600} {"reviewerID": "3689286c8658f54a2ff7aa68ce589c81f6cae4c4d9de76fa0f66d5c114f79837", "asin": "8939d791e9dd035aa58da024ace69b20d651cea4adf6159d984872b44f663301", "overall": 4.0, "helpful": {"nHelpful": 21, "outOf": 22}, "reviewText": "This is a great truck GPS. I've tried others and nothing seems to come close to the Rand McNally TND-700.Excellent screen size and resolution. The audio is loud enough to be heard over road noise and the purr of my Kenworth/Cat engine. I've used it for the last 8,000 miles or so and it has only glitched once. Just restarted it and it picked up on my route right where it should have.Clean up the minor issues and this unit rates a solid 5.Rand McNally 528881469 7-inch Intelliroute TND 700 Truck GPS", "summary": "Great Unit!", "unixReviewTime": 1280016000} 

I am trying to convert it to a Pandas DataFrame using the following code:

 train_df = pd.DataFrame() count = 0; for l in open('train.json'): try: count +=1 if(count==20001): break obj1 = json.loads(l) df1=pd.DataFrame(obj1, index=[0]) train_df = train_df.append(df1, ignore_index=True) except ValueError: line = line.replace('\\','') obj = json.loads(line) df1=pd.DataFrame(obj, index=[0]) train_df = train_df.append(df1, ignore_index=True) 

However, it gives me "NaN" for the nested values, i.e. "useful" attribute. I want the result to be such that both keys of the nested attribute are a separate column.

EDIT:

PS: I use try / except because I have a "\" character in some objects that gives me a JSON decoding error.

Can anyone help? Is there any other approach that I can use?

Thanks.

Answer 1

Use json_normalize in a list of dictionaries, which works much faster on a large number of json objects.

 from pandas.io.json import json_normalize my_list = [] with open('train.json') as f: for line in f: line = line.replace('\\','') my_list.append(json.loads(line)) # avoid transposing if you want to keep keys as columns of the dataframe result_df = json_normalize(my_list).T 

Answer 2

to try:

 pd.concat([pd.Series(json.loads(line)) for line in open('train.json')], axis=1)