OpenFeign Compatibility with Android SDK < 26

[masc3d]

feign-10 is now incompatible with < sdk 24 (when using eg. JAXRSContract which uses java8 lambdas), equalling 50% of android devices out there.

[masc3d]

throwing NoClassDefFoundError on

removeValues(consumes.value(), (mediaType) -> emptyToNull(mediaType) == null, String.class);

maybe it's not the lambda though, as those should be supported on all devices by now.

[masc3d]

caused by java.util.function.Predicate in feign.Util which is unsupported on sdk <24.

[kdavisk6]

I was not aware of this specifically. We made a conscious decision to move to Java 8 knowing that we would have some issues with Android. I'm open to suggestions on how do you think we should proceed.

[masc3d]

We made a conscious decision to move to Java 8 knowing that we would have some issues with Android

was there a public discussion I could read up on @kdavisk6?

well it probably implies that feign >= 10 will not be feasible to use for majority of android projects for what I would estimate to be 5-10 years until < sdk 24 devices become irrelevant.

a pretty harsh decision. I would rather have opted to move feign to kotlin to maintain portability and have better language features at the same time.

is it too late to re-evaluate this?

[kdavisk6]

It was discussed in #656. We knew that this would be something that will come up. The consideration towards Android was, let's see how it goes. And now, here we are. Moving to Kotlin is a more extreme move than moving to Java 8, and is something that we would encourage by creating a new library.

Knowing that, what do you suggest we do? From my perspective I see a few options.

• Restrict Java 8 language usage to that of Android
• Require that Android developers use an SDK higher than 24 when using OpenFeign 10.x
• Create a feign-android specific version, that deals with this issue specifically.

We need to strike a balance between providing support for as many platforms as we can, while still moving the project forward, adding capabilities that allow us to do so. Java, which has been stagnant for some time, is now moving very quickly forward. From my perspective, which is just one of many, this is just the first step toward Java 10 and 11, which will bring another set of challenges.

Thoughts?

[masc3d]

Restrict Java 8 language usage to that of Android

not feasible if you want to extensively leverage language features plus requires verification mechanisms to prevent unsupported java8 features from leaking in.

Require that Android developers use an SDK higher than 24 when using OpenFeign 10.x

won't happen with > 50% unsupported atm and current rate of android moving forward. I cant imagine any project which can afford to ignore 50% or say even 25% of android devices.

Create a feign-android specific version, that deals with this issue specifically.

lots of forking and maintenance. that could easily become more extreme than moving to kt.

[masc3d]

Moving to Kotlin is a more extreme move than moving to Java 8, and is something that we would encourage by creating a new library.

I don't see it that extreme. can be applied incrementally as interoperability is excellent. I would rather go this route and get more balanced, elegant and powerful language instantly while preserving android compatibility.

then let's see where java is in 5 years. compared to kt its now (at least) 5 years behind.

[kdavisk6]

Kotlin is an different programming language. While there is interoperability, it's akin to suggestion that we port to NodeJS, Python, or any other language. It's the opinion of OpenFeign that large integrations/ports, such as the one you are proposing, be done as separate repositories. See HACKING, for more context on that.

With that being said, the only major change in 10.x is moving to Java 8. Is there a feature in particular that is in 10.x that you would like to be able to use? Would a backport of the feature into 9.x (which is still on Java 6) be adequate?

[masc3d]

it's akin to suggestion that we port to NodeJS, Python

no it's not. those don't even relate to jvm or would compile to jvm bytecode without considerable footprint.

Is there a feature in particular that is in 10.x that you would like to be able to use?

no. not yet.

is android support relevant to you and will you accept PRs that remove java8 features incompatible with android for 10.x?

[kdavisk6]

Removing features that have already been accepted feels like bad form to me. However, if you were to propose an alternative implementation of a feature, adding support for other platforms, that's something I think everyone could get behind.

And to make something clear, I speak for myself and myself alone. There are other maintainers on this project with their own opinions and concerns. They may agree or disagree with me on this issue. Regardless, we made the decision to move to Java 8 on knowing the risks. I understand and respect your position and your advocacy for continuing to support a wider range of Android platforms. There is no reason, in my mind, that we cannot continue to do so, to a point.

It's not feasible for us (the maintainers), to impose platform restrictions on new features and contributions, as not all of those submitting changes are familiar or aware of those platform restrictions. On the other hand, the community cannot expect us (the maintainers) to know about where this library is used all of the nuance that goes along with that. We will continue to rely on the community to help here.

[velo]

is android support relevant to you and will you accept PRs that remove java8 features incompatible with android for 10.x?

We discussed using https://github.com/luontola/retrolambda to create a java 6 compatible feign.... But, no one demonstrated any interested back then.

We could keep feign 9.x active for java 6 compatibility.

@masc3d Are you interested in maintaining that branch?

[masc3d]

Removing features that have already been accepted feels like bad form to me.

@kdavisk6 I was referring to your initial suggestion Restrict Java 8 language usage to that of Android

Are you interested in maintaining that branch?

@velo no, but I may kind of have to )

alright, so I understand your decision is set, that feign-10 will move forward without android support being (very) relevant. I will stick with feign-9 and we'll see how it goes..

Im open to discuss this further, but you can also close this issue, I will leave it to you.