The state of Clojure on Android

The state of Clojure on Android(blog.ndk.io)

55 points by Kuytu 11 years ago | 17 comments

robinricard 11 years ago |

This is a nice, complete and well-documented review of why Clojure is not made for Android.

I disagree with the conclusion though: yeah, the classic JVM-clojure seems doomed on Android. Why ? It's not made for that, it's made for servers, it's not even used for desktop GUI programming that much. Even if we beat the startup time, it's not likely that we'll get nice UI-bindings for Android before a long time (I like the Java-interop but writing serious Android apps with the interop alone seems awful to me) and yes, there are some libraries abstracting this a little bit such as neko [1] but it's still too close to the Java-development paradigms (it's not as clojure-ish as I wish, but that's my opinon!).

So, the conclusion: no we don't need a swift for Android, and clojure is not swift anyway.

iOS already has clojure support, it's not really stable but it has decent startup times and a nice API. It's not the JVM-clojure, it's clojurescript. The UI is managed by React Native and om. It works and you get a nice REPL running too. [2] No it's not finished and has lots of bugs and it's not even publishable on the AppStore for now but it's just a matter of weeks here before addressing those problems (you can follow the work of David Nolen & Mark Fikes on this particularly, they often communicate on what they're doing but there are other awesome people working on this too!)

So for Android ? ReactNative is coming too [3]. When it's there, the community will try the same thing as what they did on iOS, and it's likely to succeed too. The community behind cljs is growing and it's becoming the main clojure implementation for ui-related stuff. You can thank React and om for that but I think that core.async has a lot to do with it too.

Anyway, I don't want to stop believing in Clojure for mobile development.

[1]: https://github.com/clojure-android/neko [2]: https://github.com/omcljs/ambly/wiki/Driving-React-Native-wi... [3]: https://facebook.github.io/react/blog/#when-is-react-native-...

zak_mc_kracken 11 years ago | |

> So, the conclusion: no we don't need a swift for Android,

I think we do, and to me, that language is Kotlin:

Analysis: http://goo.gl/wNjXoU

The language: http://kotlinlang.org

on_and_off 11 years ago | | |

Kotlin also looks like the most serious contender to me. Sure, something like Rust might be more interesting for various reasons, but the engineering effort needed to port the Android framework to another language would be huge (and the official word from the Android framework team is that it is not worth the effort). Not to mention that third party developers would have to relearn the framework from scratch and that we either have a windows mobile 6 / WP7 rift or the framework team has another nightmarish problem to solve. Kotlin looks like it could both guarantee retro-compatibility with Dalvik and brings a better language to the platform.

zserge 11 years ago | | |

Totally agree with you! Kotlin is great replacement for Java on Android. I like to see that more and more interesting libraries appear for Kotlin, like Anko. I recently made Anvil compatible with Kotlin (http://zserge.com/blog/anvil-kotlin.html), which makes UI development a pleasure

daniel_solano 11 years ago | |

Just a few thoughts:

1. Clojure is a general purpose programming language. Just because it isn't used much for desktop GUIs doesn't mean it's not usable for that. What percentage of Java/Scala development is used for GUIs?

2. Clojure in iOS isn't just available via ClojureScript. There is also RoboVM, though it doesn't support REPL-driven development, if I understand it correctly.

3. I am very interested in what React Native can do. For me, the key questions about using it are: Can you create applications that fit will into the UX idioms of the host platform? What happens when something goes wrong? Do the various levels of indirection make debugging much more difficult? Do you still have access to the full platform? If not, then only certain classes of applications can be implemented using React Native.

pjmlp 11 years ago | |

> So, the conclusion: no we don't need a swift for Android, and clojure is not swift anyway.

At very minimum we need Java 8, but I am willing to bet Google IO will not bring any news on that front.

Currently I have been using Java/C++ for my mobile hobby programming as I also like to play around with WP (C++/CX).

Regarding Clojure, maybe a static Lisp with Clojure syntax but with type inference could be a way to tackle the problem.

It wouldn't be pure Clojure, but maybe good enough.

th0ma5 11 years ago | | |

Maybe Pixie-Lang with the NDK. You would have to wrap all the Java GUI calls through their C interface. This sort of messes with my head a little.

j_m_b 11 years ago | |

I've been using Phonegap with Clojurescript on iOS. Perhaps that would be a viable alternative for Android development as well.

daniel_solano 11 years ago | | |

It is certainly doable. I haven't done it myself, but I know of people who have done it successfully on Android.

Unfortunately, I don't think Phonegap is a very good solution. Depending on your application, it may be good enough, but it's overall not as good as true native development.

on_and_off 11 years ago | | |

It is not. Not as a general purpose solution anyway. Phonegap can be useful in order to quickly publish a very simple application on both platforms. Outside of that, its disadvantages pile up VERY quickly.

th0br0 11 years ago |

From a user perspective, I find the statement that "Lean Clojure" works based off these charts a bit daring. Why?

a) Without having access to the numbers, (Skummet) startup is ~5 times as slow as the Java version. That is a noticeable difference. Plus, even Java apps sometimes tend to take a second or more to start up if they're reasonably complex.

b) The "Lean Clojure works" statement is derived from comparing _only_ the startup times. The fact that the task time is around two times Java's for Skummet is a big thing. Ignoring startup time, and without further data/tests, one could argue that Clojure Android apps run twice as slow as Java ones... that's a big no-go given the high amount of sparkly UI features of most modern apps. Or at least a no-go for having a Clojure-only stack?

(Skummet's probably in its early stages and I'd expect improvements but ... from those graph I'd argue 'Lean Clojure' does not work (yet?))

hoodunit 11 years ago | |

To clarify my statement: the idea of lean Clojure for making Clojure viable for Android development basically works. More static compilation of vars dramatically reduced startup times; I think further "lean Clojure" changes like dependency shaking would bring equally dramatic improvements. I'm handwaving a bit here at the challenges involved in making further changes, but the direction seems quite promising, at least for reducing startup times.

General performance is more difficult to speak for. If it is within a factor of two of Java that seems pretty similar to Clojure vs Java on the desktop and is a tradeoff many are willing to make. It's also true that Clojure relies heavily on Java libraries and Android app performance is highly dependent on underlying library and native performance. So basically for run time performance I don't think the performance of Clojure itself is the most significant factor.

cp9 11 years ago |

At clojure/west Zachary Kim had a talk about the state of mobile app dev in clojure [0].

[0]: https://www.youtube.com/watch?v=AgjRo54ImUMa

daniel_solano 11 years ago |

So, I've been involved on Clojure/Android for a long time. I really like what Alexander has done with Skummet, and I think there is still room for improvement. However, I have to wonder to what extent is it good enough?

I use a lot of applications which have terrible load times. Some of these may be native, others are probably using some sort of cross-platform development framework. While I would prefer better performance, I am coming to the conclusion that startup time really isn't an absolute deal-breaker for a lot of people.