Return the oldest transcript matching the video ID #1163
Conversation
Just return the oldest saved transcript in the DB that matches the given `video_id`, instead of also searching for the hard-coded `transcript_id` value of `"en"`.
seanh
force-pushed
the
query-for-oldest-transcript
branch
from
afbf2b0
to
f431a6d
Compare
| transcript = Sequence( | ||
| lambda n: [ | ||
| { | ||
| "text": f"This is the first line of transcript {n}", | ||
| "start": 0.0, | ||
| "duration": 7.52, | ||
| }, | ||
| { | ||
| "text": f"This is the second line of transcript {n}", | ||
| "start": 5.6, | ||
| "duration": 4.72, | ||
| }, | ||
| ] | ||
| ) |
There was a problem hiding this comment.
Fixed TranscriptFactory so that each Transcript object has a different value for Transcript.transcript. This is needed by the tests below, and is probably a good idea in general: TranscriptFactory always returning Transcript's with the same Transcript.transcript is a bit of a footgun (tests that compare the values of two Transcript.transcript's could have false positives thinking they came from the same Transcript when in fact they belonged to different ones).
| @@ -116,7 +117,6 @@ def test_get_transcript(self, db_session, svc, YouTubeTranscriptApi): | |||
| ] | |||
|
|
|||
| @pytest.mark.usefixtures("db_session") | |||
| @pytest.mark.parametrize("transcript__transcript_id", ["en"]) | |||
There was a problem hiding this comment.
This parametrize was fixing the value of transcript.transcript_id to "en" because get_transcript()'s DB query would only find transcripts with Transcript.transcript_id="en". It's no longer necessary for the test to do this.
There was a problem hiding this comment.
Looks good 👍
Tested in my local DB with an existing transcript from main and then:
insert into transcript (select now(), now(), 2, video_id, transcript_id || 'new', transcript from transcript limit 1);
update trasnscript set transcript='[]' where id = 2;
all good
update trasnscript set transcript='[]' where id = 1;
no transcript lines now
Just return the oldest saved transcript in the DB that matches the given
video_id, instead of also searching for the hard-codedtranscript_idvalue of"en".This is a slightly simplified rewrite of #1159.
YouTubeService.get_transcript()currently always saves transcripts in the DB withtranscript.transcript_id="en"(hard-coded) and it always queries the DB for a transcript withtranscript.video_id=video_idandtranscript.transcript_id="en".In #1162 we're going to start saving transcripts in the DB with different
transcript_id's (we're going to be generating unique transcript IDs that work even when a video has multiple transcripts with the same language code) so we can no longer always query for a hard-codedtranscript.transcript_id="en".Instead, this PR changes
get_transcript()to query fortranscript.video_id=video_idonly and then, since there could be more than one saved transcript with the samevideo_id, it just returns the oldest saved transcript.This should work, and it'll also mean that the default transcripts of previously-annotated videos and assignments don't change when we change the default transcript selection algorithm.
What we should really do here (probably) is persist the choice of default transcript by adding a nullable
video.default_transcriptcolumn to thevideotable (foreign key to thetranscripttable). But for the sake of expedience this will do for now. We can always back-fill avideo.default_transcriptcolumn in future by finding the oldest saved transcript for eachvideo_id.