Duplicate key error when tagging media "tag" and "#tag"

[ayleph]

I have a user who was getting server errors when uploading media. He was tagging his upload as "tag,#tag". "tag", "#tag", "Tag", and possibly other permutations are given the same tag number in my postgresql database.

mediagoblin-prod=# select distinct tag from core__media_tags where name='tag';
 tag
------
 1742
(1 row)

mediagoblin-prod=# select distinct tag from core__media_tags where name='#tag';
 tag
------
 1742
(1 row)

mediagoblin-prod=# select distinct tag from core__media_tags where name='Tag';
 tag
------
 1742
(1 row)

This causes a foreign key constraint violation when attempting to add "#tag" to core__media_tags.

Error - <class 'sqlalchemy.exc.IntegrityError'>: (IntegrityError) duplicate key value violates unique constraint "core__media_tags_tag_media_entry_key"
DETAIL:  Key (tag, media_entry)=(337, 3534) already exists.
 'INSERT INTO core__media_tags (media_entry, tag, name) VALUES (%(media_entry)s, %(tag)s, %(name)s) RETURNING core__media_tags.id' {'tag': 337, 'name': u'#tag', 'media_entry': 3534}

I can reproduce this error by uploading by own media tagged as "tag,#tag". Interestingly, if I tag the media as "tag,tag" then I do not get the foreign key constraint violation. If we can ignore duplicate identical tags in the upload (tag, tag), then I think it would be good if we could also ignore tags which result in duplicate tag numbers in the table.

breton suggested I tag with under the 0.8.0 milestone.

[Boris Bobrov]

thanks for the report.

Note to self: this seems to be because of db/models.py +734 -- tags are matched based on slug. Check from there.

[Christopher Allan Webber]

I don't think this is 0.8.0 critical but if it happens by then I'd be happy to merge it.

[ayleph]

Thanks for the patch loic! I'll test locally and update with results.

[ayleph]

Patch applies successfully and appears to fix the issue. Without patch, setting tags = "tag, #tag, Tag" results in server error. With patch, settings tags = "tag, #tag, Tag" results in media being added with "Tag" as tag. Fine by me, as "tag", "#tag", and "Tag" all end up in the same collection.

Good to push to master in my opinion.

[ayleph]

Applied as of ​679f729. Thanks for the patch.