fix(xhr): make performance observer work with relative urls

[mhennoch]

Which problem is this PR solving?

• Similar to fetch (fix(instrumentation-fetch): perf observer bugs #1868) make perf observer work with relative urls.

Some minor fixes

• I see no reason to use url.match, changed it to .test which is faster and get rids of the !! conversion
• Remove unnecessary slice() as sortResources already slices everything.

[codecov]

Codecov Report

Merging #2226 (a1b5412) into main (bd2a005) will increase coverage by 0.03%.
The diff coverage is 100.00%.

❗ Current head a1b5412 differs from pull request most recent head 5d5262c. Consider uploading reports for the commit 5d5262c to get more accurate results

@@            Coverage Diff             @@
##             main    #2226      +/-   ##
==========================================
+ Coverage   92.70%   92.74%   +0.03%     
==========================================
  Files         145      145              
  Lines        5184     5182       -2     
  Branches     1060     1059       -1     
==========================================
  Hits         4806     4806              
+ Misses        378      376       -2     

Impacted Files	Coverage Δ
packages/opentelemetry-core/src/utils/url.ts	100.00% <100.00%> (ø)
...s/opentelemetry-instrumentation-fetch/src/fetch.ts	96.91% <100.00%> (-0.08%)	⬇️
...emetry-instrumentation-xml-http-request/src/xhr.ts	97.58% <100.00%> (+0.99%)	⬆️
packages/opentelemetry-web/src/utils.ts	94.80% <100.00%> (ø)

[obecny]

lgtm, one thing

[obecny]

lgtm, thx for changes

[johnbley]

Good catch, thanks!

[MSNev]

Note: This won't work with all IE versions (depending on how it's hosted) as it (IE) can rewrite the URL as http://xxxx:443/yyy (which is just annoying).

For Application Insights we added a helper function that we run everything through to avoid the issue to ensure that all URL's (page and dependency (XHR)) calls get reported as expected. https://github.com/microsoft/ApplicationInsights-JS/blob/master/shared/AppInsightsCommon/src/UrlHelperFuncs.ts#L86-L93

[MSNev]

And the URL passed into the above function is generated by the same href trick being used here, we just "cache" a couple to avoid recreating new elements everytime.

[vmarchaud]

I'm not sure which versions of IE we support right now ? cc @obecny @dyladan

[mhennoch]

Sorry was on vacation. Anyways IE doesn't have PerformanceObserver support so this should not be an issue here as this code is never ran in IE. But this .href trick is used elsewhere also so maybe lets create a separate issue to figure it out?

[dyladan]

@vmarchaud officially i'm not sure we claim any specific browser support. Unofficially, I think we only really work in IE 11

@MSNev is this a blocker for this PR?

[MSNev]

I don't believe it's a blocker, and it can be fixed later (if people complain).

And it should be fixed more generally so that any checked / reported URLs don't get logged as http:///domain.xxx:443/ which is the main issue and only becomes a code bug when comparing different (unnormalized) Urls.