Python 3: How to upload a pandas dataframe as a csv stream without saving on disc?

Question

I want to upload a pandas dataframe to a server as csv file without saving it on the disc. Is there a way to create a more or less "fake csv" file which pretends to be a real file?

Here is some example code: First I get my data from a sql query and storing it as a dataframe. In the upload_ga_data function I want to have something with this logic

 media = MediaFileUpload('df',
                      mimetype='application/octet-stream',
                      resumable=False)

Full example:

from __future__ import print_function
from apiclient.discovery import build
from oauth2client.service_account import ServiceAccountCredentials
from googleapiclient.errors import HttpError
from apiclient.http import MediaFileUpload
import pymysql
import pandas as pd
con = x

ga_query = """
    SELECT XXXXX
    """

df = pd.read_sql_query(ga_query,con)

df.to_csv('ga_export.csv', sep=',', encoding='utf-8', index = False)

def upload_ga_data():
    try:
        media = MediaFileUpload('ga_export.csv',
                          mimetype='application/octet-stream',
                          resumable=False)
        daily_upload = service.management().uploads().uploadData(
                accountId=accountId,
                webPropertyId=webPropertyId,
                customDataSourceId=customDataSourceId,
                media_body=media).execute()
        print ("Upload was successfull")
    except TypeError as error:
      # Handle errors in constructing a query.
      print ('There was an error in constructing your query : %s' % error)

Answer 1

The required behavior is possible using stream :

to create a more or less "fake csv" file which pretends to be a real file

Python makes File Descriptor (with open ) and Stream (with io.StringIO ) behave similarly. Then anywhere you can use a file descriptor can also use a String Stream.

The easiest way to create a text stream is with open(), optionally specifying an encoding:

 f = open("myfile.txt", "r", encoding="utf-8") 

In-memory text streams are also available as StringIO objects:

 f = io.StringIO("some initial text data") 

The text stream API is described in detail in the documentation of TextIOBase.

In Pandas you can do it with any function having path_or_buf argument in its signature , such as to_csv :

DataFrame.to_csv( path_or_buf =None, sep=', ', na_rep='', float_format=None, columns=None, header=True, index=True, index_label=None, mode='w', encoding=None, compression=None, quoting=None, quotechar='"', line_terminator='\\n', chunksize=None, tupleize_cols=None, date_format=None, doublequote=True, escapechar=None, decimal='.')

Following code exports a dummy DataFrame in CSV format into a String Stream (not physical file, in-memory octet-stream):

import io
import pandas as pd

df = pd.DataFrame(list(range(10)))

stream = io.StringIO()
df.to_csv(stream, sep=";")

When you want to get access to the stream content, just issue:

>>> stream.getvalue()
';0\n0;0\n1;1\n2;2\n3;3\n4;4\n5;5\n6;6\n7;7\n8;8\n9;9\n'

It returns the content without having the need to use a real file.

Answer 2

Though the other answer is an excellent start, there may be some who are confused on how to complete op's whole task. Here is a way to go from writing a dataframe to a stream to preparing that stream for upload using Google apiclient.http module. A key difference from op's attempt is that I pass the stream itself to a MediaIOBaseUpload instead of a MediaFileUpload. The file is assumed to be utf-8 like OP's file. This runs fine for me until the media is being uploaded, then I have an error " self._fp.write(s.encode('ascii', 'surrogateescape')) UnicodeEncodeError: 'ascii' codec can't encode character '\xe9' in position 2313: ordinal not in range(128)"

import io
import pandas as pd
from googleapiclient.errors import HttpError

from apiclient.http import MediaIOBaseUpload  # Changed this from MediaFileUpload

df = pd.DataFrame(list(range(10)))

stream = io.StringIO()
# writing df to the stream instead of a file:
df.to_csv(stream, sep=',', encoding='utf-8', index = False)
try:
    media = MediaIOBaseUpload(stream,
                          mimetype='application/octet-stream',
                          resumable=False)

#### Your upload logic here using media just created ####

except HttpError as error:

    #### Handle your errors in uploading here ####

Because I have a unicode character, I developed the alternative code which accomplishes the same thing but can handle the unicode characters.

import io
import pandas as pd
from googleapiclient.errors import HttpError

from apiclient.http import MediaIOBaseUpload  # Changed this from MediaFileUpload

df = pd.DataFrame(list(range(10)))

records = df.to_csv(line_terminator='\r\n', index=False).encode('utf-8')
bytes = io.BytesIO(records)

try:
    media = MediaIOBaseUpload(bytes,
                          mimetype='application/octet-stream',
                          resumable=False)

#### Your upload logic here using media just created ####

except HttpError as error:

    #### Handle your errors in uploading here ####

Answer 3

I used:

from googleapiclient.http import MediaIoBaseUpload

versus @Katherine's:

from apiclient.http import MediaIOBaseUpload 

But other than that, @Katherine's alternative solution worked perfectly for me as I was developing a solution to write a dataframe to a csv file in Google Drive running from a Google Cloud Function.