[atlas] Feature Request - Atlas Latest

Hi Daniel

On 14/03/14 17:47, "Daniel Quinn" <dquinn at ripe.net> wrote:

>On 14/03/14 15:34, John Bond wrote:
>> I would like to request an additional API call which just returns the
>> latest results for a specific measurement.  For use cases I would point
>> you to the "RIPE Atlas - Map Visualisations²[1].  I have also seen
>>places
>> in the dnsmon beta where you only show the latest results.   this would
>> also be very useful for trouble shooting, if I see an issue its
>> inconvenient to download xMB of json data just to see the whats going on
>> now.
>
>I wrote this call yesterday actually, though it'll still be a few weeks
>before it sees the light of day.  We recently updated our datastore
>table structure to allow simpler querying of the latest results, so this
>call is now relatively low-cost to our infrastructure.  I'm currently in
>the process of porting it to work against all of our measurement maps,
>including the new measurement pages, so they will probably be released
>at the same time.
Great news Thanks 

>
>> An additional feature would be to ask for a result for X seconds in the
>> past, the call would return a result closest to that time.  This is for
>> investigating historic events, if I have an issue that lasts one hour I
>> would normally only need one result to perform the initial
>>investigation.
>> This would also be useful to have a history/timeline function to the
>> maps[1]. 
>
>Unfortunately, because of how we store the data, we can't simply grab a
>perfect set of results nearest time X.  At best, I can give you that
>which you've already accomplished with your snippet of code here: Take
>all results within a designated window, loop over them, throwing out
>duplicates by probe id, returning the remainder.  It would require the
>server to do much of the work you've managed to do client-side, in only
>a few lines, so there isn't a lot of motivation on our side to do it...
>yet.  We do have intentions to make the maps you mentioned historically
>relevant, at which point such a call may make more sense.  I can't make
>any promises here though.
Understood I can live with doing this client side.

Thanks John