Yes, right.

Think need more survey settings here:

• Set completed at first completed
• Set completed each time completed
• Set completed at last completed (default)

And send confirmation at same thing.

But thing we have to move it in feature request

In my opinion the data in the token table is redundant.

Instead just search for the responses related to each token and show whatever information you prefer (first completed, last completed..)

The only real information is the usesleft column; the other information is just redundant.

When anonymous, you can not link to the responses table so it is not redundant

Redundance only for anonymous token , if not completed have the date.

But with more settings:

• Set completed at first completed
• Set completed each time completed
• Set completed at last completed (default)
  It's not redundant.

And , look at : http://bugs.limesurvey.org/view.php?id=7797#c25333

Good point Menno, forgot about that!

With anonymous survey : before enter the survey we test the completed state and the useleft state.

1st : we test the useleft with common_helper/function usedTokens
2nd : we test the completed state (survey/index line 464)

This second test is redudant.

There should be one test in the token model: that test should take care of checking the usesleft and/or completed status. This way we don't need to pass all kinds of rules around the application and keep them in one place.

Keep in mind that usesleft is a feature introduced way after the completed date. The completed date was the only way of checking completion before so that will be all over the application. When usesleft was added some places in the application were updated to make use of usesleft, while at other places i can imagine the completed date being N is still the only check. So to change the behaviour of the compelted field with a setting we must be very careful that the right completion check is done otherwise that might break the application at places we don't expect.

Oh, and for http://bugs.limesurvey.org/view.php?id=7797#c25333 solving that in rules would be the best. Make sure the interface is clear about the possibilities.

@mdekker : I know for old system without usesleft, like a lot of old function in survey/index.php ;).
And you're right: need to test whole right acess in one function only.

replace the common_helper/function usedTokens by a:
SurveyRuntimeHelper/tokenAcess function maybe ? Or you want it in some model ? (Tokens_dynamic?)

What do you mean with "solving that in rules" ? I think first the same think but after more thinking : have an option for a "global" password survey can be a good think to keep.

Denis