Open Stack

Hello,

I might have found the problem.

The problem is that there can be several 'Translated' results within a week,
however, those results are saved with the same primary key.

I have just proposed https://review.openstack.org/316460 .

I am not sure whether 'date' field is used on other Stackalytics sources,
but at least, generating the primary key for translation records on 
record_processor.py
does not use 'savedDate' field, as Shengjing mentioned.
If 'date' field is used on other Stackalytics sources, please tell me, 
then I will change the review :)


With many thanks,

/Ian


SJ Zhu wrote on 5/16/2016 12:14 AM:
> On Sun, May 15, 2016 at 10:42 PM, Ian Y. Choi <ianyrchoi at gmail.com> wrote:
>> I have a question, although it is not mainly related to wrong Stackalytics
>> stats.
>>
>> I have found that API 'stats/user/%s/%s..%s'
>> and API 'stats/project/%s/version/%s/contributor/%s/%s..%s' generate
>> different results.
>>
>> For example, from Apr 1 2016 to today, my count is 34320 with the first API,
>> but the send API generates 29490 for the same period.
>>
>> I do know why the two APIs generate different numbers.
>> Can statistics numbers on user API and the sum of results from each project
>> API be different?
> Hi all,
>
> Previously Stackalytics uses API
> 'stats/project/%s/version/%s/contributor/%s/%s..%s'.
> But I change it to 'stats/user/%s/%s..%s' in
> https://review.openstack.org/#/c/288871/
>
> I may find the BUG but I have no time to fix it these days. So I hope
> someone to help.
>
> Stackalytics will get the results every week. And
> 'stats/user/%s/%s..%s' will return
> everyday's count. However Stackalytics seems not to sum a week's work.
>
> To fix it, I think we can change the way how Stackalytics uses
> Zanata's API. Based on
> 'stats/user/%s/%s..%s' we can get 'savedDate'. So just use the
> 'savedDate' to display
> on Stackalytics. I haven't read the code carefully but I think the
> change is possible.
> What's more we can debug the wrong results on Stackalytics more conveniently.
>
> But there's another question. How can we fix the old results on Stackalytics?
>