empty Entry update() method about django-yarr HOT 6 CLOSED

Interesting. That shouldn't be empty, but if it was there, arguably it should be the only way an old item should resurface as unread - if it has changed, it's been republished so should probably be shown to users. Instead you're seeing the opposite, which suggests the problem is elsewhere.

Do you have an example of a feed and item(s) which reappeared so I can check exactly what's going wrong?

The most likely reason would be if the RSS publisher removed an item from their feed; that would cause an old item to re-appear at the end of their feed, and Yarr would add it as unread, thinking it was a new item. To work around this issue, I'll delay deletion (probably by adding a new "expires" Entry field) - but at this stage that's just a guess, so although useful, it may not actually be fixing the problem you saw.

(The other reason for items reappearing is if the guid, link, and title or date were missing in the feed, but added later; that sounds like a really weird edge case though, so seems unlikely - and if that is the case, I'm not sure what the right action would be anyway!)

For now I will add the expiry field, add an exception for missing guid/link/title+date, and fix a bug where unread items would get cleaned prematurely.

from django-yarr.

It has happened with nytimes feeds, with this site's feed http://blog.sanctum.geek.nz , and with my own feed. Others as well, but those I definitely remember. Can't say about the other two, but my own feed had not changed in any way when it suddenly reappeared as unread. http://onethirdpi.com/blog/feed.rss .
These are items that reappear in bulk. It seems like all of the items of the feed become reset to unread. It probably wouldn't show up in the All view, but I hang out in the Unread view, so these "new" items appear as a block of entries by one feed at the bottom of the list.

from django-yarr.

Ah, thanks for the extra info - sounds like something's going wrong with the feed parser on one request, causing the entries to be wiped, then it works the next check and everything goes back to normal. The expiry field would act as a workaround, but I'll try to find the underlying cause too.

from django-yarr.

Just want to add that the blog.sanctum.geek.nz feed has been freaking out this way fairly consistently for the past couple of days. Nothing has been added to it. I'll download a copy of it now and the next time it comes back unread, and will report back.

EDIT: Actually, the problem might be, as you said, in retrieving the feed. wget took a long time to get the request through. Looks like it had to deal with some temporary redirects to feedburner. So that retrieval may have just been timing out usually, and going through quickly enough on a rare occasion. Will compare later just in case.

from django-yarr.

Thanks for the extra info! I still couldn't recreate it, but a timeout could explain it - if so, I've just pushed 0.3.6 with some changes which should fix the problem:

• Sometimes when a feed fails, it can still be parsed; I think this may have been where it was going wrong, so it now checks that at least some entries were found: https://github.com/radiac/django-yarr/blob/master/yarr/models.py#L320
• New YARR_ITEM_EXPIRY, which will double as a workaround if that wasn't problem - by default entries hang around for a day after they would have previously been deleted.

As I said, I've not seen the problem myself and can't recreate it, but I'm running a Yarr installation for a couple of friends, and one of them noticed it happen - I'm going to run with ITEM_EXPIRY=0 and will continue to monitor the situation to check I've found the underlying problem.

I'm going to tentatively mark this as closed, but will reopen it if it still happens with 0.3.6+.

from django-yarr.

Just wanted to note that there is pretty much no difference between the feed that I saved this morning and the one that I just got now. There is just a different generation time and timestamp comment. The feed does not change besides that. Will install the updated version now to see if the expiry helps.

from django-yarr.