csv-etl 0.0.1

No project description provided

csv-etl

A rules based approach to performing ETL operations on csv files.

Installation

pip install csv-etl

Quick Introduction

The goal of this project is to provide a re-usable way to perform ETL operations on csv files. This implementation takes a given row of data from a csv file and applies a set of rules to generate a new format of the data. A rule knows three things

• Where the original data is coming from (source)
• Where the new data is going to (target)
• The data type the new data is expected to conform to (output_type)
  • output_type can be the following options and will result in the correlating python types
    • Integer : int
    • Decimal : decimal.Decimal
    • String : str
    • Date : datetime.datetime

The root Rule class in this library contains those three properties.

The Rule class is subclassed, and the subclasses are where the logic for extracting/modifying the data takes place.

The current Rule subclasses are

• StaticRule
  • Use Case: Static value for every row
• SingleSourceRule
  • Use Case: Extract value from source column, place in new target column
• DateRule
  • Use Case: Gather date part values from multiple columns, generate date value

Usage

With a config file

from csv_etl.rules import load_rules_from_yaml
from csv_etl import CSVConverter

rules = load_rules_from_yaml("path/to/config/file")
csv_converter = CSVConverter(rules)
result = csv_converter.convert("path/to/csv/file")

Configuration examples can be found here

The basic structure of the configuration file is:

rules:
  -
    target: The new key
    operation: extract || static
    to_type: Integer || String || Decimal || Date
    source: Where to pull data from

Creating rules programmatically

from csv_etl.rules import (
  StaticRule,
  SingleSourceRule,
  DateRule
)
from csv_etl import CSVConverter

rules = [
  StaticRule(source='kg', target='Unit', output_type='String')
  SingleSourceRule(source='Order Number', target='OrderId', output_type='Integer'),
  DateRule(source={'day': 'Day', 'month': 'Month', 'year': 'Year'}, target='OrderDate', output_type='Date'),
]
csv_converter = CSVConverter(rules)
result = csv_converter.convert("path/to/csv_file")

# By default, the result is a json array. But you can receive is back as a csv string
result = csv_converter.convert("path/to/csv/file", to='csv')

# You can also write the result to a specified file
result = csv_converter.convert("path/to/csv/file", to='csv', outfile="path/to/outfile")

A step by step use case of this can be found here

Developing

Make Commands

# run test suite
pytest
# generate test coverage
test-cov
# generate docs
docs
# build distribution files
build-dist
# run in docker container
play

What Next?

Expand on Date Rule

The current implementation of the DateRule is fairly rigid. It only covers the use case of there being day/month/year columns in the given csv. It also assumes the data values are numbers and not string representations of the months like March

This could be expanded upon to

• Handle month strings
• Handle different date formats
  • YYYY-mm-dd
  • mm-dd-YYYY
  • dd-mm-YYYY

Calculation Rules

Introducing a new CalculationRule that takes

• a list of column names for its source
• an operation_type ("+", "-", "*", "/")

When executed, the rule would gather the data, perform the calculation and assign it to the target.

Better error handling

Currently the errors are just printed to the console. Might be nice to have an option to gather them and have them represented in the resulting data set somehow.