I'm not sure if this is related: When using EMs rand() function to create a random integer number and then applying conditions to it, do i have to use
((equationQuestionCode == "1"))
...or does that compare the rand number with a string value so hat I should better use
((equationQuestionCode == 1))

?!?

(equationQuestionCode == "1") should be fine (but you can also use the number). Strings that are numbers are automatically cast to numeric.

Mazi-

When I import this survey, it shows an error for the relevance. The first randnumber.NAOK is shown as the SGQA code (e.g. it isn't recognized on import).

When I manually edit the condition to fix that problem, the generated equation is correct:

((car_owner.NAOK == "A1") and (randnumber.NAOK == "0")) or ((home_owner.NAOK == "A1") and (car_owner.NAOK == "A1") and (randnumber.NAOK == "1"))

So, I can't reproduce this yet.

However, oddly, I do see the problem relevance equation within the .lss file. Perhaps there is also an import problem since the variable is declared within the same group that the condition is used. Although, I haven't seen that problem before either.

Hi Tom,

I have tried to reproduce the problem but whatever I tried, it now shows up correctly. I wonder how I managed to get this done so it convert from, maybe it was just my fault.

nevertheless you are right, there is a problem at import. The GSQA value for the question question isn't converted correctly, see attached screenshot.

The import is working correctly.

There seems to be an error in the source database:

<row>
<cid><![CDATA[763]]></cid>
<qid><![CDATA[3627]]></qid>
<scenario><![CDATA[1]]></scenario>
<cqid><![CDATA[3626]]></cqid>
<cfieldname><![CDATA[19966X84X3633]]></cfieldname>
<method><![CDATA[==]]></method>
<value><![CDATA[0]]></value>
</row>

Note the invalid cfieldname, if should end with 3626. Question is how that happened in the first place? How old is that database, which versions were used to edit it? Why are there duplicate question codes at all? (randnumber?)

You can see the error in the first of Mazis screenshots, too.
'randnumber' has two different question IDs if you look closely.

Since this is so hard to reproduce, I suggest we close this matter until someone finds how how exactly to reproduce it.

Carsten, this survey was created on Limeservice (account: http://actions.limequery.com/admin). The database is pretty fresh. There is just one test survey and this one.

Tom/Carsten: Since this is an important survey that should go live soon, do you recommend to delete the equation question and re-create it?
If so, can I leave the conditions? They were not created using the condition editor but using the relevance field and if I use the same question code, should it all work fine?

@Tom, using the attached survey, can you reproduce the error that the progress bar doesn't work correctly on some pages?
I couldn't always reproduce the problem for every page, seems to be related to conditions. I attached another screenshot.

I remember there was an error in the conditions editor a little time ago.
Maybe this is/was an after-effect.
Just dropping and creating the conditions again should do the trick, no need to delete the equation question.

Do I have to re-create the conditions at all for technical reason or to be sure that it all works well?

Because I thought that when using question codes and relevance only (no condition editor), there is no reference to any question ID anymore but only the code is used?
If so, there would be no use to re-create all the conditions I would say?!

Mazi-

To recreate the relevance from conditions, you can either re-import the survey, or use the EM menu tool to convert conditions to relevance for the entire survey. Then check the Show Logic file for any errors. As long as there are no errors, the survey should be fine.

I mean only the conditions where you are using the conditions editor and that don't work as intended.

Tom, one more general question since I'm trying to debug some other odd behavior: if there is a question with some customized JS at the first question of a group at a survey run in group by group mode, when does EM process that JS?

• Only if there are any conditions at one of those questions?
• Only if an equation question exists at this group?
• Always?

Different topic (but I wanted to ask before creating a bug report): Did you notice any odd/different behavior when using the rand() function inside an IF statement which refers to previous questions at an equation question if you run it in group by group or all in one mode?
Should EM be able to check the previous questions in all in one mode and then apply the rand function to generate a random number, or does the equation question need to be placed on a following page?

Mazi-

Custom JS will probably run regardless of when EM does it processing. EM gets called (via checkconditions) at load time, and then every time a value is changed. When custom JS gets called depends on when it gets loaded relative to document.ready(). I'm not an expert on that stuff. So, I recommend you pick some sample use cases and use FireBug to step through the code and see when the custom JS is called. Bottom line is that EM doesn't try to encapsulate custom JS to prevent it from being called except for situations where the question is relevant.

I didn't try your custom rand() functions. However, if they are part of relevance or EM in general, rand() will be called multiple times, which often isn't what you want. So, see this example, which ensures that rand() is only called once. http://docs.limesurvey.org/Expression+Manager+Sample+Surveys#Randomly_Ask_One_Question_Per_Group

/Tom