That's a good idea, but there is one problem.

The validation rules state that if you have min/max values, and the question is n on-mandatory, then a blank value is OK (good). This may seem non-intuitive, but that is the way the validations work in 1.91 too (they just don't show color-coding).

I guess an alternate process would be to do a two step validation:

if (empty) { / set to now class / }
else if (passes validation) { / set to .good class / }
else { / set to .error class /

<q>The validation rules state that if you have min/max values, and the question is n on-mandatory, then a blank value is OK (good).</q>
Then there are problem with mandatory ( check at the second question in exemple: mandatory but good class on em_value).

I think the best way is an .empty class:
if (empty) { .removeClass('error good).addClass('empty') }
else if (passes validation) { .removeClass('error empty).addClass('good') }
else { .removeClass('good empty).addClass('error') }

:)

PS: in 1.91: all validation are made in PHP, and if validation don't work : input-error

PS2: it's important to have good/error/empty for each em_validation because if we hav 2 the one can be red and another green ( can't use .input-error .good{color:red;} )

I like the idea of two-step validation, but it will require 4-8 hours worth of refactoring and re-validation, so not sure if can get to it before 1.92 RC4

I don't understand PS2. Currently, you can do the following for each em_validation type:

(1) .error {color:purple} // errors when you first see them - before submitting with error
(2) .good (color:green}
(3) .hide-tip .good { display: none } // hide the tip if no problem
(4) .hide-tip .error { } // show the tip in appropriate color if there is error
(5) .input-error .error {color:red} // so visible and red if submit an invalid answer

Through inheritance (e.g. you don't need to explicitly put these), you get
(6) .input-error .hide-tip .good { display:none } // so continue to hide if good
(7) .input-error .good { color:green } // so still green if OK even if input-error in rest of question.

(7) .input-error .good { color:green } // so still green if OK even if input-error in rest of question.

Yep, but actually , we have .good on error tip ( i think)

Look at the second question of http://ls-dev.sondages.pro/index.php?sid=76597&amp;newtest=Y&amp;lang=fr the we can't take
.mandatory .good{color:green}

It's why i think a .empty class are the best.

( for 1.92 RC4, no problem even if it's for 1.92 release 2000, i know i can't make modification directly, maybe look for some patch ;), i put low on priority).

Yes, there is .input-error .good there, since you can a whole question fail at least one validation rule (so gets .input-error), but not fail all of the rules.

So, in your example, it fails the mandatory rule, but not the validation rules (which currently allow blank values).

I think you've hit on a different issue. Right now, most validations let you keep the answer blank; and mandatory is handled separately. Perhaps they should be handled together, so only consider a blank value OK if the question is non-mandatory.

We should probably discuss, as a team, the full set of different styles we'd want for +/- mandatory, +/- empty, +/- any em_validation +/- input-error +/- fails validation rule to be sure we've covered all the desired possibilities.

All the more reason to hold off on this for now rather than rush it into RC4.

Is this still a problem?

I think there are a difference between NA/empty and error.

Allowing blank value aren't a Validation succes, it's an Validation not-applicable

:)

I don't make test with different combination, but allways think best way are a good way.

Note to self. To make this work:

if (empty) { .removeClass('error good').addClass('empty'); }
else if (passes validation) { .removeClass('error empty').addClass('good'); }
else { .removeClass('good empty').addClass('error'); }

However, to do if(empty), need to refactor EM-generated validation JavaScript. Currently, validations are like if(empty(x) || x <= 5); and those are concatenated . Would need to split apart empty tests and validity tests for each question attribute, then do the same in GenerateRelevanceAndTailoringJavaScript()

Also requires tweak of _ValidateQuestion to do same logic (to ensure that isempty() and regular validations are both done to confirm that submitted data is accurate). - so update of qstatus to add 'isemptyJS'

I had 2 idea.

Add 3 javascript function for em_validation:
addValidationEmpty
addValidationSuccess
addValidationError

And make something like that:
if (empty) { $(this).addValidationEmpty(); }
else if (passes validation) { $(this).addValidationSuccess(); }
else { $(this).addValidationError(); }

But you can put this "bug" in developpment and assign to me.

Or maybe i had to put a new bug/developpment, test this one, and if class is good: close this one.

For input.text, isempty is easy
For radio-list (in Yii actually) :, $(#questionID .radio-list .radio-item:not(noanser-item)).leght() == 0

But #questionID aren't good for array for exemple.

I have an idea for that : send it to mailing list:
The idea

Put #questionSubQuestionID to all SubQuestion, i think after it's more easy to manage EM-validation by subQuesqtion: code is the same.

Problem : actually we have some javaConditionsSubID: we have to replace with questionSubID in HTML AND in javascript.

I think it can be cool to have:

#questionQuestionID and in database surveyIDXgroupISXQuestionID, the we can allways split SESSION var to work on #questionQuestionID

What do you think of this ?

Denis-

Thanks for the ideas. However, the hard part is refactoring EM to generate the

if(empty) { }
else if (passes validation) { }
else { }

parts. Once they are in place, adding/removing classes is really easy.

Also, there is so much interaction between relevance, hidden, validity, and mandatory status

So, why don't you give me a few days to see if I can tackle this. If not, I can guide you as to what needs to be re-factored in EM to make it work.

I'll work first on updating the in-line and wiki documentation for developers to clarify the architectural design and goals of EM.

Although I know we are a community and should be allowed to do things our own way, EM was designed with a specific approach in mind that, if followed, should make it much easier to extend and maintain.

I commit a fix for citronade:
function addClassEmpty(){
$('.answers-wrapper input.text[value=""]').addClass('empty');
$('.answers-wrapper input[type=text][value=""]').addClass('empty');
$('.answers-wrapper textarea').each(function(index) {
if ($(this).val() == ""){
$(this).addClass('empty');
}
});

$("input.text,input[type=text]").live("blur", function(){
  if ($(this).val() == ""){
    $(this).addClass('empty');
  }else{
    $(this).removeClass('empty');
  }
});
$("textarea").live("blur", function(){
  if ($(this).val() == ""){
    $(this).addClass('empty');
  }else{
    $(this).removeClass('empty');
  }
});

}

Seems to work .

( I like external JS function, more easy to fix and replace).

Denis-

That does seem like a reasonable solution. And, it doesn't require any EM changes.

So, we could keep the EM logic intact (which does this):

if (passes validation) { .removeClass('error').addClass('good'); }
else { .removeClass('good').addClass('error'); }

And your function would add/remote .empty

And template.css could be updated to always show default text style if .empty is present.

So, handing this one over to you...

/Tom

For 2.0 or 1.92 ?

2.0 please.

http://git.io/01VaFg

Put on a new feature

For difference between good and not-applicable: better to use directly EM.
See http://bugs.limesurvey.org/view.php?id=6080 for example.

Version 2.00RC2 released.