Freegle / Iznik / 24990

72%

Build Type

push

circleci

Committed by invalid-email-address

Commit Message

feat(upload): instrument the client-side compression step directly

The closed PR #963 claimed ~13% of selected photos were "abandoned during
compression", but nothing measured that: no Loki event bracketed the compress
step (preprocess-progress was only console.logged), and with autoProceed the
`upload` event (upload_started) fires in @uppy/core's createUpload BEFORE
runUpload executes the Compressor pre-processor — so the selected->started gap
is before compression by construction, and the 87->76 delta was a per-file
(file-added) vs per-batch (upload) counting artifact.

Instrument it properly instead of guessing:

- upload_compress_started / upload_compress_finished bracket each file's
  compression via @uppy/compressor's per-file preprocess-progress /
  preprocess-complete events, capturing original_size, compressed_size,
  elapsed_ms, and whether the bytes actually shrank.
- A `compressed` flag, set from the batch `compressor:complete` event (which
  carries only successfully-compressed files), records silent compression
  failures — @uppy/compressor swallows its own errors and uploads the file
  UNCOMPRESSED, so absence from that event is the real, measurable failure
  mode the PR only speculated about.
- Make the funnel per-file end-to-end (selected -> compress_started ->
  compress_finished -> succeeded) so stages are comparable, fixing the
  per-file/per-batch mismatch that caused the original misdiagnosis.
- Tag every event with `uploader` (our|photo) so desktop (OurUploader) and
  mobile/tablet (PhotoUploader) can be separated in Loki. PhotoUploader had
  no upload telemetry at all; it now emits the same funnel.

No change to compression config or upload behaviour — this establishes a
baseline against the current default Compressor so any future downscaling
change can be judged on measured data, not inference.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VqZianMqz6i4FKMdPgnvWM

Coverage Stats

12352 of 16346 branches covered (75.57%)

Branch coverage included in aggregate %.

33 of 89 new or added lines in 2 files covered. (37.08%)

514 existing lines in 9 files now uncovered.

131319 of 183852 relevant lines covered (71.43%)

38.1 hits per line