Opened 17 years ago
Closed 17 years ago
#4720 closed (fixed)
feed class get_object has no opportunity to object to not having a url
Reported by: | Owned by: | cablehead | |
---|---|---|---|
Component: | contrib.syndication | Version: | dev |
Severity: | Keywords: | ||
Cc: | cablehead | Triage Stage: | Ready for checkin |
Has patch: | yes | Needs documentation: | no |
Needs tests: | no | Patch needs improvement: | no |
Easy pickings: | no | UI/UX: | no |
Description
I have a feed that looks almost exactly like the "complex example" given in the manual for chicagocrime.org:
class DocumentComments(Feed): feed_type = Atom1Feed def get_object(self, bits): if len(bits) != 1: raise InternetDraft.DoesNotExist return InternetDraft.objects.get(filename=bits[0])
I assumed that since that method existed, it would be called for all requests, and a request for the "base" url of the feed would result in the DoesNotExist exception. Instead, I get "AttributeError: 'NoneType' object has no attribute 'filename'", since in link() I use obj.filename
and obj is getting set to None.
I don't like having a known url for a 500 error, since I expect users to try to hack the urls and see what they can see.
I've attached a proposed diff. It always calls get_object, and creates a default get_object method that returns None. This should keep the existing behavior (and, in fact, works fine with my other feeds that don't take objects), and allows the user's class to handle a bits=[] if they would like to.
With the existing setup, the only way I can think of to not provide a 500 error when visiting this feed without an additional URL portion is to test for obj == None in the link method, and I can only do that because I know that the link one is the first one called - way too much internal knowledge of the feed class.
Attachments (3)
Change History (10)
by , 17 years ago
Attachment: | syndication-500.diff added |
---|
comment:1 by , 17 years ago
Cc: | added |
---|
comment:2 by , 17 years ago
Triage Stage: | Unreviewed → Accepted |
---|
Yeah, this is an annoyance in the feeds system and should be fixed.
comment:3 by , 17 years ago
The attached patch has been working well at http://rst2a.com for the last 2 or so months.
comment:4 by , 17 years ago
Needs documentation: | set |
---|---|
Needs tests: | set |
Owner: | changed from | to
Grabbing to add tests, and to update doco for this patch.
by , 17 years ago
Attachment: | syndication-500-2.diff added |
---|
comment:5 by , 17 years ago
Needs documentation: | unset |
---|---|
Needs tests: | unset |
Triage Stage: | Accepted → Ready for checkin |
- fixed up usage of tabs in the original patch
- added stripping out empty strings from bits to the patch - it seems like
if len(bits) != 1: raise InternetDraft.DoesNotExist
should handle the case of an empty url, instead of having:
return InternetDraft.objects.get(filename=bits[0])
hit the db with an empty string.
- added a failing (now passing) unit test
- updated docs
comment:6 by , 17 years ago
I feel like I'm going against the grain by strictly testing at the unit level.
syndication-500-2-alt.diff is an alternate patch, which tests things by invoking the full request stack. It's less unit-y, but might be more inline with the django way of doing things.
I was banging my head reading the docs, trying to work out how this could have worked as is, and finally checked chicagocrime.org. They are throwing a 500 error there: http://www.chicagocrime.org/rss/beats/