A python function to read data from multiple csvs, each of them can have extra rows or columns or both between them. Also needs a primary column header called key_header

multiple_csv.py

from real_type import get_type
#real type can be taken from https://github.com/ankitml/real_type or https://pypi.python.org/pypi/real_type/0.1

def read_multiple_csv(files, key_header):
    """
    takes a list of filesnames and a key_header string to merge the files.
    Can merge if there are different columns in the files, different records in
    the files.
    Assumption - all the files should have the column named key_header
    example : files = ['grades.csv', 'previous_grades.csv']
    key_header = 'email'
    merged_data = read_multiple_csv(files, key_header)
    """
    data = {}
    # files = ['a.csv', 'b.csv', 'c.csv']
    # key_header = 'email'

    for filename in files:
        file_generator = open(filename)
        headers = file_generator.next().split(',')
        headers = [h.strip() for h in headers]

        for line in file_generator:
            splits = line.split(',')
            inner_dict = {}
            for k,header in enumerate(headers):
                #assumes header is a string. probably it is
                if header is not '':
                    inner_dict[header] = int(splits[k]) if get_type(splits[k]) is int else splits[k]
            data_key = inner_dict.pop(key_header)
            try:
                z = data[data_key].copy()
            except KeyError:
                z = {}
            z.update(inner_dict)
            data[data_key] = z
    return data
    
def write_combined_csv(data, key='id', file_name='combined.csv'):
    import csv
    headers = [key]
    headers.extend(data[data.keys()[0]].keys())
    list_data = []
    for key_value, inner_dict in data.items():
        inner_dict[key] = key_value
        list_data.append(inner_dict)
    with open(file_name, 'wb') as combined_file:
        dict_writer = csv.DictWriter(combined_file, headers)
        dict_writer.writeheader()
        dict_writer.writerows(list_data)