crunch_json_odm.py

import json
from pprint import pprint
from datetime import datetime

def load_json_data(path):
    start_time = datetime.now()
    f = open(path, 'r')

    # stripping out data contaminants from Crunchbase
    first_line = f.readline()
    decoded = first_line #.decode('utf-8')
    start_index = decoded.find('{')
    end_index = decoded.rfind('}')


    clean_string = decoded[start_index:(end_index + 1)]

    opening_count = clean_string.count('{')
    closing_count = clean_string.count('}')
    escaped_open = clean_string.count('\{')
    escaped_closed = clean_string.count('\}')
    print opening_count
    print closing_count
    data = clean_string
    # data = json.loads(clean_string)
    end_time = datetime.now()
    f.close()

    return {"start_time": str(start_time), "end_time": str(end_time), "data": data}

results = load_json_data('odm.json')
print results["start_time"]
print results["end_time"]
print results['data'][:50]
print results['data'][-50:]

# OUTPUT:
# 672351  <- '{' count
# 672348  <- '}' count. Why is this not the same?
# 2015-09-07 12:28:13.457119
# 2015-09-07 12:28:18.955385
# {"root":[{"crunchbase_uuid":"ed13cd36fe2b3707197b0
# sts worldwide for technologically cutting-edge"}]}