Using Bulk 2.0

You can use this library to access Bulk 2.0 API functions.

Create new records:

"Custom_Id__c","AccountId","Email","FirstName","LastName"
"CustomID1","ID-13","contact1@example.com","Bob","x"
"CustomID2","ID-24","contact2@example.com","Alice","y"
...

sf.bulk2.Contact.insert(data="./sample.csv", batch_size=10000)

Create new records concurrently:

sf.bulk2.Contact.insert(data="./sample.csv", batch_size=10000, concurrency=10)

Update existing records:

"Custom_Id__c","AccountId","Email","FirstName","LastName"
"CustomID1","ID-13","contact1@example.com","Bob","X"
"CustomID2","ID-24","contact2@example.com","Alice","Y"
...

sf.bulk2.Contact.update(data="./sample.csv")

Upsert records from csv:

"Custom_Id__c","LastName"
"CustomID1","X"
"CustomID2","Y"
...

sf.bulk2.Contact.upsert(data="./sample.csv", external_id_field='Custom_Id__c')

Upsert records from dict:

data = [
      {'Custom_Id__c': 'CustomID1', 'LastName': 'X'},
      {'Custom_Id__c': 'CustomID2', 'LastName': 'Y'}
    ]

sf.bulk2.Contact.upsert(data=df.to_dict(orient='records'), external_id_field='Custom_Id__c')

Query records:

query = 'SELECT Id, Name FROM Account LIMIT 100000'

results = sf.bulk2.Account.query(
    query, max_records=50000, column_delimiter="COMMA", line_ending="LF"
)
for i, data in enumerate(results):
    with open(f"results/part-{i}.csv", "w") as bos:
        bos.write(data)

Download records(low memory usage):

query = 'SELECT Id, Name FROM Account'

sf.bulk2.Account.download(
    query, path="results/", max_records=200000
)

Delete records (soft deletion):

"Id"
"0000000000AAAAA"
"0000000000BBBBB"
...

sf.bulk2.Contact.delete(data="./sample.csv")

Hard deletion:

sf.bulk2.Contact.hard_delete(data="./sample.csv")

Retrieve failed/successful/unprocessed records for ingest(insert,update…) job:

results = sf.bulk2.Contact.insert(data="./sample.csv")
# [{"numberRecordsFailed": 123, "numberRecordsProcessed": 2000, "numberRecordsTotal": 2000, "job_id": "Job-1"}, ...]
for result in results:
    job_id = result['job_id']
    # also available: get_unprocessed_records, get_successful_records
    data = sf.bulk2.Contact.get_failed_records(job_id)
    # or save to file
    sf.bulk2.Contact.get_failed_records(job_id, path=f'{job_id}.csv')