How should the Ranking question behave?
What does it mean a ranking question to be mandatory?
Does it mean that all of the options need to be ranked?

Then that number of options to be ranked can be updated using max?

Am I right?

Hi Gabriel

Thanks for picking this up. IMHO a mandatory question is one that "Must" receive a response in order to proceed. However I believe that any conditions / sub options that are also set need to be respected.

For the purposes of the example it is assumed there are 4+ SubQuestions in the Ranking Question:

So for instance, in the ranking question that is mandatory and has the MIN (blank) AND MAX 3 answers then the behaviour should be:

• 0 Ranked: Mandatory Condition Not Met - Do Not Continue.
• 1 - 3 Ranked: Mandatory Condition Satisfied, MIN 0 & MAX 3 Condition met - OK to Continue

However in the event that MIN == 3 & MAX == 3 then user must rank 3 SQ's to continue.

To me this is logical. However currently if you set a MAX value less than the total number of SQ's and Mandatory is on then it is not possible to complete the ranking question. Hence you have to turn Mandatory off which means you risk a question being skipped.

I hope that all makes sense, but reply if anything is not clear.

Thanks Marc

Yes, I think the issue here is that mandatory is making all SQ to be ranked, while it should be, at least 1.
Doing it just like that seems a strong thing to change, as per backward compatibility.

So, I would do,
If mandatory is set and no min nor max, I wouldn't update the behaviour (prior compatibility)
If mandatory is set and min or max, mandatory would mean at least 1 option ranked.
@c_schmitz, what do you think about this?

That would work assuming it didn't interfere with the options.
but as a purest it would make send to code around the conditions

Let me rephrase that last comment.

That would work assuming it didn't interfere with the options.
but as a purest it would make sense to code around the options (MIN / MAX) conditions

But I suppose, if Mandatory was set and that condition was satisfied by selecting one choice, if the option Min 3 was set then it would trigger the error condition around the option.

@c_schmitz, what do you think about this?

I agree to the concept of making a smaller change, to not break BC.

Maybe Fredrik has an opinion too? Will ping him.

@f_funke ?

Fredrik is on vacation this week.

I could reproduce the issue with 5.0.4. The behavior illustrates a general flaw of LimeSurvey, which also affects other question types like multiple choice.

Currently, logical rules implicitly make a question mandatory. So if minimum or maximum answers are set in the logic section a question has to be answered to proceed in the questionnaire. But sometimes a survey researcher just wants to communicate how answers should be ideally answered while allowing the respondent not to comply with these instructions (in paper and pencil questionnaires respondents can neglect instructions, too).

I'd prefer to disentangle the option for making a question mandatory and the question logic and suggest the following general behavior:

• the mandatory option (on | soft | off) is used to determine if a response is required
• the logic (e.g., min/max answers) is used to specify how a question should be answered

The independent combination of mandatory and logic gives researchers most flexibility:
case 1: mandatory == no and logic == no: no answer has to be given, question can be skipped
case 2: mandatory == yes AND logic == no: question has to be fully answered (here: all SQs have to be ranked) to proceed
case 3: mandatory == yes AND logic == yes: the rules described in logic (here: min/max) have to be met to proceed
case 4: mandatory == no AND logic == yes: information on max/min responses/numbers is displayed in the questionnaire but respondent can give any or no answer and proceed

Currently, only cases 1 and 2 are implemented in LimeSurvey. Case 3 and case 4 most probably mean a bigger change without backward compatibility. But in my eyes, this would make LimeSurvey more versatile.

Additional remark on min and max answers:
case 1: If only minimum answers are set, the instruction is "Please select at least n answer(s)".
case 2: If only maximum answers are set, the instruction is "Please select at most n answer(s)".
case 3: If minimum and maximum answers are set, the instruction is "Please select from n to m answers".
case 4: If minimum answers equals the maximum answers, the instruction is "Please select n answers".

I think that it would be better to update the wording for case 4: "Please select exactly n answers"

So @f_funke, we can apply the change suggested here?
https://bugs.limesurvey.org/view.php?id=17330#c64688

I would apply it and then review how the system reacts and compare it to your notes.
What do you think?

Sounds good to me. Also to @c_schmitz?

PR: https://github.com/LimeSurvey/LimeSurvey/pull/1941

Finally, we made changes for this behaviour:

Mandatory Validation

• If min_answers or max_answers is set, there must be at least one response. Failing message is "Please rank the items.".
• If neither limit is set, all options must be ranked. Failing message is as usual: "Please rank all items."

It is not exactly how Fredrik wanted as mandatory validations and max/min validations are separate processes.
Still, I think it is OK.

Still, there is a case in which redundant messages appear.

When

• a min or max is set
• and you do not select any option.

In this case, two messages will appear:

• one saying that a question was not answered (mandatory validation)
• other saying that a question was not answered in a valid way (in / max validation).

I still didn't understand : it's a conceptual logical issue …

I can report an issue on numeric question type where i put min to 10 and max to 5 ?

Fix committed to 3.x-LTS branch: http://bugs.limesurvey.org/plugin.php?page=Source/view&id=32185

PR for Master: https://github.com/LimeSurvey/LimeSurvey/pull/1966

Fix committed to master branch: http://bugs.limesurvey.org/plugin.php?page=Source/view&id=32304