Normalize flatten

I think it would greatly simplify downstream usage of twarc.expansions.flatten in plugins if flatten() always returned a list of tweet objects regardless of whether it was passed a bundled tweet {"data": {}}, a bundled list of tweets {"data": [{...}]} or an already flattened tweet {"id": "..."}.

[edsu]

Thanks for those pointers @igorbrigadir – it would be great if they solved this problem for us. But I think we should move forward with our own solution until it materializes.

[edsu]

Because it is seen as additive, not destructive, I’m pretty sure flatten() returns the same structure it was given, e.g {"data": []} or {"data": {}}. It can be a bit tedious to have to unpack that in plugins, and I found myself copy pasting the same code.

After a night’s sleep I think @SamHames is right and that flatten should be left as non-destructive, and what is needed is something new that will create a singular JSON representation based off of either the v1 or v2 representation. So I’m going to close this and focus on #412 instead. It’s just a bigger job because I need to read the standard more closely.