Project Setup

• Project Setup
  • Data Modelling Principles & Guidelines
    • Be Analyst Friendly
    • Be Subject-Oriented
    • Be Relevant
    • Be Cost Efficient
    • Be Easy to Maintain
    • Avoid complex dependencies
  • Enforcing Code Quality
    • SQL Linting
    • YAML Linting
    • pre-commit hooks
    • dbt pre-commit hooks
  • Setting up Local Testing Environments

---

This project sets up the data modelling and day-to-day-operations of theLook e-commerce DWH leveraging:

• dbt-core

• BigQuery

• Cloud Composer

• Google Cloud Provider for Terraform

---

Data Modelling Principles & Guidelines

The DWH transformations of theLook e-commerce data were architected under the following principles and guidelines

Be Analyst Friendly

• Analysts shouldn't have to do multiple joins to retrieve meaningful data

Be Subject-Oriented

• Tables are organized around major topics of interest, such as customers, products, orders

• Each subject represents One-Big-Table with nested arrays and structs

  • child objects should never be orphans
  • child objects will always be queried within the context of the parent object

Be Relevant

• Data should reflect how current underlying platform functions

• Data should reflect the topics of interest to business

Be Cost Efficient

• Only process pieces of information that have changed

• Avoid scanning too much data per run

Be Easy to Maintain

• Backfilling historical data should be possible via the scheduled run without the need for extra code adjustments

• Changes in data should be easy to trace and audit

Avoid complex dependencies

• Processing by topic instead of monolitic schedules of all topics together

---

Enforcing Code Quality

The following linters are in place

• SQL linting with custom configuration for .sqlfluff

• YAML linting with custom configuration for .yamllint

• Python linting with default configuration via pylint

• Markdown linting with default configuration with pymarkdownlint

SQL Linting

To see if your SQL is compliant to the defined standard, you can run the following commands

# lint a specific file
sqlfluff lint path/to/file.sql

# lint a file directory
sqlfluff lint directory/of/sql/files

# let the linter fix your code
sqlfluff fix folder/model.sql

• SQL linting (and fixing) is enforced via pre-commit hooks for sqlfluff

YAML Linting

# check which files will be linted by default
yamllint --list-files .

# lint a specific file
yamllint my_file.yml

# OR
yamllint .

pre-commit hooks

Pre-commit have been set up in this repo to check and fix for:

• missing lines at the end
• trailing whitespaces
• violations of sql standards
• errors in yaml syntax

dbt pre-commit hooks

dbt pre-commit hooks have been set up to check that:

• there are no compilation errors

• no dbt script is directly referring to a table

• script contains only existing sources or macros

• no semi-colons have been forgotten at the end of sql queries

Hence, when working with the repo, make sure you've got the pre-commit installed so that they run upon your every commit

# install the githook scripts
pre-commit install

# run against all existing files
pre-commit run --all-files

---

Setting up Local Testing Environments

• Airflow: 2.9.3

• dbt-core: 1.8.2

• GCP infra