How to attach large files to an email using Python - Gmail API

Question

I'm trying to send an email with an attachment (ideally multiple attachments) that are larger than 10 MB and smaller than the limit of 25 MB in total. The reason I mention 10 MB is because it seems to be the lower bound for when the normal way of attaching files stops working and you get Error 10053.

I've read in the documentation that the best way to do this would be by using the resumable upload method but I haven't been able to get it to work nor have I been able to find any good examples in Python. Most of the SO questions on this simply link back to the documentation which doesn't have a Python example or their code resulted in other errors.

I'm looking for an explanation in Python because I want to make sure I understand it correctly.

Questions I've looked through:

• Attaching a file using Resumable upload w/ Gmail API
• Gmail Api resumable upload Rest( attachment larger than 5MB)
• using /upload urls with Google API client
• How to upload large messages to Gmail
• Error 10053 When Sending Large Attachments using Gmail API
• Sending email via gmail & python
• MIMEMultipart, MIMEText, MIMEBase, and payloads for sending email with file attachment in Python

Code:

import base64
import json
import os
from email import utils, encoders
from email.message import EmailMessage
from email.mime import application, multipart, text, base, image, audio
import mimetypes

from apiclient import errors
from googleapiclient import discovery, http
from google.oauth2 import service_account

def send_email(email_subject, email_body, email_sender='[email protected]', email_to='', email_cc='', email_bcc='', files=None):

    # Getting credentials
    with open(os.environ.get('SERVICE_KEY_PASSWORD')) as f:
        service_account_info = json.loads(f.read())

    # Define which scopes we're trying to access
    SCOPES = ['https://www.googleapis.com/auth/gmail.send']

    # Setting up credentials using the gmail api
    credentials = service_account.Credentials.from_service_account_info(service_account_info, scopes=SCOPES)

    # This allows us to assign an alias account to the message so that the messages aren't coming from 'ServiceDriod-8328balh blah blah'
    delegated_credentials = credentials.with_subject(email_sender)

    # 'Building' the service instance using the credentials we've passed
    service = discovery.build(serviceName='gmail', version='v1', credentials=delegated_credentials)

    # Building out the email 
    message = multipart.MIMEMultipart()
    message['to'] = email_to
    message['from'] = email_sender
    message['date'] = utils.formatdate(localtime=True)
    message['subject'] = email_subject
    message['cc'] = email_cc
    message['bcc'] = email_bcc
    message.attach(text.MIMEText(email_body, 'html'))


    for f in files or []:
        mimetype, encoding = mimetypes.guess_type(f)

        # If the extension is not recognized it will return: (None, None)
        # If it's an .mp3, it will return: (audio/mp3, None) (None is for the encoding)
        # For an unrecognized extension we set mimetype to 'application/octet-stream' so it won't return None again. 
        if mimetype is None or encoding is not None:
            mimetype = 'application/octet-stream'
        main_type, sub_type = mimetype.split('/', 1)

        # Creating the attachement:
        # This part is used to tell how the file should be read and stored (r, or rb, etc.)
        if main_type == 'text':
            print('text')
            with open(f, 'rb') as outfile:
                attachement = text.MIMEText(outfile.read(), _subtype=sub_type)
        elif main_type == 'image':
            print('image')
            with open(f, 'rb') as outfile:
                attachement = image.MIMEImage(outfile.read(), _subtype=sub_type)
        elif main_type == 'audio':
            print('audio')
            with open(f, 'rb') as outfile:
                attachement = audio.MIMEAudio(outfile.read(), _subtype=sub_type)          
        elif main_type == 'application' and sub_type == 'pdf':   
            with open(f, 'rb') as outfile:
                attachement = application.MIMEApplication(outfile.read(), _subtype=sub_type)
        else:                              
            attachement = base.MIMEBase(main_type, sub_type)
            with open(f, 'rb') as outfile:
                attachement.set_payload(outfile.read())

        encoders.encode_base64(attachement)
        attachement.add_header('Content-Disposition', 'attachment', filename=os.path.basename(f))
        message.attach(attachement)



    media_body = http.MediaFileUpload(files[0], chunksize=500, resumable=True)
    print('Uploading large file...')
    body = {'raw': base64.urlsafe_b64encode(message.as_bytes()).decode()}


    message = (service.users().messages().send(userId='me', body=body, media_body=media_body).execute())

Note: Right now, in the MediaFileUpload I'm using files[0] because I'm only using one file for testing and I just wanted to attach one file for now until it works.

Error:

Exception has occurred: ResumableUploadError
<HttpError 400 "Bad Request">
  File "C:\Users\CON01599\AppData\Local\Continuum\anaconda3\Lib\site-packages\googleapiclient\http.py", line 927, in next_chunk
    raise ResumableUploadError(resp, content)
  File "C:\Users\CON01599\AppData\Local\Continuum\anaconda3\Lib\site-packages\googleapiclient\_helpers.py", line 130, in positional_wrapper
    return wrapped(*args, **kwargs)
  File "C:\Users\CON01599\AppData\Local\Continuum\anaconda3\Lib\site-packages\googleapiclient\http.py", line 822, in execute
    _, body = self.next_chunk(http=http, num_retries=num_retries)
  File "C:\Users\CON01599\AppData\Local\Continuum\anaconda3\Lib\site-packages\googleapiclient\_helpers.py", line 130, in positional_wrapper
    return wrapped(*args, **kwargs)
  File "C:\Users\CON01599\Documents\GitHub\pipelines\components\email\send_email.py", line 105, in send_email
    message = (service.users().messages().send(userId='me', body=body, media_body=media_body).execute())

Answer:

import base64
import io
import json
import os
from email import utils, encoders
from email.message import EmailMessage
from email.mime import application, multipart, text, base, image, audio
import mimetypes

from apiclient import errors
from googleapiclient import discovery, http
from google.oauth2 import service_account


def get_environment_variables():
    """ Retrieves the environment variables and returns them in
        a dictionary object.
    """
    env_var_dict = {
        'to': os.environ.get('TO'),
        'subject': os.environ.get('SUBJECT'),
        'body': os.environ.get('BODY'),
        'file': os.environ.get('FILE')
    }

    return env_var_dict


def send_email(email_subject, email_body, email_sender='[email protected]', email_to='', email_cc='', email_bcc='', files=None):

    # Pulling in the string value of the service key from the parameter
    with open(os.environ.get('SERVICE_KEY_PASSWORD')) as f:
        service_account_info = json.loads(f.read())

    # Define which scopes we're trying to access
    SCOPES = ['https://www.googleapis.com/auth/gmail.send']

    # Setting up credentials using the gmail api
    credentials = service_account.Credentials.from_service_account_info(service_account_info, scopes=SCOPES)
    # This allows us to assign an alias account to the message so that the messages aren't coming from 'ServiceDriod-8328balh blah blah'
    delegated_credentials = credentials.with_subject(email_sender)
    # 'Building' the service instance using the credentials we've passed
    service = discovery.build(serviceName='gmail', version='v1', credentials=delegated_credentials)

    # Building out the email 
    message = multipart.MIMEMultipart()
    message['to'] = email_to
    message['from'] = email_sender
    message['date'] = utils.formatdate(localtime=True)
    message['subject'] = email_subject
    message['cc'] = email_cc
    message['bcc'] = email_bcc
    message.attach(text.MIMEText(email_body, 'html'))


    for f in files or []:
        f = f.strip(' ')
        mimetype, encoding = mimetypes.guess_type(f)

        # If the extension is not recognized it will return: (None, None)
        # If it's an .mp3, it will return: (audio/mp3, None) (None is for the encoding)
        # For an unrecognized extension we set mimetype to 'application/octet-stream' so it won't return None again. 
        if mimetype is None or encoding is not None:
            mimetype = 'application/octet-stream'
        main_type, sub_type = mimetype.split('/', 1)

        # Creating the attachement:
        # This part is used to tell how the file should be read and stored (r, or rb, etc.)
        if main_type == 'text':
            print('text')
            with open(f, 'rb') as outfile:
                attachement = text.MIMEText(outfile.read(), _subtype=sub_type)
        elif main_type == 'image':
            print('image')
            with open(f, 'rb') as outfile:
                attachement = image.MIMEImage(outfile.read(), _subtype=sub_type)
        elif main_type == 'audio':
            print('audio')
            with open(f, 'rb') as outfile:
                attachement = audio.MIMEAudio(outfile.read(), _subtype=sub_type)          
        elif main_type == 'application' and sub_type == 'pdf':   
            with open(f, 'rb') as outfile:
                attachement = application.MIMEApplication(outfile.read(), _subtype=sub_type)
        else:                              
            attachement = base.MIMEBase(main_type, sub_type)
            with open(f, 'rb') as outfile:
                attachement.set_payload(outfile.read())

        encoders.encode_base64(attachement)
        attachement.add_header('Content-Disposition', 'attachment', filename=os.path.basename(f))
        message.attach(attachement)

    media_body = http.MediaIoBaseUpload(io.BytesIO(message.as_bytes()), mimetype='message/rfc822', resumable=True)
    body_metadata = {} # no thread, no labels in this example

    try:
        print('Uploading file...')
        response = service.users().messages().send(userId='me', body=body_metadata, media_body=media_body).execute()
        print(response)
    except errors.HttpError as error:
        print('An error occurred when sending the email:\n{}'.format(error))


if __name__ == '__main__':

    env_var_dict = get_environment_variables()
    print("Sending email...")
    send_email(email_subject=env_var_dict['subject'], 
            email_body=env_var_dict['body'], 
            email_to=env_var_dict['to'],
            files=env_var_dict['file'].split(','))

    print("Email sent!")

Answer 1

The issue you're having here is that your MediaUpload is a single attachment.

Instead of uploading a single attachment as a resumable MediaUpload, you need to upload the entire RFC822 message as a resumable MediaUpload.

In other words:

import ...
...
from io import BytesIO
from googleapiclient.http import MediaIoBaseUpload

SCOPES = [ 'scopes' ]

creds = get_credentials_somehow()
gmail = get_authed_service_somehow()

msg = create_rfc822_message(headers, email_body)
to_attach = get_attachment_paths_from_dir('../reports/tps/memos/2019/04')
add_attachments(msg, to_attach)

media = MediaIoBaseUpload(BytesIO(msg.as_bytes()), mimetype='message/rfc822', resumable=True)
body_metadata = {} # no thread, no labels in this example
resp = gmail.users().messages().send(userId='me', body=body_metadata, media_body=media).execute()
print(resp)
# { "id": "some new id", "threadId": "some new thread id", "labelIds": ["SENT"]}

I pieced this together from your provided code, reviewing this GitHub issue and Google's Inbox-to-Gmail email importer, specificially this bit.

When sending replies to existing messages, you will almost certainly have some sort of metadata that you should provide to help Gmail keep track of your new response and the original conversation. Namely, instead of an empty body parameter, you would pass informative metadata such as

body_metadata = { 'labelIds': [
                    "your label id here",
                    "another label id" ],
                  'threadId': "some thread id you took from the message you're replying to"
                }

Other good refs:

• API Client's Gmail PyDoc
• Actual code used