umputun / feed-master / 41242272240

75%

push

github

Date finished YouTube live streams by download time (#184)

A YouTube live stream gets its `published` time when it is scheduled, not when it airs. `update()` moves an entry's date to the download time only when `published` is under 24h old, so a stream scheduled more than a day ahead kept the scheduling time and showed up in feeds back-dated, below newer episodes. Real case: Echo's "Один с Дмитрием Быковым" was scheduled 27 Oct 19:11 UTC, failed with "This live event will begin in 23 hours" until it ended, downloaded 28 Oct 21:21 UTC, and entered the feed dated 27 Oct.

The worker now records when it first reaches the download step for an entry (`first_seen` bucket in the youtube store, written before every download attempt, never overwritten, dropped by `SetProcessed`). `update()` resets the date when that first sighting was within 24h of `published`. A stream is first seen right after it is scheduled, so it gets the download time once it finally downloads. A new channel's backlog is first seen days after publishing and keeps its dates, as before.

`updated` from the YouTube feed is not used: it moves on any metadata edit, so recently edited old videos would jump to the top when a channel is added.

Behavior change: a regular video held back by a download outage longer than 24h also gets the recovery time as its date, the same way outages under 24h already did. Entries with no `first_seen` record, including everything pending at deploy, behave exactly as before.

48 of 66 new or added lines in 2 files covered. (72.73%)

1558 of 2068 relevant lines covered (75.34%)

20.94 hits per line