Poll Variants and Voting methods.

[artlog]

Poll variant generic #4043 requires to discuss architecture, design and implementation.
Root of this discussion can be found in #3897.

paraphrasing #3897.

poll type - What kind of options are proposed:
    date
    text
    pictures
    ...

voting variant - How participants can vote:
    yes|no|maybe|null ( default : "simple" legacy 'Opted variant' ) 
    configurable list  ( actualy "generic" variant #4200 )
    ordered
    points ( not yet implemented )
    stars ( not yet implemented )

result variant (currently missing) - How is the winner determined:
    counting points
    most points
    Condorcet method
    Majority judgement
    most yes
    ...

How the winner is determined is not enough, representation of result can differ, by example how ex-aequo are displayed, if second choices are kept etc,,,

[artlog]

For generic variant i wonder if answers could be seen as another kind of Options.
In majority judgement case those can be named Mentions and are finaly just an integer value of the order, a reference of a Mention.
This Mention can even be abstract whenever it is an order ( 'first','second',...) especially for condorcet method and could then easily be handled with i10n.
A vote is then an association of a Mention for each Option. Options or Mentions are a list of choices.
Current implementation in #4200 is based on a string value, but could be a reference in Mentions.

[dartcafe]

For generic variant i wonder if answers could be seen as another kind of Options.

Answers are no options. The options are the poll creator's offers, votes/opt-ins/mentions/orders (generic or opted) are the participant's responses.

[artlog]

yes i understand that in our app context an Option is a Candidate.
Idea was more to reuse some properties of poll_option table, like fact that it is a list of possible items dedicated for a poll and that they handle ordering. And that by creating a new table for mentions/answers not hardcoded and mightbe supporting internationalisation this way.
I was interested too to reuse/copy some frontend parts of Options for Mention support.
But this is not that important, this is only on subtopic.

[dartcafe]

I don't get. What do you want to achieve? Any use case or user story?

[artlog]

To go further the result poll should be studied.

This result should display the Best Option computed on all votes, this is the minimal requirement.
Constraints are different given the vote method.
It is a real question to be able to enforce constraints during poll. afterward it will disappointing if at the end vote should be redone since constraints where not compatible with vote method.
It seems legit then to consider that vote method should be configured at poll creation time.
But here we could support alternate method grouped by compatible constraints, that is allowing to compare results over multiple (compatible) methods without requiring to recreate another poll. This seems complicated still.
Focusing on Condorcet and Majority Judgement by example both are not compatible or through tweaks that can defeat the purpose and not reflect voters will.
Condorcet requires that there are as many mentions as options and that mention is used for one option only, acting as an ordering of options.
Majority judgement fits in any generic polls without additional constraint given missing mentions are seens as reject.

And then comes the submission of full vote to obtain the result. I would prefer a submission fully anonymous, without any hint on semantic, just option id and mention, then this can be delegated to an external vote API without disclosure.
And best would be to support existing voting formats.

[artlog]

constraints :

Each poll method comes with constraints.
Whenever a vote is casted then constraint are checked for the given method.
Then constraint is somehow part of voting api.
A feedback should be given to accept or refuse vote with indication of what is refused.
This feedback should be displayed to user to help him to adapt to constraints.

constraints are in any case at least checked on server side, but there could be a frontend only dedicated method to ease process.
there can be specific variants that embed frontend constraints and apply only to a subset of methods.