amazon tracking link redirect loop

[rrotter]

fyi, I was getting a redirect loop yesterday w/ the amazon tracker stripping rule. Amazon kept adding psc=1 in addition to th=1. I added th=1&psc=1 to the redirect, only to get another redirect loop of amazon trying to remove psc=1 today! So I added the following exclude pattern which should deal with both situations:

https://(www\.)?amazon\.(com|(com?\.)?[a-z][a-z])/dp/[^/?#]+\?th=1(&psc=1)?(#.*)?

I don't know whether my experience yesterday was a fluke or if psc=1 is a new parameter that we'll see more (or maybe they'll start adding other parameters), so not sure if this is something you'd want to add right away vs. waiting for more user reports.

Also note, I greatly simplified the tld pattern vs. the original rule because domain false positives don't really matter for an exclusion rule.

[mshibanami]

@rrotter Thanks a lot for letting me know that!

It seems that psc=1 indicates whether the user has selected a size option on the web page for T-shirts, for example. I couldn’t replicate Amazon automatically removing this parameter, but we’d probably better not touch it if it’s present.

Handling Amazon URLs is getting harder now, since we have to manage multiple parameters that can appear in different orders. So, I now think it’s time to rely mainly on capture-group processing, allowing us to deal with each parameter sequentially.

Like this:

• Redirect From: (https://(?:www\.)?amazon\.(?:com|(?:com?\.)?[a-z]{2})/)(?:.+/)?(dp/[^/?#]+)(.*)
• Redirect To: $1$2?th=1$3
• Capture Group Processing:
  1. $3: Replace Occurrences
     • Target: /ref=[^#?$]*
     • Replacement: `` (empty)
  2. $3: Replace Occurrences
     • Target: (?:(\?)|&)(?!psc=)[^&#]+
     • Replacement: $2
  3. $3: Replace Occurrences
     • Target: ^\?
     • Replacement: & (empty)
  4. $3: Replace Occurrences
     • Target: &$
     • Replacement: `` (empty)

This is the screenshot of the rule:
https://github.com/user-attachments/assets/e145de51-4795-478c-99d6-cb0f14ddecdf

It looks a bit awkward but should work fine without any performance problem.

I'll dive into it a bit more later on, but I'm sure I'll submit it soon.

By the way, thanks again for your report and the improved rule! If you're happy, could I add you as a contributor to the rule in the library? Thanks!