It looks as though it depends on the exact nature of the query.

The following always return up to date results, even with a since_id
(I haven't included those since_ids here)

http://search.twitter.com/search.json?q=hong+kong+OR+kowloon&rpp=100
http://search.twitter.com/search.json?q=%23iphone&rpp=100

but the following will just return 200 OK with no results:

http://search.twitter.com/search.json?q=from%3ADavidFeng+OR+from%3ABeijingWithKids+OR+from%3ABlueJDMBA+OR+from%3Akaiserkuo+OR+from%3Acharlieflint+OR+from%3Aourmaninsh&rpp=100
http://search.twitter.com/search.json?q=%23beijing+OR+%E5%8C%97%E4%BA%AC+OR+beijing+OR+%E5%8D%97%E7%BD%97%E6%95%85%E4%B9%A1+OR+nanluoguxiang+OR+%E4%B8%89%E9%87%8C%E5%B1%AF+OR+%E4%BA%94%E9%81%93%E5%8F%A3+OR+%E6%9C%9B%E4%BA%AC+OR+wudaokou+OR+sanlitun&rpp=100

Interesting: change the above first query to:

http://search.twitter.com/search.json?q=hong+kong+OR+kowloon+OR+tsim+tsa+shui&rpp=100

and then the results STOP coming if there is a since_id ....


I've filed a support ticket with Twitter ( 623447 ) with this info,
and hopefully we'll see some progress on it.

Mark.


On Oct 28, 9:56 am, Marc W <marcwanchipm...@gmail.com> wrote:
> A number of people are seeing similar things, especially if you
> specify a since_id:
>
> http://groups.google.com/group/twitter-development-talk/browse_thread...
>
> My current (extremely bad) solution is to just "hire hose" and get all
> the tweets every time, and then filter out those I've seen before by
> id.  Gaaak!
>
> I'll see if opening a support ticket or whatever helps with this
> instead.
>
> Mark.
>
> On Oct 27, 6:28 am, briantroy <brian.cosin...@gmail.com> wrote:
>
>
>
> > This is happening RIGHT NOW for the following:
>
> > 1) Go to search.twitter.com and enter "tweetsforboobs OR
> > tweetforboobs" as the search.
>
> > 2) Go tohttp://tweetsforboobs.organdsee the twitter feed on the
> > left.
>
> > Notice that the last tweet from 2 hours ago (VerticalMeasures) is not
> > in the twitter feed on tweetsforboobs.org. Also note the ID of the
> > tweet - from VerticalMeasures that is missing from tweetsforboobs.org:
> > 5181937429
>
> > Now here is the log file of the Twitter API call:
>
> > DEBUG: 06:18:01 PM on Mon October 26th Doing CURL fetch with User
> > Agent: justsignal/1.0 (+http://justsignal.com) and 
> > RFERER:http://justsignal.com/widgets/20ab5e90bf116397d6fb84ca80321928/widget...
> > DEBUG: 06:18:01 PM on Mon October 26th Twitter responded with 200 HTTP
> > Status Code.
> > DEBUG: 06:18:01 PM on Mon October 26th MaxID: 5182676703
> > DEBUG: 06:18:01 PM on Mon October 26th There are: 0 results in this
> > fetch.
> > Updating number for api hits for hour: 18 to: 6
> > THROTTLE-69: 06:18:01 PM on Mon October 26th Slowing collection...
> > Avg: 0 returning delay: 180
> > DEBUG: 06:18:01 PM on Mon October 26th Checking for next page... ****
> > DEBUG: 06:18:01 PM on Mon October 26th There is NOT another page of
> > results...
> > DEBUG: 06:18:01 PM on Mon October 26th Old max: 5182676703 New max:
> > 5182676703
> > DEBUG: 06:18:01 PM on Mon October 26th Old max: 5182676703 New max:
> > 5182676703
>
> > Note that our id is already > the last tweet ID from VerticalMeasures,
> > yet we never got that tweet. Our id from the log snip: (5182676703) is
> > NOT in our database (we never got it). It does not match the tweet ID
> > before Vertical Measures: 5180513610
>
> > Somehow the API is returning a new (and bigger) max id on 200
> > responses with no tweets in them OR on 403 (those are the only two
> > http codes in the log for today). Either way, that shouldn't be
> > happening.
>
> > Brian Roy
> > justSignal
>
> > On Oct 26, 12:47 pm, briantroy <brian.cosin...@gmail.com> wrote:
>
> > > Everything below ONLY PERTAINS TO THE SEARCH API:
>
> > > 1) Since late last week I've noticed a significant number of 403
> > > errors (403 Error from JSON: since_id too recent, poll less
> > > frequently). These usually indicate I'm hitting a server with an
> > > "older" view of the search index - since it thinks the ID I sent in
> > > since_id is newer than the newest it has. These trouble me because
> > > when I get a 200 after the 403 sometimes I get everything back to my
> > > since_id, sometimes I don't. I appears some indexes have gaps until
> > > they catch up.
>
> > > QUESTION: Are there any ongoing search indexing issues that you are
> > > aware of?
>
> > > 2) Since late last week I've noticed that some search API requests
> > > appear to get "stuck" returning an empty json result (no new tweets).
> > > This can go on for HOURS (today one got stuck like this for 12 hours).
> > > When I restart my process sometimes this clears up (I get the backlog)
> > > - other times it does not (I continue to get 0 tweets in the json).
> > > All of the requests return HTTP 200 and valid json.
>
> > > QUESTION: Are they any ongoing caching issues with the search API?
>
> > > These issues are new in the last 7 days (since about last Thursday).
> > > My IP is whitelisted. I'm sending both a valid user agent and referrer
> > > header. My processes are throttled by the volume of tweets the
> > > receive. I've made no changes to my processing since late September.
>
> > > Any assistance would be appreciated. My user's are comparing what they
> > > see from my service to search.twitter.com and telling me we are
> > > broken.
>
> > > Regards,
>
> > > Brian Roy
> > > justSignal

Reply via email to