I am taking some notes here, basically for deeper learning and reminding me of knowledge. If it is helpful for others, cool.
It just started. I warmed up talking to Jacek, we both thought it starts at 15:00 but it didn't so it gave us 1h to warm up.
Wow, more then 5000 participants. I guess that is the advantage of online conferences, the number of people joining is basically unlimited.
They have all kinda stuff set up, discord (a chat), spatial (a kinda virtual env where you can work around) and of course you can watch the video in live.
Talk 1: Evolution of Browser Testing
from Saucelabs starts with a 25min talk on how browser
automation has evolved over the last year. So I expect a high level overview.
Let's see if there is something to learn.
Half in he very well explained the history and also the tech setups of all the browser testing tools.
I start to get the feeling that the conference is going to be just about browser testing (?), which I would call integration testing. I am not sure this is going to satisfy what I expected from "Test JS Summit". That makes me think about what I did expect. I beleive I expect to hear about good testing approaches, which start at a lower level for me. I actually see testing more as a way of driving code, which means drive design and architecture of a software.
Aha, he is introducing webdrive-bidi now.
WebDriver BiDi is a proposed bidirectional protocol for browser automation, building on and extending WebDriver.
WebDriver BiDi is not ready.
Christian is saying "you can finally talk to your Safari-friend". Seeing the latest focus that Apple puts on browser development I would not be so sure that they will help us on that front.
I am glad that there are things moving forward in the space of browser testing.
In the Q&A at the end the question came up "How can I prevent the
Christian says "it is something that you need there".
This is what I also see as the main issue, it causes lags, flaky tests, devs pulling out their hair
and tests to become hard to maintain. I have no solution either, I am also thinking about this
for years already ;).
Talk 2: It's not About Your Assertion Library
The talk was not about assertion libraries or anything alike, it was an even higher
level talk. Ok. I didn't expect that either.
Still it was very good to have this kind of talk, we don't talk about why and how to test right often enough. Still it felt a bit like he was not able to find the rights words to get through to me, but I know what he means. 🤔
Mark says in the after-talk Q&A.
Back in the days, before docker, when you installed node on your machine ...
In the Q&A "What is your definition of integration tests?". Define it yourself. Align with your team what are the right tests for your product and call them whatever you want.
What? There are companies that really label themselves to do "Enterprise Jamstack Websites"? Wow, what happens when the Jam is not sweet anymore? Pffff... 🤷🏽
Talk 3: Your Tests Lack Vision
Angie Jones starts with a video on inattentional blindness, which already makes very clear where she is going (since I know what applitools does). Then comes a lot of slides for nodding, especially if you have done visual testing a couple of times.
When you are writing your tests, you are scripting in your inattentional blindness.
She came across a broken website, and then:
I did what any of you would do, I opened the dev tools.
I love that. She still finished the process but also mentioned that she thought about the non-devs that don't even know about devtools. Yo yo.
She is getting her point across quite well, she bascially sells the AI browser testing approach the company she works for does. I am totally on board and I think this is a very good way to do it, hence my ML tidbit on "Machine Learning vs. Screenshot Comparing". I just can't stop having the impression that this is more of a sales talk than a talk for a conference.
First question is about Github Actions. Kent C. Dodds touches on the monopolistic trend just briefly, not enough I think. I have my problems with github actions being so dominating. I have a draft post that was triggered by me learning github actions, leading to go to codeberg (without having anything like github actions). The marketplace makes it "open source" was a sense I got from the discussion, I am not sure I can agree.
Also here it sounds like the browser testing is what people like a lot. Unit tests are always mentioned on the side, but I am not sure it gets the love it should. Maybe I just have an old style view on this.
Oren says "integration testing you get high coverage really fast", without touching on the tradeoffs. Then he mentions that he also hates flakiness and prefers 20% of his code tested and be reliable, over 100% tested but flaky. Later he does state the big difference of testing at the top of the test pyramid and at the bottom. Well done.
Jason Palmer says "keeping quality high pays dividence over time". nodding hard
"When should I write a unit test?" and "What is a component test?" Yoni asks the panel. Kent states that more integration tests prevent needing unit tests (how I heard it). Aha. Though he also says that algorithmic things (as he calls them) are where he sees unit tests being needed. Yet another time that I understand that we don't really have the same view on testing.
Nancy says it: "integration tests is a term we are not aligned on". I agree.
Mmmh, Yoni summarizes the panel kinda like this: test on the user's side, network and UI! And (he said) we still have the "old tools" like unit testing. And that gives me food for thought, having learned testing from JB Rainsberger's and Integrated Tests Are A Scam. Do I look at it from the wrong angle. But the test quality I have seen in the last years and I see in most places is so low, that I am not sure we are on the right track dismissing testing the small things and just testing from the user's chair.
I need a break ... bbl
Talk X: Contract Testing with pact.io
Matt Fellows shows us how integration tests actually will fail you when your system grows. Thanks. I am not loosing faith. Yep, contract tests, this is the thing we want between parts of our system.
Ok, I realize the integration tests things like *-testing-library is a different flight level to what pact.io and the integration tests Matt talked about.
Matt says it "End to End Testing means too many things". Where does contract testing fits in the test pyramid? It's rather close to a unit test Matt says. On the provider side it sits more in the middle of the pyramid. "They are closer to unit tests, but they still run through a couple layers". "Contract tests removes tests from the end-to-end test." Matt says in the Q&As.