First of all, apologies for being able to provide limited version info here. The sticky article said to find version info in the footer of the admin page but for some reason I see no such version info in the footer.

If our documented info is still correct then our current build is as follows:

limesurvey 3.17.3+190429, PHP 7.3.5, SQL Server 2016

I have physical access to the web and SQL servers so I can check on those locations if there's a file that will contain the relevant version information?

Finally, the reason why I logged this post, is one of our users has recently created a survey, which is set to non-anonymized, however 3 of the responses have an entry for their token in the survey table, but there is no matching entry for their tokens in the tokens table. Is there any combination of configuration which will allow anonymized responses when the survey was set to non-anoymized, or is this an error with the database?

Additionally, this same survey has also had two people respond who appear in both tables, however in the survey table, all of the potential questions shows as blank in the SQL, not NULL, as though a response reached the server but it was corrupt.

I assume this is a different problem, since the symptoms are not the same as the users missing from the tokens table.

d_williamson wrote: I assume this is a different problem, since the symptoms are not the same as the users missing from the tokens table.

The root cause looks outside of LimeSurvey. It's not a simple setting which can cause a missing token inside the response table (if it's not anonymized).
Issues around database or database connection can be the route cause.

If LimeSurvey is running under Windows too, you might check PHP SQL driver too.

If the version of LimeSurvey is not displayed at the bottom (every page on the backend), something else is wrong too.

You are using an old version of LimeSurvey. So you might consider an update to the lastest LS3 version too.

Thanks for the reply, you suggested checking the PHP SQL driver.
Did you mean checking some for specific log related to the driver for errors, or did you mean just to check the version we were running to see how old it is?

Version check and update. There isn't a lot of testing with LimeSurvey on the Windows platform. I cannot remember reports about the issue you're describing here.

We managed to re-create the 2 problems, as detailed below:


1) A users' response was logged in the survey table, but no mention of the token for the user in the tokens table, meaning that vote was essentially anonymous.

• This can be caused by adding and removing the participants on the surveys.

For example, the survey is setup without specified participants, anyone can then navigate to the survey and complete it, leaving no token code in the survey table. If participants are then added, it will add in the token code into the survey table which in turn would result in one entry having a token code and the other not.

Another scenario would be that a survey is created, and participants added to the survey. Those participants then complete the survey, and the owner then removes all participants. When all participants are removed, the token table is dropped entirely, which leaves survey response with tokens applied to them in the survey table but no corresponding tokens in the token table. If participants are then added again, the token table is recreated but the original participants to complete the survey are not re-added to the Token table.

2) A users' response was logged in the survey table, but their response to all questions is blank, not NULL, and no specific answers which should be impossible since it's a dropdown multiple choice (NULL is valid entry). For this situation, the user's details were logged in the tokens table.
• This can be caused by proceeding through the survey without selecting an option from the drop down. For example if an individual presses next whilst the word "Choose Option" is displayed in the drop down menu, it will insert a blank entry into the survey table rather than the answer code as that option has no answer code available.


This would suggest that these are database design problems, no constraint between survey entry and token entry for the first problem, and code needs to be added to prevent users selecting "Choose option".

Can you confirm if this behaviour has already been patched in the current application version?

Anyone?
Ideally I don't want to update the software if it doesn't remediate the problem we're having.

d_williamson wrote: if it doesn't remediate the problem we're having.

I'm currently not coming to the conclusion that the two "problems" you described in your post are "bugs" where you can expect a fix via update soon. Perhaps I haven't fully understand your post. But the response table will just contain the token code with not relation to the participants list. When you remove the participants list, you will need to reimport data when recreating it to match the response data-table.

If you want a reaction from the developers, I suggest to submit two bugreports. In the end it might turn out to be a feature request.

d_williamson wrote: This can be caused by adding and removing the participants on the surveys.

Clearly not a technical issue. Not fixable.
When need to reset a survey : best is always disable/enable.

d_williamson wrote: This can be caused by proceeding through the survey without selecting an option from the drop down. For example if an individual presses next whilst the word "Choose Option" is displayed in the drop down menu, it will insert a blank entry into the survey table rather than the answer code as that option has no answer code available.

Set the question to mandatory, can not submit mofve like this ...

d_williamson wrote: Can you confirm if this behaviour has already been patched in the current application version?

No issue, the no fix.

@jelo

In terms of understanding my post, in scenario 1, we would have expected that if the participants (token) table is removed, then the responses (in the survey table) associated with those participants is also removed.

As for scenario 2, we would have expected that "Choose Option" in the dropdown list is not a valid response to submit.

@DenisChenu
I'll put forward the recommendation you provided for Scenario 1, we'd have liked to have seen something such as a database constraint, that would delete both the token and their response at the same time, but since Lime Survey supports anonymized surveys, I imagine that might not be possible.

As for scenario 2, we'll test with mandatory question as you say.

d_williamson wrote: in scenario 1, we would have expected that if the participants (token) table is removed, then the responses (in the survey table) associated with those participants is also removed.

That will be a feature request for a clean routine, to remove responses with the same token.

A relation would need to consider the anonymous mode, where the token is not saved into the response of a participant.

As for scenario 2, we would have expected that "Choose Option" in the dropdown list is not a valid response to submit.

Choose Option is a valid response as long as a Non-response is valid. To change that you would make the question mandatory.

As suggested, I tried to set the question to mandatory in order to prevent users selecting "please choose" from the dropdown menu, and it still accepted the response without an error. The only difference now is from the survey admin panel, I get the attached screenshot, showing that the participant successfully completed the survey, but the response was not inserted into the table.

I would have expected the user to get an error message if the database isn't going to accept the response, however from the user's perspective, they've completed the survey and there was no error so everything must have gone through okay.