I'm getting the same issue as Marc and harshavs too. I'm using since_id to start a search from the last set of resuts but generally get 0 results most of the time, even though the same search without since_id yields lots of results.
I could start logging tweet IDs against user IDs, and then work backwards down the list of past results until I get a since_id that yields results, but it does seem like a bit of a pain. This is my first time using the search API so I can't even tell you when this started happening. It just happens! For info: search is using negation operators, OR and from: Dave On Oct 23, 10:29 am, harshavs <harsh...@gmail.com> wrote: > We are experiencing the same thing as Marc. We are not using from: > though. > But intermittently it gets tweets in significant number. If I try to > get the same set again using the since_id for this particular crawl, I > again get no tweets. > > Any Idea on this issue? > > Thanks. > > On Oct 23, 10:48 am, Marc W <marcwanchipm...@gmail.com> wrote: > > > We've started seeing this too. If we specify a since_id, then > > periodic refetches will frequently return "no new tweets" (but > > sometimes they will return new ones). If we simply drop the > > since_id, then all new tweets are fetched. > > > Some new server optimization thing gone wrong? > > > Thanks! > > > On Oct 20, 10:06 pm, Christopher Warren <christopher.war...@gmail.com> > > wrote: > > > > We have an app that runssearchesregularly, and recently stopped > > > receiving newtweets. After investigating we found a search > > > combination that seems to break the search API. Instead of getting a > > > response withnotweets, an .atom request errors and a .json request > > > 404s. > > > >http://search.twitter.com/search.json?q=from:silent_tester02&since_id...... > > > > Changing the query to not use from:username works as expect, but I've > > > put several usernames in and they all respond the same way. I haven't > > > managed to narrow down the cause of the problem much further than > > > that, but we're handling it in our code by rescuing any failedsearchesand > > > appending since: with the date of the most recent tweet > > > to the q. > > > > Any thoughts on what might be causing this would be appreciated.