Import Kaggle csv from download url to pandas Data

问题:

I've been trying different methods to import the SpaceX missions csv file on Kaggle directly into a pandas DataFrame, without any success.

I'd need to send requests to login. This is what I have so far:

import requests
import pandas as pd
from io import StringIO

# Link to the Kaggle data set & name of zip file
login_url = 'http://www.kaggle.com/account/login?ReturnUrl=/spacex/spacex-missions/downloads/database.csv'

# Kaggle Username and Password
kaggle_info = {'UserName': "user", 'Password': "pwd"}

# Login to Kaggle and retrieve the data.
r = requests.post(login_url, data=kaggle_info, stream=True)
df = pd.read_csv(StringIO(r.text))

r is returning the html content of the page. df = pd.read_csv(url) gives a CParser error: CParserError: Error tokenizing data. C error: Expected 1 fields in line 13, saw 6

I've searched for a solution, but so far nothing I've tried worked.

回答1:

You are creating a stream and passing it directly to pandas. I think you need to pass a file like object to pandas. Take a look at this answer for a possible solution (using post and not get in the request though).

Also i think the login url with redirect that you use is not working as it is. I know i suggested that here. But i ended up not using is because the post request call did not handle the redirect (i suspect).

The code i ended up using in my project was this:

def from_kaggle(data_sets, competition):
    """Fetches data from Kaggle

    Parameters
    ----------
    data_sets : (array)
        list of dataset filenames on kaggle. (e.g. train.csv.zip)

    competition : (string)
        name of kaggle competition as it appears in url
        (e.g. 'rossmann-store-sales')

    """
    kaggle_dataset_url = "https://www.kaggle.com/c/{}/download/".format(competition)

    KAGGLE_INFO = {'UserName': config.kaggle_username,
                   'Password': config.kaggle_password}

    for data_set in data_sets:
        data_url = path.join(kaggle_dataset_url, data_set)
        data_output = path.join(config.raw_data_dir, data_set)
        # Attempts to download the CSV file. Gets rejected because we are not logged in.
        r = requests.get(data_url)
        # Login to Kaggle and retrieve the data.
        r = requests.post(r.url, data=KAGGLE_INFO, stream=True)
        # Writes the data to a local file one chunk at a time.
        with open(data_output, 'wb') as f:
            # Reads 512KB at a time into memory
            for chunk in r.iter_content(chunk_size=(512 * 1024)):
                if chunk: # filter out keep-alive new chunks
                    f.write(chunk)

Example use:

sets = ['train.csv.zip',
        'test.csv.zip',
        'store.csv.zip',
        'sample_submission.csv.zip',]
from_kaggle(sets, 'rossmann-store-sales')

You might need to unzip the files.

def _unzip_folder(destination):
    """Unzip without regards to the folder structure.

    Parameters
    ----------
    destination : (str)
        Local path and filename where file is should be stored.
    """
    with zipfile.ZipFile(destination, "r") as z:
        z.extractall(config.raw_data_dir)

So i never really directly loaded it into the DataFrame, but rather stored it to disk first. But you could modify it to use a temp directory and just delete the files after you read them.