I am having a problem with group randomization combined with token based response persistence.

The issue seems to only happen if you have a survey with token response persistence, multiple groups assigned to the same group randomization, partially complete that survey leaving some of those groups unanswered, then re-enter the survey and get assigned a new randomization order. The page renders with your question(s) hidden with the classes "ls-irrelevant ls-hidden input-error".

I am not exactly sure what is happening but you can replicate it with the logic above. You might need to enter in a new private window a few times to get a scenario where it happens.

If using the latest version, please file a bug report including a sample .lss (survey export) file.

I'm on Version 3.8.1+180529. I just made a fresh install with the latest and it appears group randomization doesn't work at all in 3.14.3+180809 (on groups at least). Ill put something in the bug report when I get a chance. For now we will just remove response persistence since the survey is short.

Bigred01 wrote: I just made a fresh install with the latest and it appears group randomization doesn't work at all in 3.14.3+180809 (on groups at least)

You mean even without using token the whole group randomization feature is currently not working at all?

please, could you upload your lss file so we can reproduce?

Attached is the lss export and some screen shots.

When group randomization is on, the question number beside the asterisk in the question text does change to a random order, but the groups and their questions are still asked in the same order as they are laid out in the survey. When the group randomization is removed, the question numbers match the questions that show.

I have only tried with on the groups themselves. I have not tried question level randomization inside a group in 3.14.

Fresh install of Version 3.14.3+180809
Using Vanilla theme with ajax mode disabled.

I can confirm that the group randomization is not working with this file.
Used version 3.14.2 via demo.limesurvey.org.

Finally getting back to this.


Every time this issue happens $em->currentQset is populated with the original question one would see without group randomization, but the page is rendering some other groups question as it should with randomization. There is no $lemQuestionInfo set so the rendered question is defaulting to a "generic_question", invalid and hidden and you are unable to proceed next.

Setting newtest=Y seems to correct the em instance so I will use it as a workaround until I fix the issue in my install

Has anyone else happened to figure this one out or have any insight on what the problem might be?

Thanks

We have just created a lengthy explanation of the different Limesurvey randomization features and that included group randomization. It worked fine for us using the latest Limesurvey version without tokens.
Here is a link to the setup and a link to a sample survey .

You can test the whole sample survey with all randomization features at this link .

Did anyone file a bug report so we can follow the progress and double check if this is now fixed?

Ill make a note to test this.

Group randomization (at the group level) has always worked with no tokens for me. Its just when you mix in response persistence, a new session and a partially answered survey where they stopped in the middle of the randomized groups. It re-randomized the groups putting an unanswered mandatory question before the last question answered, then tried to send you back to the last question answered which resulted in the a group loading with error classes and hidden questions, unable to proceed forwards.

I think I still have my demo survey. I can re-load this to the lime demo site and try to replicate again.

I just encountered this issue with question group randomization and token based persistence. I had quite a few respondents report issues before I found this thread.

I'm using LS 3.x build 190529 and this is still an issue.

Any thoughts on how this can be fixed?

And if not possible to fix, how about a check during survey activation that displays an alert if question groups are randomized, and tokens are being used?

blocka wrote: Any thoughts on how this can be fixed?

Where is the bugreport for that issue? I thought the seed value was introduced to allow routines to make randomization reproducible. The code seems to be not using that value.