Make model parameter optional for Backbone.sync

[creynders]

I encountered a situation in which I needed to call Backbone.sync directly, w/o a model, which works perfectly except it will throw when trying to trigger a 'request' event on model. Obviously I could easily pass a trigger:noop object or something, but thought this was cleaner. Test included btw.

[akre54]

Sorry but -- what's the use case? Every extra conditional is one more potential bug.

[creynders]

It's a bit hard to explain, but basically it comes down to this: I have a number of things that are automatically attached to all Backbone.sync requests (e.g. a unique ID for each request, centralised error handling et cetera), which is why I'd like to use Backbone.sync for a multipart/formdata RPC request, to benefit from the extras, but all the data it's sending is massaged and collected, i.e. there's no specific model that I need to use. As I said, I can easily solve it, but I assumed that model was in fact meant to be optional since on this line it's truthiness is being checked as well, i.e. why bother checking if it's required anyway? If you dig through the function you'll see there's no real dependency on model except to use as data if none is provided and to send out a 'request' notification, which struck me as it being optional. But no problem if it's too much a hassle.

[akre54]

To be honest, I'm not sure why that's there (but it's been there for years), along with the model.trigger. I think wrapping sync with a dummy object is going to be your best bet here, as yours is probably not a use case we want to be supporting. I'll leave this issue open a little longer in case any other contributors want to chime in.

[jridgewell]

We should either add this, or remove the&& model guard.

[Florian-R]

I was curious about this guard and for posterity it was introduced by 4c1bdb46a8aeba6a6fc9803614b074dbca2ec050. Seems safe to remove it.

[creynders]

Back then the guard made totally sense, since model was optional... Which I think TBH it still should be 😁

[akre54]

or remove the && model guard.

I thought of that (removing it doesn't fail any tests), but what's the harm in keeping it?

In this case it's probably fine to remove, but I think recent history has shown that we tend to get bit by edge cases we think are safe to get rid of. Let's just make sure we have tests covering the expected behaviors.