dq-module 1.1.7

data profiling and basic data quality rules check

dq-module

dq-module is a tool which can be used to perform validations and profiling on the datasets.This tool is compatible with two run_engines pyspark and polars.

Features

1. Data Validation

This library contains a SingleDatasetQualityCheck() class which can used to validate the dataset against a defined set of rules. This class contains the following rules which can be used to validate the dataset:

• null_check
• schema_check
• range_min_check
• range_max_check
• length_check
• unique_records_check
• unique_keys_check
• allowed_values_check
• min_length_check
• column_count_check
• not_allowed_values_check
• date_range_check
• regex_check
• row_count_check

How To Use

import dataqualitycheck as dq
from datetime import date
import time

Step 1: Adding a datasource

We have 4 classes which can be used to connect to a datasource:

1. AzureBlobDF() - This class can be used to interact with the datasources on azure blob storage .
2. DatabricksFileSystemDF() - This class can be used to interect with the datasources on databricks filesystem.
3. DatabricksSqlDF() - This class can be used to interact with the datasources on databricks databases.
4. LocalFileSystemDF() - This class can be used to interact with the datasources on your local filesystem.

Each of the above class provides the functionalities to read and write from the respective datasources.

Example: Pass the configuration of the blob connector in blob_connector_configand add a datasource by defining a data_read_ob and data_write_ob.

blob_connector_config = {"storage_account_name" : "<storage-account-name>", "container_name" : "cooler-images", "sas_token" : <vaild-sas-token>}

data_read_ob = dq.AzureBlobDF(storage_name=blob_connector_config["storage_account_name"], sas_token=blob_connector_config["sas_token"])
data_write_ob = dq.DatabricksFileSystemDF()

tables_list is a dictionary that contains the list of sources along with the container_name , source_type , layer , source_name , filename , read_connector_method and latest_file_path for the tables on which the validations has to be applied .

#This is optional.It is required when you are calling individual rules.
tables_list = {}

Step 2: Add a DataContext

Instantiate a DataContext by passing tables_list,interaction_between_tables,data_read_ob,data_write_ob, data_right_structure,job_id,time_zone,no_of_partition and output_db_name . You can also pass the run_engine with which you want to apply the quality checks. By default, the run_engine is pyspark.

dq_ob = dq.SingleDatasetQualityCheck(tables_list={}, 
                                     interaction_between_tables=[],  
                                     data_read_ob=data_read_ob, 
                                     data_write_ob=data_write_ob, 
                                     data_right_structure='file',
                                     job_id=<pipeline_run_id>,
                                     time_zone=None,
                                     output_db_name=<database_name_where_report_has_to_be_written>,
                                     no_of_partition=4)

Step 3:

Passing a rules_diagnosys_summery_file_path and config_df as an input and apply validations on various columns of respective table defined in the config_df.

Here is an sample of the config_df.

container_name	source_type	layer	source_name	filename	rule_name	column_to_be_checked	value	date_column_config	date_format_dictionary	ruletype	active	read_connector_method	latest_file_path	output_folder_structure	failed_schema_source_list
		processed			null_check		null	null	null	Mandatory	1	blob	path-to-file	processed/
		processed			range_min_check		10	null	null	Not Mandatory	1	blob	path-to-file	processed/
		processed			range_max_check		1000	null	null	Not Mandatory	1	blob	path-to-file	processed/

Example:

rules_diagnosys_summery_folder_path = <folder-path-for-the-output-report>
 
config_df = spark.read.option("header",True).csv(<path-to-the-config>)

dq_ob.apply_validation(config_df, write_summary_on_database=True, failed_schema_source_list=[], output_summary_folder_path=rules_diagnosys_summery_folder_path)

2. Data Profiling

• We can generate a detailed summary statistics such as mean, median, mode, list of uniques, missing count etc. of a dataset using the DataProfile() class.
• This class can also be used to recommend some data quality rules based on the profiling report generated on the dataset.

How To Use

Step 1: Add a Datasource

data_write_ob = dq.DatabricksSqlDF()
data_write_ob = dq.DatabricksSqlDF()

Step 2: Add a DataContext

import pytz
time_zone = pytz.timezone('US/Central')
dq_ob = dq.DataProfile(tables_list=tables_list,
                       interaction_between_tables=[],
                       data_read_ob=data_read_ob,
                       data_write_ob=data_write_ob,
                       data_right_structure='table',
                       job_id=<pipeline_run_id>,
                       time_zone=time_zone,
                       no_of_partition=4,
                       output_db_name=<database_name_where_report_has_to_be_written>,
                       run_engine='polars')

Step 3: Pass config_df as an input and apply data profiling on various columns of respective table defined in the config_df.

# You can create a config_df in pyspark/polars run_engine directly also rather than reading as a csv.
config_df = spark.createDataFrame([{"container_name" : "<blob-storage-container-name>",
                                    "source_type" : "<source-type>",
                                    "layer" : "raw",
                                    "source_name" : "<data-source>",
                                    "filename" : "<filename>",
                                    "latest_file_path" : "<new-filename>",  
                                    "read_connector_method" : "databricks sql",
                                    "output_folder_structure" : "<directory-path-to-store-result>"}])

# Generating a data profiling report.
dq_ob.apply_data_profiling(source_config_df=config_df, write_consolidated_report=True)

# Generating a data profiling report as well as recommending the quality rules based on the profiling report.
rules_config = dq_ob.data_profiling_based_quality_rules(config_df, list_of_columns_to_be_ignored)

3. Consistency Check

You can check the consistency of common columns between two tables using the ConsistencyCheck() class.

How To Use

Step 1: Add a datasource.

data_read_ob = dq.DatabricksFileSystemDF()
data_write_ob = dq.AzureBlobDF(storage_name=blob_connector_config["storage_account_name"], sas_token=blob_connector_config["sas_token"])

Step 2: Add a DataContext

dq_ob = dq.ConsistencyCheck(tables_list={}, 
                            interaction_between_tables=[],
                            data_read_ob=data_read_ob,
                            data_write_ob=data_write_ob, 
                            data_right_structure='file',
                            job_id=<pipeline_run_id>,
                            time_zone=None,
                            no_of_partition=4,
                            output_db_name=<database_name_where_report_has_to_be_written>)

Step 3: Pass config_df and output_report_folder_path as an input and apply consistency check. Here is a sample consistency check config.

container_name	base_table_source_type	base_table_layer	base_table_source_name	base_table_filename	base_table_col_name	base_table_file_path	mapped_table_source_type	mapped_table_layer	mapped_table_source_name	mapped_table_filename	mapped_table_col_name	mapped_table_file_path	read_connector_method	output_folder_structure
		processed						processed					dbfs	processed/data_consistenct_test/
		processed						processed					dbfs	processed/data_consistenct_test/

config_df = spark.read.option("header",True).csv(<path-to-the-consistency-check-config>)

output_report_folder_path = <folder-path-for-the-output-report>

dq_ob.apply_consistency_check(config_df, output_report_folder_path)