microlinkhq / metascraper / 31421500642

95%

push

github

fix(youtube): date a video by its upload date (#869)

metascraper-youtube owned title, author, publisher and image but left
date to the generic rules, where `$jsonld('dateCreated')` searches the
JSON-LD tree recursively, walks into `comment[]` and answers with a
viewer's timestamp:

    https://www.youtube.com/watch?v=arj7oStGLkU
    2020-05-11T19:29:04.000Z   <- a comment
    2016-04-06T16:59:35.000Z   <- the upload

Declare `date` here like the other fields, read from the `uploadDate`
the watch page carries. Vendor rules run before the generic ones, so
YouTube no longer depends on how the shared date rules are ordered.

The fixture is the watch page's own JSON-LD, comment included, with the
commenter's details replaced. Removing the rule fails it with the exact
production value. The snapshot test alongside it destructured `date`
out and asserted only that it was a string, so it could not have caught
this; pin it.

The recursive search reaching into comments is a wider bug that also
affects `review` nodes. It belongs in its own change, against
@metascraper/helpers.


Claude-Session: https://claude.ai/code/session_01DiDJXMPg2e98kgL3drjg6f

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

807 of 858 branches covered (94.06%)

Branch coverage included in aggregate %.

4 of 4 new or added lines in 1 file covered. (100.0%)

2830 of 2983 relevant lines covered (94.87%)

42.23 hits per line