Get rid of Optional (#52)

[gavinking]

As proposed in #52, this change kicks Optional out of our code base.

Justification:

• it doesn't really protect you from NPEs because nothing stops you from calling get() on an uninitialized one (unlike other langs, Java doesn't have sum or union types) ... even worse, nothing stops you from assigning null to the type Optional
• indeed, it actually causes more bugs than it avoids, since Optional is assignable to Object
• it makes you create extra local variables to assign the result of get() to (unlike Ceylon, Java doesn't have intersection types)
• its comes with a performance overhead (Java doesn't have value types or union types)
• it makes method signatures uglier (unlike other langs, Java doesn't have a syntax sugar for it)

Does anybody want to step up and defend this 'orrible thing?

[DavideD]

Can you squash this in the previous commit, please?

[DavideD]

Isn't the idea of Optional to allow a more functional approach to code (it has functions like map, flatMap, ifPresent and so on)?

[DavideD]

Do you need this last thenApply?

[gavinking]

Yup, it does a widening cast.

[gavinking]

(This wouldn't be necessary if Java had declaration-site variance grumble.)

[gavinking]

Isn't the idea of Optional to allow a more functional approach to code (it has functions like map, flatMap, ifPresent and so on)?

That's not what "functional" means. (Not for any meaningful definition of "functional" anyhow.)

Look, we have a reasonable-sized codebase that uses Optional throughout. We don't call map() or flatMap() on it anywhere. And we have no reason to do so: the whole reason those operations have those names in ML or Haskell is that there's an abstraction (a type class) called Functor that lets you operate polymorphically over Lists and Maybes and all sorts of other things, letting you write a special sort of highly generic code that is utterly impossible to write in Java (without or without Optional).

So this is just cargo-cult stuff. The only people who think that calling the monomorphic function Optional.flatMap() is making their code more "functional" are people who don't actually know a single thing about functional programming.

[gavinking]

P.S. @DavideD If you like, I can write you a function applyWhenNotNull() that does exactly what Optional.flatMap() does, but that doesn't require you to wrap every one of your objects in a garbage wrapper object.

[gavinking]

The only people who think....

(That comment wasn't aimed at you, of course, @DavideD, more at the people who designed this thing.)

[DavideD]

(That comment wasn't aimed at you, of course, @DavideD, more at the people who designed this thing.)

Sure, but even so... was it really helpful as far as comments go?

Anyway, my point is that Optional is what Java provides and some user might find it useful in some circumstances. It's not that I am a big fan of it, I'm mostly trying to understand if users might find it useful to have it in the API.

[DavideD]

As far as I can tell most programmers don't like it, so I'm not against removing them from the API. I'm not sure this is the right place to discuss it though, as people won't see it.

[gavinking]

I'm not sure this is the right place to discuss it though, as people won't see it.

Right, that's why I opened #52 a couple of months ago ... to try and get some feedback.

(And we have to make a decision on this ASAP.)

[vietj]

Optional is evil

[NathanQingyangXu]

There is a serious drawback for the Optional from JDK. It doesn't implement Serializable for JDK team wanted to avoid the burden to maintain the Serializable backward compatibility.
That being said, Optional is definitively useful in general scenario. Sometimes we wanna differentiate between the following two possibilities:

• something exists but with null value;
• something doesn't exist at all;

see https://softwareengineering.stackexchange.com/questions/364211/why-use-optional-in-java-8-instead-of-traditional-null-pointer-checks for some other reasons

No idea about the hibernate-rx relevance, though.

[aguibert]

+1 on removing optional

[gavinking]

Sometimes we wanna differentiate between the following two possibilities:

• something exists but with null value;
• something doesn't exist at all;

This is indeed something that arises occasionally (really the only case I know of is caching/memoizing), it's not a distinction we need to make in the API of hibernate-rx.

[DavideD]

NIce, thanks