Bug still exists in Build 130108. I will donate/pay if this is fixed within 24h

Menno, can you take a look at this one as soon as possible, please?

I attempted to export a survey structure that included group randomization to a .txt file, and found that no information regarding the group randomization was exported. Thought this might be related to the general randomization problem rather than representing a new bug. WILL DONATE IF RANDOMIZATION IS FIXED SOON!

This won't be an easy fix and will take some time to get it right. It seems to work for all-in-one mode not not in group by group mode

Fix committed to master branch: http://bugs.limesurvey.org/plugin.php?page=Source/view&id=11372

Fix committed to master branch: http://bugs.limesurvey.org/plugin.php?page=Source/view&id=11373

Did not look at export/import yet but this two patches should fix the group randomization problem.

Please test and report if you still have trouble. I tried with a simple survey using three groups that each had just one question to keep it simple. Because it is random sometimes the order can be the normal order. If you clear results and restart the survey you get a fresh randomization.

I think in my export (limesurvey_survey_543192.lss) there is info about randomization and it seems to me it is imported correct.

I do some test: with a group by group survey : i have empty group.

I try some patch in 2.1 for group ranbdomization, never find a good way.

But think we didn't need to unset session group list each time.

Can you have a look at this patch:
https://github.com/LimeSurvey/LimeSurvey/commit/cb4487bc44c11ae7c1238c0a04b2f1783340879b

I don't iunset session group list if is already exist, just update if language is different.

Then it don't break actual group order.

I try to do some test again.

for me it doesn't work, neither for your test survey nor for my survey

@mokoe Just to make sure: did you use the updated files attached to this bug or your own unaltered installation?

both. for me it is always g1, g2 and then g3.

Menno : Got it \o/

killsurveysession kill $SESSION['survey'.$surveyid]

but seems EM use:
$_SESSION['fieldmap-' . $surveyid . $SESSION['survey'.$surveyid]['s_lang']]

For the group order.

This $_SESSION['fieldmap-' . $surveyid . $SESSION['survey'.$surveyid]['s_lang']]
MUST be :

• put in $SESSION['survey'.$surveyid]

But better : removed (think it's the same than:
$SESSION['survey'.$surveyid]['fieldnamesInfo'] )

In 2.1 : there are $SESSION['survey'.$surveyid]['question'] object.
Then a lot of change to do ....

Maybe for 2.0 we can test with killing
$_SESSION['fieldmap-' . $surveyid . $SESSION['survey'.$surveyid]['s_lang']] too

Denis

Fix committed to master branch: http://bugs.limesurvey.org/plugin.php?page=Source/view&id=11387

it's now working for me. thank you very much for the support!

Menno : great job here :)

Menno:
Seems i have some:
PHP notice
Undefined offset: 0
at /application/helpers/SurveyRuntimeHelper.php
$g = $_SESSION[$LEMsessid]['grouplist'][$gseq];

Strangely, i have again some emty group now. Maybe with token (and no randomized group).

Empty group only with debug=2 (or 1 surely), but error not shown in survey. Hard to debug this one :).

Denis

Fix committed to master branch: http://bugs.limesurvey.org/plugin.php?page=Source/view&id=11409

Think you found the last problem :) Great job! I made sure the grouplist has index groupseq, so a number starting at 0 (the order).

Also found some places where for the elements of grouplist it was looking at [0][1][2] instead of ['gid']['group_name']['description']

It's allways a bad idea to use [0] [1] ...

Seems better now.

and now the hard part, find all relevant changes and port them to 2.1 ;)

I review it after.

Think i use usort and function to sort group_list and not unset/reste his.
Same think for question object, surely the question object is harder to port.

Put http://bugs.limesurvey.org/view.php?id=7295
for 2.1/Plugin

Think this one can be closed, but a lot of code to copy/paste ;)