Skip to main content

Carol Python API and Tools

Project description


Getting Started

Run pip install pycarol to install the latest stable version from PyPI. Documentation for the latest release is hosted on readthedocs.

This will install the minimal dependencies. To install pyCarol with the dataframes dependencies use pip install pycarol[dataframe], or to install with dask+pipeline dependencies use pip install pycarol[pipeline,dask]

The options we have are: complete, dataframe, onlineapp, dask, pipeline

To install from source:

  1. pip install -r requirements.txt to install the minimal requirements;

  2. pip install -e . ".[dev]" to install the minimal requirements + dev libs;

  3. pip install -e . ".[pipeline]" to install the minimal requirements + pipelines dependencies;

  4. pip install -e . ".[complete]" to install all dependencies;

Initializing pyCarol

Carol is the main object to access pyCarol and all Carol’s APIs.

from pycarol import PwdAuth, Carol
carol = Carol(domain=TENANT_NAME, app_name=APP_NAME,
              auth=PwdAuth(USERNAME, PASSWORD), organization=ORGANIZATION)

where domain is the tenant name, app_name is the Carol’s app name, if any, auth is the authentication method to be used (using user/password in this case) and organization is the organization one wants to connect. Carols’s URL is build as

It is also possible to initialize the object with a token generated via user/password. This is useful when creating an online app that interacts with Carol

from pycarol import PwdKeyAuth, Carol
carol = Carol(domain=TENANT_NAME, app_name=APP_NAME,
              auth=PwdKeyAuth(pwd_auth_token), organization=ORGANIZATION)

Using API Key

To use API keys instead of username and password:

from pycarol import ApiKeyAuth, Carol

carol = Carol(domain=DOMAIN,
              connector_id=CONNECTOR, organization=ORGANIZATION)

In this case one changes the authentication method to ApiKeyAuth. Noticed that one needs to pass the connector_id too. An API key is always associated to a connector ID.

It is possible to use pyCarol to generate an API key

from pycarol import PwdAuth, ApiKeyAuth, Carol

carol = Carol(domain=TENANT_NAME, app_name=APP_NAME, organization=ORGANIZATION,
              auth=PwdAuth(USERNAME, PASSWORD), connector_id=CONNECTOR)
api_key = carol.issue_api_key()

print(f"This is a API key {api_key['X-Auth-Key']}")
print(f"This is the connector Id {api_key['X-Auth-ConnectorId']}")

To get the details of the API key you can do:

details = carol.api_key_details(APIKEY, CONNECTORID)

Finally, to revoke an API key:


Good practice using token

Never write in plain text your password/API token in your application. Use environment variables. pyCarol can use environment variables automatically. When none parameter is passed to the Carol constructor pycarol will look for:

  1. CAROLTENANT for domain

  2. CAROLAPPNAME for app_name

  3. CAROL_DOMAIN for environment

  4. CAROLORGANIZATION for organization

  5. CAROLAPPOAUTH for auth

  6. CAROLCONNECTORID for connector_id

  7. CAROLUSER for carol user email

  8. CAROLPWD for user password.

e.g., one can create a .env file like this:


and then

from pycarol import Carol
from dotenv import load_dotenv
load_dotenv(".env") #this will import these env variables to your execution.
carol = Carol()

Ingesting data

From both Staging Tables and Data Models (CDS Layer)

Use this method when you need to read most of the records and columns from the source.

from pycarol import Carol, Staging

staging = Staging(Carol())
df = staging.fetch_parquet(

From both Staging Tables and Data Models (BQ Layer)

Use this method when you need to read only a subset of records and columns or when data transformation is needed.

from pycarol import BQ, Carol

bq = BQ(Carol())
query_str = "SELECT * FROM stg_connectorname_table_name"
results = bq.query(query_str)

In case one needs a service account with access to BigQuery, the following code can be used:

from pycarol import Carol
from pycarol.bigquery import TokenManager

tm = TokenManager(Carol())
service_account = tm.get_token().service_account

After each execution of BQ.query, the BQ object will have an attribute called job. This attribute is of type bigquery.job.query.QueryJob and may be useful for monitoring/debug jobs.

PyCarol provides access to BigQuery Storage API also. It allows for much faster reading times, but with limited querying capabilities. For instance, only tables are readable, so ‘ingestion_stg_model_deep_audit’ is ok, but ‘stg_model_deep_audit’ is not (it is a view).

from pycarol import BQStorage, Carol

bq = BQStorage(Carol())
table_name = "ingestion_stg_model_deep_audit"
col_names = ["request_id", "version"]
restriction = "branch = '01'"
sample_size = 1000
df = bq.query(

From Data Models (RT Layer): Filter queries

Use this when you need low latency (only if RT layer is enabled).

from pycarol.filter import TYPE_FILTER, TERM_FILTER, Filter
from pycarol import Query
json_query = Filter.Builder() \
    .must(TYPE_FILTER(value='ratings' + "Golden")) \

FIELDS_ITEMS = ['mdmGoldenFieldAndValues.mdmaddress.coordinates']
query = Query(carol, page_size=10, print_status=True, only_hits=True,
              fields=FIELDS_ITEMS, max_hits=200).query(json_query).go()

The result will be 200 hits of the query json_query above, the pagination will be 10, that means in each response there will be 10 records. The query will return only the fields set in FIELDS_ITEMS.

The parameter only_hits = True will make sure that only records into the path $hits.mdmGoldenFieldAndValues will return. If one wants all the response use only_hits = False. Also, if your filter has an aggregation, one should use only_hits = False and get_aggs=True, e.g.,

from pycarol import Query
from pycarol.filter import TYPE_FILTER, Filter, CARDINALITY

json_query = Filter.Builder() \
    .must(TYPE_FILTER(value='datamodelname' + "Golden")) \
    .aggregation(CARDINALITY(name='cardinality', params = ["mdmGoldenFieldAndValues.taxid.raw"], size=40))\

query = Query(carol, get_aggs=True, only_hits=False)

From Data Models (RT Layer): Named queries

from pycarol import Query

named_query = 'revenueHist'  # named query name
params = {"bin":"1d","cnpj":"24386434000130"}  #query parameters to send.
results = Query(carol).named(named_query, params=params).go().results

It is possible to use all the parameters used in the filter query, i.e., only_hits , save_results, etc. For more information for the possible input parameters check the docstring.

What if one does not remember the parameters for a given named query?

named_query = 'revenueHist'  # named query name
> {'revenueHist': ['*cnpj', 'dateFrom', 'dateTo', '*bin']}  #Parameters starting by * are mandatory.

Sending data

The first step to send data to Carol is to create a connector.

from pycarol import Connectors
connector_id = Connectors(carol).create(name='my_connector', label="connector_label", group_name="GroupName")
print(f"This is the connector id: {connector_id}")

With the connector Id on hands we can create the staging schema and then create the staging table. Assuming we have a sample of the data we want to send.

from pycarol import Staging

json_ex = {"name":'Rafael',"email": {"type": "email", "email": ''} }

staging = Staging(carol)
staging.create_schema(staging_name='my_stag', data = json_ex,
                      crosswalk_name= 'my_crosswalk' ,crosswalk_list=['name'],

The json schema will be in the variable schema.schema. The code above will create the following schema:

  'mdmCrosswalkTemplate': {
    'mdmCrossreference': {
      'my_crosswalk': [
  'mdmFlexible': 'false',
  'mdmStagingMapping': {
    'properties': {
      'email': {
        'properties': {
          'email': {
            'type': 'string'
          'type': {
            'type': 'string'
        'type': 'nested'
      'name': {
        'type': 'string'
  'mdmStagingType': 'my_stag'

To send the data (assuming we have a json with the data we want to send).

from pycarol import Staging

json_ex = [{"name":'Rafael',"email": {"type": "email", "email": ''}   },
           {"name":'Leandro',"email": {"type": "email", "email": ''}   },
           {"name":'Joao',"email": {"type": "email", "email": ''}   },
           {"name":'Marcelo',"email": {"type": "email", "email": ''}   }]

staging = Staging(carol)
staging.send_data(staging_name = 'my_stag', data = json_ex, step_size = 2,
                 connector_id=connectorId, print_stats = True)

The parameter step_size says how many registers will be sent each time. Remember the the max size per payload is 5MB. The parameter data can be a pandas DataFrame.

OBS: It is not possible to create a mapping using pycarol. The Mapping has to be done via the UI


To log messages to Carol:

from pycarol import Carol, CarolHandler
import logging

logger = logging.getLogger(__name__)
carol = CarolHandler(Carol())

logger.debug('This is a debug message') #This will not be logged in Carol. Level is set to INFO'This is an info message')
logger.warning('This is a warning message')
logger.error('This is an error message')
logger.critical('This is a critical message')

These methods will use the current long task id provided by Carol when running your application. For local environments you need to set that manually first on the beginning of your code:

import os
os.environ['LONGTASKID'] = task_id

We recommend to log only INFO+ information in Carol. If no TASK ID is passed it works as a Console Handler.


We can use pyCarol to access the settings of your Carol App.

from pycarol.apps import Apps
app = Apps(carol)
settings = app.get_settings(app_name='my_app')

The settings will be returned as a dictionary where the keys are the parameter names and the values are the value for that parameter. Please note that your app must be created in Carol.

Useful Functions

  1. track_tasks: Track a list of tasks.

from pycarol import Carol
from pycarol.functions import track_tasks
carol = Carol()
def callback(task_list):
track_tasks(carol=carol, task_list=['task_id_1', 'task_id_2'], callback=callback)

Release process

  1. Open a PR with your change for master branch;

  2. Once approved, merge into master;

  3. In case there are any changes to the default release notes, please update them

Project details

Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pycarol-2.55.0.tar.gz (123.9 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page