Roster ID problem with new version

I use the World Bank’s cloud data server for my survey, and my instance is “fp2017.mysurvey.solutions”.

Before the 18.06 version was implemented from the back end, in any Stata export file for a roster of my questionnaire, the combination

interview__id interview__key __id

uniquely identified each observation. This has now broken. Specifically, it looks like for some questionnaires, two observations now have the __id of 1, where before one of them was a 0 and the other was a 1. To be clear, the same questionnaire, exported before 18.06 was implemented (i.e. before 6 June 2018), had no duplicates, but does so now.

Could you please look into this ASAP? Thank you so much!

Dear Hemanshu Kumar,

Could You please send us the interview keys of both “problematic” interviews and specify the roster file where issue occurs?

Best regards,
Survey Solutions Support Team

Dear Tisorn, Hemanshu, Peter,

thank you very much for reporting the issue. We were able to reproduce the issue and are investigating now.

Sincerely, Sergiy Radyakin

Dear Sergiy,

I second Hemanshu’s observation. We run several surveys at the moment on WB cloud server and after exporting, the id’s (of rosters) do not uniquely identify items any longer.

I am currently looking into it if there is any systematic explanation and will also share with you an address of a server and an exemplary interview__key.

Best
Peter

Dear Tisorn, Hemanshu, Peter,

the Survey Solutions has just been updated with a fix for the problem you have reported. Please re-export in version 18.06.20 and confirm that the problem is no longer appearing with your data.

Thank you very much, Sergiy

This seems to be fixed now Sergiy, thank you so very much!

However, there is another place that my code is breaking now, and I wanted to ask you whether this is a bug or an intentional change:

The paradata.tab file that I am now downloading seems to be formatted slightly differently. In particular, in the earlier versions, when I would download the file using a command like

. import delimited using paradata.tab, delim("\t") clear

I would get the following 9 variables:

ïinterview__id
v2
action
responsible
role
timestamp
parameters
v8
v9

Now, however, I get the following 7 variables:

ïinterview__id
v2
action
responsible
role
timestamp
parameters

On closer inspection, it appears that what was earlier in “v8” is now separated from the preceding string by a “||” rather than by a tab, which is why it is not being recognized as a separate variable by the above Stata command.

Is this correct and intentional?

Dear Sergiy,

I can confirm that this problem does not occur any longer.

Thanks a lot!

Best
Peter