Touchstart event logging non-existent clicks

[alexhass]

I have reviewed your code because of https://www.drupal.org/node/2671716 and like to ask how you would solve Touchstart event logging non-existent clicks.

I have a document listener bound to "mousedown keyup touchstart" as you can see in http://cgit.drupalcode.org/google_analytics/tree/js/google_analytics.js to overcome the other serious bugs autotrack currently has like:

• Target _blank will cause serious failures #62
• Tracking only bound to A tags / Massive page rendering slowdown with many links #63

Monitoring the mouse movement looks also a problem as I found on the net. Any other ideas?

[philipwalton]

There's always going to be a trade off here because there's no 100% reliable way to know for sure when a user action initiates a browser navigation.

If you bind to only the click event (as autotrack does) you'll miss some links, but if you bind to touchstart, mousedown, and keyup, you get a lot of false positives. It's a trade off.

That being said, I'm open to giving users the option to set that dial wherever they want. For example, we could add a configuration option to listen for additional event types and then indicate those event types in the event hits so users could distinguish between them when reporting on outbound link tracking.

Would that appeal to the Drupal team? It would mean your reports might looks something like this:

Event Category	Event Action	Event Label	Total Events
Outbound Link	click	http://example.com	112
Outbound Link	mousedown	http://example.com	61
Outbound Link	touchstart	http://example.com	38
Outbound Link	contextmenu	http://example.com	14

---

Also, FWIW, in a comment on that page someone says:

Google has documented in past that Google use mousedown keyup touchstart events to trigger tracking on cross domain auto link tracking. This was just overtaken.

This recommendation (which was removed from the documentation because it was too often misinterpreted) was for adding a linker param in cross domain tracking, which is a different thing entirely. In the case of adding a linker param, false positives have no negative impact. But in the case of outbound link tracking, false positives lead to inaccurate results, so the same methodologies shouldn't be naively applied to both use cases.

[philipwalton]

The suggestion I proposed in #65 (comment) has been implemented in the 1.0 release.

[alexhass]

I tend to say you know it better. I used click only in past and many complained that no mobile users have been tracked. It looked like click does not fire on mobile. How can we implement this in a safe way so it works all times on all devices without false detections?