simpleanalytics / scripts Goto Github PK

Hey guys

I'm trying to setup analytics on a website but it seems like the url of simple analytics is hardcoded in the library. I can do a PR myself but I do know where to find the source code for the npm package.

Thanks in advance
Tim

It appears that the script has a high number of unique visits for some websites:

It's comparing a www domain with a non-www domain.

My Sentry is showing errors thrown in the SA script, looks like it's this line:

Here's a link to the Sentry issue:

https://sentry.io/share/issue/20005bc054fb4e48a65b8c01209ae3d2/

We send a session id with events, but it makes sense to send it with page views as well.

That way we can link events (that happen on one page or multiple in a SPA) to a page view.

Add this for iOS:

https://github.com/duckduckgo/iOS/blob/28f2c58201a6b74abba12fc3ace057d76322b135/Core/UserAgentManager.swift#L99-L103

And this for desktop:

window.navigator._duckduckgoloader_

Sometimes you don't want to skip the Simple Analytics on local scripts. For example with Chrome extensions.

I am currently trying to collect some usage information about a chrome extension using SimpleAnalytics - is that possible?
I want to track some events in the extension popup-window (that appears when clicking the extension icon in the top right). This extension popup is hosted by Chrome on "chrome-extension://[Random extension ID]/popup.html" so I used the "data-hostname" argument to set the hostname to another custom URL that I created for this extension.

Using this technique, Simple Analytics is still logging "Simple Analytics: Not sending request from [Random extension ID here]". It looks like the random extension ID is triggering the localhost detection as "locationHostname.indexOf(".") == -1" will eval to true. Is there a way to disable this check or any other recommendations when using for extensions?

Let's make this optional via data-force=true, data-force-run=true, or data-always-run=true. Force makes the most sense to me because we then also can bypass the check if the script is loaded twice. Maybe some people want to load our script multiple times.

For some customers it only is added to 30% of the page views:

Instead of using this:

We can use:

var scriptElement = doc.currentScript || doc.querySelector('script[src*="' + baseUrl + '"]');

• We can now support multiple scripts on one page with multiple settings

Tasks

• Update script to use above syntax
• Add to the docs: not supported in IE (https://caniuse.com/document-currentscript) but it uses regressions
• Update proxy docs page

We use onClick when somebody clicks a link. We only set this when there is no onClick handler set. But for some frameworks like Next.js all links have onClick events.

To not overwrite that behavior we can attach an additional event that does not wait for our request to be done:

if (!link.getAttribute("onclick")) {
  // current code
} else {
  // new code
  link.on('click', function() {
    // current code
  })
}

Why not always use the second method? Well, we want to wait for the event to be sent to our server before sending the user to the next page. So it's a backup method.

cc CH

Hi there,

I Was looking at your code and saw that:

scrolled = window.setTimeout(position, 500);

setTimeout returns a TimeoutID, that we can pass to clearTimeoutto cancel the future call. So it won't do what you think.

You may want to do:

window.setTimeout(function () {
  scrolled = position()
}, 500);

Anyway, great product!

In the following line, it would be nicer to use window.addEventListener('hashchange', hashHandler, false); instead of window.onhashchange = hashHandler, because it plays nicer with other code:

It just took me way longer than I care to admit to debug an issue caused by this, which only occurred in production, since my dev-builds don't use the analytics script.

Hello,

The constructor new Event() is not supported in Internet Explorer, and this is causing scripts to break when they use history.pushState(). It looks like you are checking that Event is defined before trying to use the constructor, but Event is defined in IE, just not as a function.

There are a couple of possible solutions that come to mind: polyfilling or substituting the constructor somehow, or checking not just that Event is defined, but that it is defined as a function.

Thanks,

Jonathan Sutcliffe

(navigator.brave && await navigator.brave.isBrave() || false)

Original: simpleanalytics/roadmap#609

From a very helpful customer, we got this error report:

The following is logged to console:
Simple Analytics: Uncaught TypeError: Illegal invocation
At this stage, it seems exclusively happening on Chrome/Windows.

Good read: https://xgwang.me/posts/you-may-not-know-beacon/

It does suggest that if a URL has an error, sendBeacon will throw "TypeError: Illegal invocation".

This also seemed to effect the next.js project as well: vercel/next.js#23856 (fix: https://github.com/vercel/next.js/pull/26601/files)

Maybe with domains separated? data-hostname="example.com,example2.com" or data-extra-domains="example.com" so customers don't have to type their domain.

Detect this as a bot.

cc @Kilian

There are bots that overwrite random, let's run a test to see if they do:

const bot = Math.random() === Math.random();

When retrieving the following Simple Analytics file via my custom CNAME:
https://sa.example.com/v8/app.js

This file has an embedded "sourceMappingURL" set to "v8.js.map", which does not exist. It seems like this should point to "app.js.map", which does exist. Currently this results in a console warning.