I've seen people doing:

private val _foo by lazy { MutableLiveData()}

val foo: LiveData by lazy { _foo } // why second lazy?

Does this second lazy really make sense? Seems like

val foo: LiveData = _foo

val foo: LiveData get() = _foo

should be enough.

I think

val foo: LiveData = _foo

would initialize

_foo

immediately. To keep it all lazy,

foo

must be lazy as well.

val foo: LiveData get() = _foo

?

That should work I think. Try it to see what happens.

yep, the lazy is redundant: https://pl.kotl.in/rko7o8XfN

Correct. You can look at the source and see that lazy will only execute the block if the stored value is

UNINITIALIZED

No need for 2nd lazy

I couldn’t understand why lazy is used to init LiveData. I know what it does but why?

You won't gain much if it's a livedata that is being observed all the time the view is alive. But you can have some livedatas that you subscribe to under a certain condition, so if condition is false — you don't create it for nothing.

I haven't really used it (I've been trying Channels instead), but given livedata is the observable, aren't you going to inevitably send data to it?

not, unless you observe it. It's pretty much a very primitive version of Rx stream (BehaviorSubject).

How are you going to post events without initializing the livedata? Whether or not you use the events is a different story. Similarly with publish/behavior subject and channels; wasn't the point that the observable did not have to be aware of the observers when sending content?

Thanks for explaining.