ukeeper / ukeeper-readability / 28722551124

78%

push

github

Stream image sizing, share client, honor context, cap concurrency

getImageSize buffered every image fully into memory via io.ReadAll, built
a fresh http.Client per image, ignored the request context, and spawned
one unbounded goroutine per <img> on the page.

It now streams to io.Discard behind a 10MB io.LimitReader (reading the body
rather than trusting Content-Length, so a lying header can't inflate an
image's rank), treats non-2xx responses as size 0 so an error page can't be
ranked as the lead image, shares a single lazily-built client, and
propagates the caller's context. The probe request no longer forces
Connection: close, so the shared client can reuse connections. extractPics
caps concurrency at 8 probes per page and bounds the whole probe phase with
an overall budget (per-image timeout kept below it) so a page full of slow
image URLs can't hold the handler past the write timeout; semaphore
acquisition is context-aware. Lead-image selection breaks size ties by URL
so it's deterministic rather than scheduling-dependent.

39 of 42 new or added lines in 2 files covered. (92.86%)

792 of 1012 relevant lines covered (78.26%)

9.87 hits per line