QA Lab · Cross-Stack Series · Stack 2

Pytest + Python
+ Allure Framework

Idiomatic Python automation framework demonstrating the pytest ecosystem patterns valued in data-engineering, DevOps, and Python-native organisations — built against the same QA Lab target as all Cross-Stack Series frameworks.

pytest 8.2 · Playwright-Python 1.44 · allure-pytest 2.13 · pytest-xdist 3.5 · Python 3.12

📊 Allure Report ⚡ CI/CD </> GitHub

📊

Allure Report

Rich Allure HTML report published to GitHub Pages after every CI run. Features @allure.feature / @allure.story / @allure.severity hierarchy, parametrized test breakdowns, and failure screenshots attached automatically.

allure-pytest 2.13 · GitHub Pages View Report →

⚡

CI/CD Pipeline

GitHub Actions workflow: smoke tests on every push, full regression on main with -n 4 parallel workers, scheduled Monday 09:00 UTC. mypy strict type check and ruff lint gate before test execution.

GitHub Actions · ubuntu-latest View Runs →

</>

Source Code

Full Python project: Page Objects with @allure.step, conftest.py fixture hierarchy, data/form_data.py dataclasses, pytest.ini + pyproject.toml configuration, GitHub Actions workflow.

Public · MIT License View Repository →

About This Project

The Purpose

Python is the dominant language in data engineering, ML infrastructure, DevOps tooling, and API-first organisations. Teams in these environments expect pytest fluency, type-annotated code, and Allure-grade reporting as table stakes — not as extras bolted on top.

This is Stack 2 of the Cross-Stack Series — the same QA Lab target tested with different frameworks to enable direct, evidence-based comparison. The framework demonstrates idiomatic Python automation: yield fixtures, conftest hierarchy, dataclasses for test data, and @pytest.mark.parametrize for data-driven coverage.

Key differentiator vs Stack 1 (Playwright TS): Python’s yield fixture model and @pytest.mark.parametrize produce more expressive, composable test code than TypeScript’s base.extend() equivalent — a real trade-off worth understanding per team context.

Architecture Constraints

Every decision reflects real Python ecosystem requirements:

conftest.py fixture hierarchy — root for browser/page, section-level for page objects
yield fixtures throughout — no setup/teardown class methods, idiomatic pytest
@allure.step on all Page Object methods — Allure report populated automatically
@pytest.mark.parametrize for data-driven tests — not manual loops or test.each
pytest-xdist -n auto — CPU-count parallelism, --dist=loadfile prevents context conflicts
mypy --strict passes with zero errors — full type annotation discipline
ruff 0.4+ replaces flake8 + isort + black in a single pass
No time.sleep() anywhere — all waits via Playwright expect() auto-wait

Test Coverage

13 QA Lab UI sections covered across 12 test files. Same automation surface targeted in every Cross-Stack Series framework.

Buttons

Click states, disabled state assertion

Forms

Validation, field interaction, @parametrize submit

Input Fields

Text, number, date, search, URL types

Checkboxes

Check/uncheck, disabled state

Radio Buttons

Selection, mutual exclusivity

Dropdowns

Single select, multi-select

Tables

Cell content, row count, edit action

Alerts & Modals

Open, confirm, cancel, dismiss

Dynamic Visibility

Checkbox-triggered panel reveal

Async Buttons

Loading → success/error transitions

IFrames

frame_locator() — same API as TS

Drag & Drop

drag_and_drop() mouse simulation

Slider

Value change via fill() assertion

Tech Stack

Python ecosystem stack — every tool justified by real-world team requirements. Zero proprietary or paid dependencies.

🐍

Python 3.12 LTS

Language — LTS, typing improvements, universally available in CI environments

✅

pytest 8.2

Test runner — fixture system, plugin ecosystem, industry standard for Python

🌐

Playwright-Python 1.44

Browser automation — same engine as Stack 1 TS — fair cross-stack comparison

📊

allure-pytest 2.13

Reporting — @allure.step, @allure.feature, failure screenshots, GitHub Pages deploy

⚡

pytest-xdist 3.5

Parallelism — -n auto CPU-count workers, --dist=loadfile for fixture safety

🔎

mypy 1.10 (strict)

Type checking — strict mode, zero errors enforced in CI before tests run

🔧

ruff 0.4

Linting — replaces flake8 + isort + black, ANN rules enforce type annotations

⚙️

GitHub Actions

CI/CD — smoke on push, full regression on main, Monday 09:00 UTC schedule

🤖

Claude Code

AI engineering accelerator — scaffolding, pattern review, CLAUDE.md specification

Performance Testing — Locust

Python-native load testing in the same virtual environment as the pytest suite. Locust is ecosystem-native to the Python stack — HttpUser with @task decorators, same venv, same CI job. No toolchain switch. Part of the Cross-Stack Performance Series.

📊

Locust HTML Report

Built-in HTML report generated via --html report.html on every CI run. Includes RPS timeline, response time percentiles, failure breakdown. Retained 14 days as GitHub Actions artifacts. FINDINGS.md documents CDN framing and SLO rationale.

Locust 2.28+ · --html report · GitHub Actions Artifacts View CI Artifacts →

⚡

CI Integration

Runs as a separate CI job after the pytest suite passes. Headless mode: --headless -u N -r R -t Ts. --exit-code-on-error 1 gates the build on SLO breach. Smoke on every push; baseline on main; weekly Monday schedule.

GitHub Actions · ubuntu-latest · actions/setup-python View Pipeline →

</>

Performance Source

Scripts in performance/locust/locustfiles/: smoke, baseline, and CDN cold/warm scenarios. catch_response=True for explicit per-response pass/fail. Interactive Web UI available at localhost:8089 for local exploration.

Python 3.12 · Public · MIT License View Source →

38ms

p95 Response

650ms

p99 Response

Error Rate

500ms

p95 SLO Target

Smoke 5 VU · 30s

Baseline 10 VU · 60s · main only

p99 note 650ms spike = CloudFront cold edge miss (warm ≤40ms)

Target QA Lab / CloudFront · SLO compliance

📊 View Full Performance Series →

Pytest + Python+ Allure Framework