Risk of data loss when opening an experiment with a missing plugin

[dev-jam]

When you open an experiment with items from a plugin which is not installed, the items are missing/invisible and if I remember correctly, there is no warning about the missing plugin.

I am not sure what happens when I open an experiment within an instance of OpenSesame where the plugin is missing, edit the experiment and then save it. Will the original items still be working and at the right location when I open the edited experiment on an instance where the plugin is installed? Or will it be messed up?

Could you add some kind of protection against this?

I was thinking something in the line of a warning at opening and then for every invalid item, substituting it with a 'missing plugin' item. So the items will still be in the experiment. The location of the item is visible, the fact that it is missing is visible and you can edit the experiment without messing up the original missing items.

Additionally, it would be even nicer to have a feature that offers you to install the missing plugin. Maybe it is possible to setup a plugin repository within OpenSesame, maintained online by OpenSesame, where plugin developers can apply by sending the pypi package name and installation can be executed internally through pypi?

Best,
Bob

[smathot]

Thanks for raising this point. Right now, if you open an experiment that contains an item that doesn't exist (i.e. a missing plugin), and then save it again, the item will simply disappear. That's of course a risk, because it means losing information.

There are a few ways to deal with this. The easiest would be to open the experiment without associating it with a file name, just like when you open an older experiment, so that users at least need to explicitly overwrite the original file. A better solution would be to remember the definition of the missing plugin, but that takes a bit more effort. For now it's on the radar.