Hide ViewActions implementation details. Javadoc normalization.

[JakeWharton]

Refs #91.

[JakeWharton]

These actions are so trivial to implement that it's a giant liability and waste to expose them as public API.

[dlew]

👍

Curious API design question: would you still use checkNotNull if we add @NotNull available?

[hamidp]

👍

Also +1 for @NotNull

[ronshapiro]

Any reason for anonymous classes vs. just package-private ones? Core seems to prefer having the implementation of the classes stowed away in their own files.

[ronshapiro]

@JakeWharton what's your reasoning for not wanting to keep a WeakReference to the view? I remember @mttkay and @dpsm expressing interest in this a while back.

[JakeWharton]

would you still use checkNotNull if we add @NotNull available?

Yes. @NotNull does no actual validation, just hints to static tooling which can be ignored by callers.

Any reason for anonymous classes vs. just package-private ones?

Just because they're like 6 lines long. In reality they're 5 lines of boilerplate and 1 line of code. They would be lambdas had we Java 8 language features.

I think actual classes (even if package scoped) are only deserved if there's something interesting happening beyond just implementing a single-method interface or abstract class with one line of code.

what's your reasoning for not wanting to keep a WeakReference to the view

It conflates the responsibility and importance of a subscription. If I assume responsibility for managing a subscription then there's no harm in library code holding a strong reference.