Now, that's an interesting idea. Yeah; currently, every individual option in the list is weighted equally, which means that sub-lists with more individual options end up with greater weight. (This is why anything generated with the Aarne-Thompson list will probably end up mostly Aarne-Thompson; with over 2000 items in it, it's hard for anything else to compete.) It seems like there should be a way to select randomized sub-lists from the selected lists as they're loaded in – it'd just take some heavy listing on the JavaScript side of things, as by the time the processing hits the server side (which is where I'm accustomed to doing the heavy lifting), there's no attached information to indicate which specific prompts came from what source.
So, the implementation would probably be to have a "Include a maximum of x prompts from each selected list" box, and when it was selected, it would go through a process where:
* It looked at all of the selected lists * It retrieved the data for each list, broke it up into an array of values, and selected a randomized set of up to x options * It loaded those into the editor view.
Which seems possible, if potentially processing-heavy. And it would probably not be terribly kind to people who wanted to edit the items included on each list before limiting them; that is, selecting a maximum of x prompts would have to select them from the entire list, not any subset of the list, unless I wanted to hugely redo the code architecture.
Re: Hmm...
Date: 2014-02-10 08:21 pm (UTC)So, the implementation would probably be to have a "Include a maximum of x prompts from each selected list" box, and when it was selected, it would go through a process where:
* It looked at all of the selected lists
* It retrieved the data for each list, broke it up into an array of values, and selected a randomized set of up to x options
* It loaded those into the editor view.
Which seems possible, if potentially processing-heavy. And it would probably not be terribly kind to people who wanted to edit the items included on each list before limiting them; that is, selecting a maximum of x prompts would have to select them from the entire list, not any subset of the list, unless I wanted to hugely redo the code architecture.
Still, it's an idea I can play around with.