Python client library for accessing the asynchronous Salesforce.com Bulk API.

pip install salesforce-bulk

To access the Bulk API you need to authenticate a user into Salesforce. The easiest way to do this is just to supply username, password and security_token. This library will use the simple-salesforce package to handle password based authentication.

from salesforce_bulk import SalesforceBulk

bulk = SalesforceBulk(username=username, password=password, security_token=security_token)
...

Alternatively if you run have access to a session ID and instance_url you can use those directly:

from urlparse import urlparse
from salesforce_bulk import SalesforceBulk

bulk = SalesforceBulk(sessionId=sessionId, host=urlparse(instance_url).hostname)
...

The basic sequence for driving the Bulk API is:

1. Create a new job
2. Add one or more batches to the job
3. Close the job
4. Wait for each batch to finish

bulk.create_query_job(object_name, contentType='JSON')

Using API v39.0 or higher, you can also use the queryAll operation:

bulk.create_queryall_job(object_name, contentType='JSON')

Example

import json
from salesforce_bulk.util import IteratorBytesIO

job = bulk.create_query_job("Contact", contentType='JSON')
batch = bulk.query(job, "select Id,LastName from Contact")
bulk.close_job(job)
while not bulk.is_batch_done(batch):
    sleep(10)

for result in bulk.get_all_results_for_query_batch(batch):
    result = json.load(IteratorBytesIO(result))
    for row in result:
        print row # dictionary rows

Same example but for CSV:

import unicodecsv

job = bulk.create_query_job("Contact", contentType='CSV')
batch = bulk.query(job, "select Id,LastName from Contact")
bulk.close_job(job)
while not bulk.is_batch_done(batch):
    sleep(10)

for result in bulk.get_all_results_for_query_batch(batch):
    reader = unicodecsv.DictReader(result, encoding='utf-8')
    for row in reader:
        print(row) # dictionary rows

Note that while CSV is the default for historical reasons, JSON should be prefered since CSV has some drawbacks including its handling of NULL vs empty string.

If you are querying a large number of records you probably want to turn on PK Chunking:

bulk.create_query_job(object_name, contentType='CSV', pk_chunking=True)

That will use the default setting for chunk size. You can use a different chunk size by providing a number of records per chunk:

bulk.create_query_job(object_name, contentType='CSV', pk_chunking=100000)

Additionally if you want to do something more sophisticated you can provide a header value:

bulk.create_query_job(object_name, contentType='CSV', pk_chunking='chunkSize=50000; startRow=00130000000xEftMGH')

All Bulk upload operations work the same. You set the operation when you create the job. Then you submit one or more documents that specify records with columns to insert/update/delete. When deleting you should only submit the Id for each record.

For efficiency you should use the post_batch method to post each batch of data. (Note that a batch can have a maximum 10,000 records and be 1GB in size.) You pass a generator or iterator into this function and it will stream data via POST to Salesforce. For help sending CSV formatted data you can use the salesforce_bulk.CsvDictsAdapter class. It takes an iterator returning dictionaries and returns an iterator which produces CSV data.

Full example:

from salesforce_bulk import CsvDictsAdapter

job = bulk.create_insert_job("Account", contentType='CSV')
accounts = [dict(Name="Account%d" % idx) for idx in xrange(5)]
csv_iter = CsvDictsAdapter(iter(accounts))
batch = bulk.post_batch(job, csv_iter)
bulk.wait_for_batch(job, batch)
bulk.close_job(job)
print("Done. Accounts uploaded.")

When creating the job, pass concurrency='Serial' or concurrency='Parallel' to set the concurrency mode for the job.