To give an example, I've seen lots of "antiviruses" inject code or perform unexpected DOM changes, causing certain things to break.
Other times a user shows up with a heavily tweaked browsers that behaves differently from what you'd expect.
I stick with Flow for now just because create-react-app doesn't support officially TypeScript, but it should be possible with the next major version that will include Babel 7.
In a reasonable language, you do something obviously wrong, and it throws an exception immediately, which gives you a stack trace that gives reasonable context for understanding what went wrong.
In JavaScript, the failure often happens silently: "foo" + 1 returns "foo1", {}.foo returns undefined, and program execution continues until it absolutely can't, finally failing far away from where the problem occurred. And when it does fail, it often fails in a callback deep within framework code with no clear, traceable execution path that leads back to the point where the error occurred. The errors have all the mysteriousness of C errors, without even the traceability.
The JavaScript community doesn't seem to be aware that there are better solutions, and seem happy with the way things are, even going so far as to use JavaScript on servers where far better tools exist. It seems that the only salvation that will come will be when mature tools exist on top of WebAssembly so that those of us who care about error reporting can use better toolchains.
But I see what you're saying. Except for the declarative style popularized by React and also implemented in Mithril, Preact and Inferno, and except for the rise in popularity of TypeScript, and except for a lot of attention paid lately to immutable data structures, and except for jslint being created in 2002 and lots of similar tools since then, and except for the wide variety of JavaScript-based unit test frameworks, it really seems like the JavaScript community has no clue that there are ways to prevent the errors you describe and no inclination to prevent them.
Eh, I don't think that logging runtime-in-the browser is the value added here. I can wrap my codebase in a try-catch that catches exceptions and calls home with the stack trace in < 100 lines of code. Avoiding that < 100 lines of code isn't the value added.
So how is it even possible that this tool works if JavaScript doesn't do this? I don't know why you're just making things up.
If you think JavaScript throws errors immediately when you do something obviously wrong, try evaluating the following expressions in your console:
{}.foo
"foo" + 1
undefined + 1
Off topic because this service is a logging service to log JS errors in production. It's not related at all with how JavaScript handles errors.
Kinda wrong because even in "a reasonable language", you still need to log errors, warnings and other stuffs. It's absolutely not specific to JavaScript. Actually, JavaScript didn't have any serious logging solutions like Sentry, Loggly or RootCause until really recently. It's great to see all these new tools.
For error tracking I've previously used Sentry [0]. I found it really easy to use and setup in a fairly large JS app. The best part is that it's open source, so you always have the option of hosting it yourself, which removes any fear of vendor lock-in. I'd say it's well worth paying for their SaaS offering.
EDIT: Looking through the docs, it seems they expect you to load their non-versioned third-party sourced script, and it's not even hosted on a CDN. That's a huge security risk, so it's an immediate blacklist for me.
We'll have a CDN up shortly, good point! You can also host the JS script yourself, or use the on-premises version (Docker image)
We use FullStory and it’s fantastic. Would be extremely difficult to convert me.
Any idea why?
We record all events for any web app so you can replay session and fire the exact same mouse+keyboard+ajax events to reproduce the issue. Very similar tools but we wanted to make something generic targeting every web app
The best error handling technique is to catch exceptions and show up a big popup on the screen with the error details.
Your users will send those details to you instead of just saying "the app crashed".
While we'd like that to be true it's sadly not in practice.
The most common example is websites with sensitive data in the URL (ex: PII query parameters) using third party analytics services that receive the entire URL as a metric.
Imho Ext is a specialty of all the existing UI frameworks, and imho it's shit. Tbh I though it would die directly after being bought by Idera. Even most of the most Senior people left or got fired after that.
Most annoying things (imho): - lock in: all their widgets, and tooling, and everything only works for Sencha products, and is by far worse than the current open source counterparts - price: for a SMB/big-company this doesn't matter, but you pay a premium to use "widgets" which are worse (imho) than open source - Sencha Widgets compile into clunky and non-semantic HTML (thousands of divs), which also impacts performance a lot: simple apps usually have 3-5MB - debugging is difficult as you have to go through their "OOP" layers (can be 5-10 layers) [Java Developers created this Web Framework]
People argue that it's great because many rich widgets are already available and because the framework tells you how do to stuff. But everyone will run into a point in time where something is just extremely hard because of extjs constraints, which usually ends in a hacky solution.
I am not sure how I feel about these session recorders and probably want to start blocking them from my browser. It is akin to someone recording your every movement inside your house and watching the video to see what issues come up. Oh and you have no idea its happening.
Also wonder about how this data is protected. If videos of how I used Facebook were leaked (messages I started to type but didnt send for instance) it would be a huge violation of my privacy.
In the same token, its like free UX research, you can see where people struggle by watching them use the app. It is like standing over someone's shoulder to see how they are using the product and helps immensely in that realm. I still feel like it should be opt in though.
1. We recommend all our users to use the opt-in prompt before an error is logged. Demo: https://app.therootcause.io/sandbox/confirmation/
2. We can also show a special icon in bottom right corner when video is recorded. Demo: https://app.therootcause.io/sandbox/videorecording/
And, it's not just about production sites - another use case is using this internally as a tool between QA <-> Developers to speed up communication when new errors are found.