Multiple question with same code , maybe something here

I can not reproduce after importing and data seems OK.

I've attached a screenshot of the data as it appears in the database. Of the 19 times I have tested the survey, 11 times the data were not saved completely.

Did you test it multiple times?

I'll change the question names and try again with a new survey. I'll keep you posted.

Thanks for the fast response today.

Ok, so in order to rule out multiple questions with the same codes I have renamed them Q1-Q6. After testing this survey (see attached v2) the saving error is still there (see screenshot v2), this time with only 5 correctly saved responses of the 22 test runs I performed.

I am at a loss here...

Could you check if you can reproduce this error? Test it multiple times?

Thanks

I test 5 times with newtest=Y.

Don't want to track down this one.

So what other options are there?

What PHP version is used? 4.1.2 is stated in the ticket.

Yes, the phpMyAdmin version is 4.1.2. I'm not very much into IT so I hope this is the same as the PHP version you guys are referring to. If not, let me know where I can find the correct PHP version.

How did you start the survey each time? What link did you use?

I have used the Survey URL as can be found in the survey menu:
http://www.aronsconsultancy.nl/surveys/index.php/136935?lang=nl

In the database it says the ref URL is this:
http://www.aronsconsultancy.nl/limesurvey/index.php/admin/survey/sa/view/surveyid/136935

I have also tested it with:
http://www.aronsconsultancy.nl/surveys/index.php/136935?lang=nl&PID=XXX&psid=YYY

Same results, I still get missing data.

Never newtest=Y or newtest/Y ....

I just did 20 surveys with http://www.aronsconsultancy.nl/surveys/index.php/136935?newtest=Y&PID=XXX

still 11 cases with missing data.

So how can we fix this?

I can reproduce this using report_error_arrays v2.lss
This seems only to happen if you check the responses at the same time in the administration repeatedly.

I can reproduce this issue when I do nested conditions like in report_error_arrays v2.

I think the error is that the relevance equation is evaluated more than once, so it's random if the two equations equal each other. Will try to confirm.

You can set the seed of the rand function using srand, e.g. srand(1), in survey index class. For me, it then fails every time to save the response.

Basically confirmed that this is the problem. If you disable re-validation in the JumpTo function, this error does not appear.

Solution 1: Pre-process all expressions in the survey and replace rand with the number. This way, rand won't be called twice for one expression.

BUT: Won't work, because where would we save the processed expressions so the survey won't read the original after reload? $_SESSION?

When user are allowed to "com back" for setting alea, i use:
An short text question, hidden: ALEA
A equation:
{ALEA=if(is_empty(ALEA.NAOK),rand(1,10),ALEA.NAOK)}

This usage is really a specific usage , and sometimes : have {rand(1,10)} in equation question type are needed for some user (please : don't preprocess : this break some other system .....)

See another example on manual: https://manual.limesurvey.org/Expression_Manager_Sample_Surveys#Randomly_Ask_One_Question_Per_Group

With , in Equation : if(EQUA > 1 , EQUA,rand())

OK, I have a suggestion for a solution where the result of the rand expression is saved in a cache:

                                // Get unique hash for this rand usage, from question id, group id, params and tokens
                                 $flattened_tokens = implode(flattenrec($this->RDP_tokens));
                                 $randHash = hash('md5', $this->questionSeq . $this->groupSeq . $params[0] . $params[1] . $flattened_tokens);

                                 // Create cache if it's not there
                                 if (!isset($_SESSION['randResultCache']))
                                 {
                                     $_SESSION['randResultCache'] = array();
                                 }

                                 // Return cache result if it's there; otherwise, store and return
                                 if (isset($_SESSION['randResultCache'][$randHash]))
                                 {
                                     $result = $_SESSION['randResultCache'][$randHash];
                                 }
                                 else
                                 {
                                     $result = $funcName($params[0], $params[1]);
                                     $_SESSION['randResultCache'][$randHash] = $result;
                                 }
                                 // Cache is cleared in function killSurveySession

This is in em_core_helper.php, lines 2080 ff.

What do you think?

You mean random order ?

This one is tricky : because:

• In admin : never have random order (else : browse or export are different each time)
• In public : must be in $_SESSION , or save somewhere in DB

It's the function createFieldMap https://github.com/LimeSurvey/LimeSurvey/blob/master/application/helpers/common_helper.php#L1784 , right ?

I always think : first part must be in cache (and not in session) : https://bugs.limesurvey.org/view.php?id=7988

Randomization etc ... must be done in a plugin : https://github.com/LimeSurvey/LimeSurvey/blob/master/application/helpers/common_helper.php#L2338

Remind createFieldMap is used in admin and in public ....

We have a lot of issue with randomization , with less system in survey ....

Did the Expression system : [if(ALEA > 1,ALEA,rand(1,10)) work ?

Fix committed to bug10017 branch: http://bugs.limesurvey.org/plugin.php?page=Source/view&id=17040

I pushed my solution to a new branch, bug10017. To be tested and merged later, I guess.

I already say :

• There are alternate existing solution

But here : you add another function, just need manual update ;)

Maybe unset $_SESSION['randResultCache']) must be in $SESSION['survey'.$iSurveyID] because some user can start (and it work) 2 different survey at a time.

Denis

@DenisChenu & @ollehar can i close this report ! Seems to be very old. Still the bug exists in LS4 ? Please create a new bug report for this for the new version. I will close this one !