symonk / pytest-playwright-enhanced

:snake: Batteries included playwright for pytest [alpha]

Hi There! 👋

---

I am currently focusing on:

• selenium - Working on the python bindings.
• asserto - Python assertions made quick & easy.
• pytest-containers - A pytest plugin for starting docker compose services around tests.
• kafkapy - A python CLI and set of utilities for inspection and managing kafka clusters.
• pytest-playwright-enhanced - A batteries included framework of playwright for pytest.
• tasq - A scalable worker pool for distributed tasks in golang.
• toodoo - An open source task/todo management system written in go/nextjs.

---

My Learning Roadmap for 2024:

• javascript. Repository
• go. Repository
• sql. Repository
• Achieve AWS cloud practitioner certification.
• continue advancing knowledge of docker.
• continue advancing knowledge of python.
• continue advancing knowledge of golang.

---

Penetration testing is a core hobby of mine and I enjoy finding vulnerabilities particularly in web applications and APIs.

pytest-playwright-enhanced's Issues

ensure `--artifacts` directory is a folder

ensuring a trailing / and is actually a folder etc.
Add some error handling to places we are using this, many things can go wrong
file lengths on windows, general errors
does running containerised make this more challenging?

manage auto installation of playwright drivers if requested

video on fail fails with xdist

reproducable:

def test_something(pw_page, pw_multi_browser):
    pw_page.goto("https://www.google.co.uk")
    assert False

pytest . --headed --browser webkit --browser chromium --browser firefox -n 3 --video-on-fail yes

support `--screenshots`

Adding support for screenshots is not feature complete until:

• Yes can provide yes
• Defaults to no
• User can provide full which means yes & provides full_screen=True to the context
• All cases above have adequate unit test(s)
• All cases above are xdist worker safe

Conditionals:

• Can we do anything fancy with the screenshot name to map it back to a test?

`--root-url` should be validated as a valid URL

USE `--pw-debug` before opening any issues

Please run with --pw-debug to view more context on failures before raising any issues, otherwise at times it is almost impossible to determine the causes.

Allowed deferring of `context` & `browser` kwargs

it is not always 'known' what the user might need, it may be required to calculate it dynamically, allow some sort of defer/callback that is invoked by the plugin machinery that can be provided as a keyword arg to:

@pytest.mark.context_kwargs(callback=lambda: {"foo": 1})
@pytest.mark.browser_kwargs(callback=lambda: {"bar": 2})

expose proxy capabilities

tracing `title` & `name`

when dumping the trace, a name & title are supported, derive them from the node.name possibly?

[High Prio] Implement a reusable binary solution for tests using `tox` and w/o

Right now we have some major problems in terms of tox/testing. Everything works fine from a normal perspective after a simple playwright install because browsers reside in ~/.cache/ms-playwright/* - however, in tox this is a pytes tmpdir, which would auto acquire binaries pretty much on every run.

Implement some sort of solution for development that allows tox to leverage / reuse existing binaries. This is also slightly tricker because pytester (probably) doesn't always pass the ENV through etc.

add `towncrier` to manage release docs

dependency is in place but add boilerplate to pyproject.toml to define types etc

support `--tracing`

similar to screenshots, deliver tracing specifics after screenshots.

tox sets a $homedir and fails with binary acquisition

reliably running tests that actually launch contexts/pages is... cumbersome right now, pytest itself will set a $HOME in its temp directory, manual IDE execution is easy, tox is slightly trickier - figure out a solution

Allow per test level context/page launch kwarg overriding

will require `xvfb` or some flavour for running `headed` tests in GH actions

if we want to run non headless;well need xvfb hooked up in gh actions

Build out mkdocs site

Docs are vital for UX, build out the mkdocs + serve them on github - the machinery is already in place for this, content is missing.

run tests on a single windows node

Good for catching oddities/os specific issues should they arise; do a single Gh actions on 1 windows version

force `--headed` off if using an IDE/debugger

headless by default is ok until debugging with an IDE, this case is almost certainly wanting to view the browser

bootstrap http server with some static content for testing

we will require an http server for integration tests in future, integration tests should stand this up and have the ability to navigate to the page(s) it returns for testing etc

implement `--artifacts` for the base directory

We need a place to persist storage/artifacts