This. And speed aside, Gradle embodies everything that is wrong with modern software development. A giant, opaque, idiosyncratic black box. I develop for Android (Java and Flutter) and it has become a sad pattern from Hell: Android Studio update mandates upgraded Gradle, which breaks previously the build in most weird ways, forcing one to trawl StackOverflow looking for proper incantations to do in the myriad of files it has created with the single purpose of building the damn app. Sometimes I wish, creating new build systems was outlawed some time after gmake or cmake...
Exactly, gradle is so slow that I ditch it into the ground. Just starting a blank project is so a bad experience. Nothing beat maven and a couple of shell script.
I am convinced if they hadn't got Google's sponsorship, when Android decided to move into Gradle during Eclipse to InteliJ migration, we would be talking about Gradle as much as Grails is still discussed at local JUGs.
OK let's say I have an online store I wrote in Spring Framework. What does my before and after development workflow look like after adopting your plugin? It's been a while since I've messed with Java, but how many seconds could it possibly need to fetch a bunch of jar files? I'd love to see an in-the-life-of kind of screencast of what the daily grind looks like for normal java devs these days, so I can understand why it's so slow.
For example, what does resolution mean? Does that mean fetching the pom.xml files from sonatype to figure out the dependency graph? Don't those HTTP requests normally go fast? Is Elide sort of like setting up an HTTP caching proxy between corp and sonatype?
> OK let's say I have an online store I wrote in Spring Framework. What does my before and after development workflow look like after adopting your plugin?
Basically, you install this plugin, and it tells the `JavaCompile` tasks normally used by Gradle to `isFork = true` and use `elide javac -- ...` for the compile command.
So, javac invocations travel over the CLI instead of through the Tooling API, which Gradle normally uses, and which is effectively running in Gradle's daemon VM, and thus is subject to JIT warmup.
Through Elide, since it's a native binary, all that JIT warmup is skipped, and you are effectively choosing to balance more toward quick startup and wall-clock time (with many small calls) instead of waiting for JIT warmup to make the Gradle daemon fast with javac.
Otherwise, it's a completely identical javac experience. Running `elide javac -help` produces identical help output. Identical inputs should produce identical outputs, and we build it at JDK 24 so it can support `--source/--target/--release X` and friends for anything older down to JDK 8.
> For example, what does resolution mean? Does that mean fetching the pom.xml files from sonatype to figure out the dependency graph?
It means resolving the graph from declared (direct) dependencies, downloading pom.xml metadata, downloading JARs, unpacking it all to disk, and providing a local `.m2` Maven-compliant root, sort of like Node does for `node_modules/`. Ours lives in `.dev/dependencies/m2` once you run `elide install`.
Elide embeds Maven's resolver, so resolution semantics are identical to Maven's. We do some small optimizations to make fetching fast (i.e. initializing related classes at build time, that sort of thing), but honestly not much. Just building Maven's resolver like this yields a major gain, and not messing with it preserves expected behavior.
Since Elide emits these deps in a Maven Local-style root, Gradle just finds them on disk when it needs them, so it doesn't need to engage its resolver or fetcher at all.
> Don't those HTTP requests normally go fast?
It is worth noting that Gradle seems confined to HTTP/1.1 and poor connection pooling even today, so it's not that hard to beat.
> Is Elide sort of like setting up an HTTP caching proxy between corp and sonatype?
We don't proxy or anything, this fetching still happens through normal Maven Central unless configured otherwise in your `elide.pkl` manifest. For now, deps are placed in the local project, but we want to move to a central cache and link like modern NPM installers do.
(Oh, if you are seeing how the plugin installs a Maven repository, that's just to ship our own plugin artifact and Gradle Catalog, for easy dependency use. maven.elide.dev doesn't proxy to central or anything like that.)
Hypothetically, if you could daemonize javac, would JIT eventually kick in over multiple recompiles of the same code? The obvious use case for this would be IDEs, but I imagine it could work in CI too if you had some kind of persistent "compiler as a service" setup that outlived the runners.
Not to detract from the cool work done here, just curious if this other approach has been tried too.
> Hypothetically, if you could daemonize javac, would JIT eventually kick in over multiple recompiles of the same code?
This is actually how Gradle works by default; in fact, even if you pass `--no-daemon`, it will start a daemon, and just kill it after the build (lol). It's daemon-first, daemon-only, because of this exact issue.
As I understand it, Gradle's daemon timeout is typically 10 minutes, but can be extended. We are guessing that this style rarely hits the 10k class threshold where JIT performance converges with native, especially since Gradle also supports incremental compilation, and that further impedes the progress toward a warm JIT. Gradle focuses hard on build caching and incremental compilation, and this is conceptually in contention with reaching a warm JIT, ironically.
Native Image has JIT capabilities (just not as mature as Hotspot yet), and keeping the daemon alive would probably still yield wins. We haven't tried it yet and that's a good idea
I'm aware of the Gradle daemon, but wasn't sure if it only handled dependency resolution and other build orchestration tasks or if it also ran the compiler in the same JVM instance. Last time I worked somewhere using Gradle I think we forked the compiler anyway to ensure the project was built on exactly the same Java version regardless of what the Gradle environment was running in, so in that case there is definitely room for startup optimization.
I do recall the Gradle daemon living much longer than 10 minutes, though. The docs say 3 hours is default, although if really trying to maximize the JIT advantage it perhaps would make sense to keep it alive as long as possible.
> or if it also ran the compiler in the same JVM instance
Honestly, Gradle's toolchain resolution mechanisms have changed so much, it is a little hard to keep track. Good point
> The docs say 3 hours is default
Huh, I'm not sure where I got 10 minutes. I'll research some more about these assumptions, but even so, having used Gradle professionally and personally for many years (like you), I wonder how often I would have hit that threshold. I was mostly working in smaller projects and companies, though, so my experience could be different from the norm. Even assuming engineers hit that threshold, it would be toward the latter end of that window and only until the end of that window. An experience optimized for cold-start and non-reflection balances this approach, and is even build-cacheable with standard tools like sccache. There are still a lot of projects where JIT-based javac would be optimal
> perhaps would make sense to keep it alive as long as possible
It would for sure. I'm not sure why Gradle was never able to execute on Bazel's full vision for remote execution. Probably a hermeticity problem, considering the challenges they already seem to face with e.g. the configuration cache.
I work professionally in Python these days, but I've just spent a bit of time digging into the contemporary situation because the JVM environment is much more interesting to me, and found this spec: https://build-server-protocol.github.io
Seems the Scala guys have doubled down on the "compiler as a service" approach, presumably because their compile time story continues to be painful. But also looks like the same solution is used for the VS Code Java/Gradle integration, so seems like this might be the more conventional way to go for traditional JVM projects.
For processes where the JIT takes a while to kick in, but also you don't want to waste memory keeping JVMs alive while not doing anything (and compilation could be a good example of that), I wondered if there was a way to snapshot and restore the JVM state and turns out some people are experimenting with that too: https://openjdk.org/projects/crac/
Elide can run Python using GraalPy, and we want to do some similar toolchain stuff in that universe. We are working on an integration with uv. Thanks also for linking this Build Server Protocol thing -- this looks very cool, very relevant, and I don't think I've seen this yet.
> Seems the Scala guys have doubled down on the "compiler as a service" approach, presumably because their compile time story continues to be painful
I wonder if we could run the Scala compiler. Reflection is a challenge in native mode and iirc Scala is reflection heavy, but I have no idea, I've never used it myself. I'll look into this
> snapshot and restore the JVM state and turns out some people are experimenting with that too
There are many efforts in this area: CrAC, Leyden, the new JDK 24 AOT stuff, and pre-warmup stuff before that. These all do make a difference but Native Image takes AOT optimization to a whole other level and seems to perform better for quick-startup quick-shutdown cases[1].
Elide also runs JVM code via Truffle/Espresso, and we're going to need a regular JVM integration. On this side CrAC and so on will be useful to cut down on startup time
For scala 2 at least, the standard library is very heavy and the compiler of course relies on it. It has a god object, predef, that cascades massive classloading if you touch anything in it. As a result, it is essential for latency to keep a warm jvm up to avoid reloading it.
Incremental rebuilds can actually be not too bad in scala when a warm jvm is used for compilation as it is by sbt and other scala build tools, but sbt also by default uses the warm jvm to run in process tests. This too can be reasonably quick but leads to problems if there are any resource leaks in the user's test suite.
Thus with sbt, one must either exercise discipline and care with tests to not leak anything or you will periodically have to restart sbt, which can be very painful because it is much heavier than the already heavy scala compiler.
It's a disappointing state of affairs because while there is a lot to like about scala, it is impossible to escape the bloat. Even if you opt out of sbt, the lack of modularity in the standard library forces workarounds, such as native-image, which had its own issues (including horrific build times for making native images and weak reflection support). Even if you or your org avoids the direct pitfalls, it is likely at some point you will end up debugging a dependency that doesn't.
I personally abandoned scala due to this bloat, which was a shame because I find it to have the best combo of expressiveness and pragmatism of any established general purpose languages.
In addition to being a runtime, Elide embeds `javac` and `kotlinc`. Here it is being used as a build toolchain, but it can also run those languages, through GraalVM/Truffle.
There are some JARs we must commit to source because they are built with patches from upstream GraalVM components. That's probably the big weight there.
We have an issue in our repo tracking those upstream PRs. Once merged, we won't need to do this anymore. We should probably publish them in our own Maven repo to avoid this kind of experience. Thanks for the push to take care of that cruft :P
You can find them in `third_party/oracle`. You also don't need to build from source -- Elide ships as a binary for Linux amd64/macOS arm64. You are totally welcome to though, and if you encounter issues, hop in our discord and we'd be glad to help: https://elide.dev/discord
(In other news, I've added Kotlin to the list there, as {...|kt|kts}, so thank you for reporting this as well! We really need some docs updates and I'll make sure to clarify this stuff)
Would this result in the same .class that javac produces? I.e. at runtime is code compiled with this fundamentally different than code compiled with javac?
This results in identical .class bytecode that `javac` produces, yes, because it's literally just `javac`, but built as a native binary instead of a Java entrypoint.
We build at JDK24 and support `--source/--target/--release` so you can build against earlier JVM targets.
If you find any output differences between Elide and JDK24 (Oracle GraalVM), then we'd consider that a bug. Similarly, we accept identical compile flags and emit identical warnings/messages, as under the hood we are just using the same compiler APIs but natively.
Elide's resolver is just Maven's, and we use Maven's own components to read poms. Native-imaging Gradle itself is really challenging because Gradle relies so much on reflection, and really is architected to account for the JIT case (i.e. emphases on the daemon).
I haven't found a way to do it yet so this plugin is the best alternative so far.
The underlying Elide tool looks amazing, and potentially solves all sorts of pain for me. Does the Kotlin compiler support plugins, e.g. serialisation? I can imagine they might be too reflective.
> Does the Kotlin compiler support plugins, e.g. serialization?
We don't support Kotlin Serialization yet, but KotlinX is fully embedded and ready. You can reach for things like KotlinX Coroutines without installing a single JAR.
Bring-and-use-your-own-plugins might be challenging, yeah, but we plan to at least build in the KotlinX plugins you would want by default (serialization included). So thank you for this note, it's helpful and I'll add a notch next to that feature :)
Benchmarking a toy case (hello world) shows a 20x performance difference; this is true for both kotlinc and javac, when benchmarked through hyperfine. YMMV, but benchmarks of different build styles show native converging with JIT at about 10k classes[1]
So, if your project is 10k classes or less, you are looking at "up to" a 20x improvement in compile time vs. stock javac.
On the dependency fetching side, we haven't benchmarked yet. It definitely feels like 100x+ though on a wall-clock basis.
Personally the solution to faster build times in Gradle, is to keep using Maven.
I never touch Gradle unless there is no way around it, like Android.
However, this looks like an interesting idea.
This. And speed aside, Gradle embodies everything that is wrong with modern software development. A giant, opaque, idiosyncratic black box. I develop for Android (Java and Flutter) and it has become a sad pattern from Hell: Android Studio update mandates upgraded Gradle, which breaks previously the build in most weird ways, forcing one to trawl StackOverflow looking for proper incantations to do in the myriad of files it has created with the single purpose of building the damn app. Sometimes I wish, creating new build systems was outlawed some time after gmake or cmake...
Exactly, gradle is so slow that I ditch it into the ground. Just starting a blank project is so a bad experience. Nothing beat maven and a couple of shell script.
I am convinced if they hadn't got Google's sponsorship, when Android decided to move into Gradle during Eclipse to InteliJ migration, we would be talking about Gradle as much as Grails is still discussed at local JUGs.
Same, but also for Kotlin
Agreed.
OK let's say I have an online store I wrote in Spring Framework. What does my before and after development workflow look like after adopting your plugin? It's been a while since I've messed with Java, but how many seconds could it possibly need to fetch a bunch of jar files? I'd love to see an in-the-life-of kind of screencast of what the daily grind looks like for normal java devs these days, so I can understand why it's so slow.
For example, what does resolution mean? Does that mean fetching the pom.xml files from sonatype to figure out the dependency graph? Don't those HTTP requests normally go fast? Is Elide sort of like setting up an HTTP caching proxy between corp and sonatype?
> OK let's say I have an online store I wrote in Spring Framework. What does my before and after development workflow look like after adopting your plugin?
Basically, you install this plugin, and it tells the `JavaCompile` tasks normally used by Gradle to `isFork = true` and use `elide javac -- ...` for the compile command.
So, javac invocations travel over the CLI instead of through the Tooling API, which Gradle normally uses, and which is effectively running in Gradle's daemon VM, and thus is subject to JIT warmup.
Through Elide, since it's a native binary, all that JIT warmup is skipped, and you are effectively choosing to balance more toward quick startup and wall-clock time (with many small calls) instead of waiting for JIT warmup to make the Gradle daemon fast with javac.
Otherwise, it's a completely identical javac experience. Running `elide javac -help` produces identical help output. Identical inputs should produce identical outputs, and we build it at JDK 24 so it can support `--source/--target/--release X` and friends for anything older down to JDK 8.
> For example, what does resolution mean? Does that mean fetching the pom.xml files from sonatype to figure out the dependency graph?
It means resolving the graph from declared (direct) dependencies, downloading pom.xml metadata, downloading JARs, unpacking it all to disk, and providing a local `.m2` Maven-compliant root, sort of like Node does for `node_modules/`. Ours lives in `.dev/dependencies/m2` once you run `elide install`.
Elide embeds Maven's resolver, so resolution semantics are identical to Maven's. We do some small optimizations to make fetching fast (i.e. initializing related classes at build time, that sort of thing), but honestly not much. Just building Maven's resolver like this yields a major gain, and not messing with it preserves expected behavior.
Since Elide emits these deps in a Maven Local-style root, Gradle just finds them on disk when it needs them, so it doesn't need to engage its resolver or fetcher at all.
> Don't those HTTP requests normally go fast?
It is worth noting that Gradle seems confined to HTTP/1.1 and poor connection pooling even today, so it's not that hard to beat.
> Is Elide sort of like setting up an HTTP caching proxy between corp and sonatype?
We don't proxy or anything, this fetching still happens through normal Maven Central unless configured otherwise in your `elide.pkl` manifest. For now, deps are placed in the local project, but we want to move to a central cache and link like modern NPM installers do.
(Oh, if you are seeing how the plugin installs a Maven repository, that's just to ship our own plugin artifact and Gradle Catalog, for easy dependency use. maven.elide.dev doesn't proxy to central or anything like that.)
Hypothetically, if you could daemonize javac, would JIT eventually kick in over multiple recompiles of the same code? The obvious use case for this would be IDEs, but I imagine it could work in CI too if you had some kind of persistent "compiler as a service" setup that outlived the runners.
Not to detract from the cool work done here, just curious if this other approach has been tried too.
> Hypothetically, if you could daemonize javac, would JIT eventually kick in over multiple recompiles of the same code?
This is actually how Gradle works by default; in fact, even if you pass `--no-daemon`, it will start a daemon, and just kill it after the build (lol). It's daemon-first, daemon-only, because of this exact issue.
As I understand it, Gradle's daemon timeout is typically 10 minutes, but can be extended. We are guessing that this style rarely hits the 10k class threshold where JIT performance converges with native, especially since Gradle also supports incremental compilation, and that further impedes the progress toward a warm JIT. Gradle focuses hard on build caching and incremental compilation, and this is conceptually in contention with reaching a warm JIT, ironically.
Native Image has JIT capabilities (just not as mature as Hotspot yet), and keeping the daemon alive would probably still yield wins. We haven't tried it yet and that's a good idea
I'm aware of the Gradle daemon, but wasn't sure if it only handled dependency resolution and other build orchestration tasks or if it also ran the compiler in the same JVM instance. Last time I worked somewhere using Gradle I think we forked the compiler anyway to ensure the project was built on exactly the same Java version regardless of what the Gradle environment was running in, so in that case there is definitely room for startup optimization.
I do recall the Gradle daemon living much longer than 10 minutes, though. The docs say 3 hours is default, although if really trying to maximize the JIT advantage it perhaps would make sense to keep it alive as long as possible.
> or if it also ran the compiler in the same JVM instance
Honestly, Gradle's toolchain resolution mechanisms have changed so much, it is a little hard to keep track. Good point
> The docs say 3 hours is default
Huh, I'm not sure where I got 10 minutes. I'll research some more about these assumptions, but even so, having used Gradle professionally and personally for many years (like you), I wonder how often I would have hit that threshold. I was mostly working in smaller projects and companies, though, so my experience could be different from the norm. Even assuming engineers hit that threshold, it would be toward the latter end of that window and only until the end of that window. An experience optimized for cold-start and non-reflection balances this approach, and is even build-cacheable with standard tools like sccache. There are still a lot of projects where JIT-based javac would be optimal
> perhaps would make sense to keep it alive as long as possible
It would for sure. I'm not sure why Gradle was never able to execute on Bazel's full vision for remote execution. Probably a hermeticity problem, considering the challenges they already seem to face with e.g. the configuration cache.
In any case, this is great feedback, thank you :)
I work professionally in Python these days, but I've just spent a bit of time digging into the contemporary situation because the JVM environment is much more interesting to me, and found this spec: https://build-server-protocol.github.io
Seems the Scala guys have doubled down on the "compiler as a service" approach, presumably because their compile time story continues to be painful. But also looks like the same solution is used for the VS Code Java/Gradle integration, so seems like this might be the more conventional way to go for traditional JVM projects.
For processes where the JIT takes a while to kick in, but also you don't want to waste memory keeping JVMs alive while not doing anything (and compilation could be a good example of that), I wondered if there was a way to snapshot and restore the JVM state and turns out some people are experimenting with that too: https://openjdk.org/projects/crac/
It's all neat stuff!
> I work professionally in Python these days
Elide can run Python using GraalPy, and we want to do some similar toolchain stuff in that universe. We are working on an integration with uv. Thanks also for linking this Build Server Protocol thing -- this looks very cool, very relevant, and I don't think I've seen this yet.
> Seems the Scala guys have doubled down on the "compiler as a service" approach, presumably because their compile time story continues to be painful
I wonder if we could run the Scala compiler. Reflection is a challenge in native mode and iirc Scala is reflection heavy, but I have no idea, I've never used it myself. I'll look into this
> snapshot and restore the JVM state and turns out some people are experimenting with that too
There are many efforts in this area: CrAC, Leyden, the new JDK 24 AOT stuff, and pre-warmup stuff before that. These all do make a difference but Native Image takes AOT optimization to a whole other level and seems to perform better for quick-startup quick-shutdown cases[1].
Elide also runs JVM code via Truffle/Espresso, and we're going to need a regular JVM integration. On this side CrAC and so on will be useful to cut down on startup time
[1]: https://github.com/simonis/LeydenVsGraalNative
For scala 2 at least, the standard library is very heavy and the compiler of course relies on it. It has a god object, predef, that cascades massive classloading if you touch anything in it. As a result, it is essential for latency to keep a warm jvm up to avoid reloading it.
Incremental rebuilds can actually be not too bad in scala when a warm jvm is used for compilation as it is by sbt and other scala build tools, but sbt also by default uses the warm jvm to run in process tests. This too can be reasonably quick but leads to problems if there are any resource leaks in the user's test suite.
Thus with sbt, one must either exercise discipline and care with tests to not leak anything or you will periodically have to restart sbt, which can be very painful because it is much heavier than the already heavy scala compiler.
It's a disappointing state of affairs because while there is a lot to like about scala, it is impossible to escape the bloat. Even if you opt out of sbt, the lack of modularity in the standard library forces workarounds, such as native-image, which had its own issues (including horrific build times for making native images and weak reflection support). Even if you or your org avoids the direct pitfalls, it is likely at some point you will end up debugging a dependency that doesn't.
I personally abandoned scala due to this bloat, which was a shame because I find it to have the best combo of expressiveness and pragmatism of any established general purpose languages.
Volker Simonis over at Amazon did some very cool benchmarks showing the impact this can have (i.e. native-imaging javac):
https://github.com/simonis/LeydenVsGraalNative
So elide is a tool to run {js|ts|py} [1]. How does it compile javac to native?
[1] https://github.com/elide-dev/elide
In addition to being a runtime, Elide embeds `javac` and `kotlinc`. Here it is being used as a build toolchain, but it can also run those languages, through GraalVM/Truffle.
What.is.happening.here. Are they committing node_modules or something?
There are some JARs we must commit to source because they are built with patches from upstream GraalVM components. That's probably the big weight there.
We have an issue in our repo tracking those upstream PRs. Once merged, we won't need to do this anymore. We should probably publish them in our own Maven repo to avoid this kind of experience. Thanks for the push to take care of that cruft :P
You can find them in `third_party/oracle`. You also don't need to build from source -- Elide ships as a binary for Linux amd64/macOS arm64. You are totally welcome to though, and if you encounter issues, hop in our discord and we'd be glad to help: https://elide.dev/discord
(In other news, I've added Kotlin to the list there, as {...|kt|kts}, so thank you for reporting this as well! We really need some docs updates and I'll make sure to clarify this stuff)
Would this result in the same .class that javac produces? I.e. at runtime is code compiled with this fundamentally different than code compiled with javac?
This results in identical .class bytecode that `javac` produces, yes, because it's literally just `javac`, but built as a native binary instead of a Java entrypoint.
We build at JDK24 and support `--source/--target/--release` so you can build against earlier JVM targets.
If you find any output differences between Elide and JDK24 (Oracle GraalVM), then we'd consider that a bug. Similarly, we accept identical compile flags and emit identical warnings/messages, as under the hood we are just using the same compiler APIs but natively.
Interesting idea. I wonder how this would compare to native-imaging Gradle or Maven itself.
Elide's resolver is just Maven's, and we use Maven's own components to read poms. Native-imaging Gradle itself is really challenging because Gradle relies so much on reflection, and really is architected to account for the JIT case (i.e. emphases on the daemon).
I haven't found a way to do it yet so this plugin is the best alternative so far.
The underlying Elide tool looks amazing, and potentially solves all sorts of pain for me. Does the Kotlin compiler support plugins, e.g. serialisation? I can imagine they might be too reflective.
> The underlying Elide tool looks amazing
Thank you!
> Does the Kotlin compiler support plugins, e.g. serialization?
We don't support Kotlin Serialization yet, but KotlinX is fully embedded and ready. You can reach for things like KotlinX Coroutines without installing a single JAR.
Bring-and-use-your-own-plugins might be challenging, yeah, but we plan to at least build in the KotlinX plugins you would want by default (serialization included). So thank you for this note, it's helpful and I'll add a notch next to that feature :)
Is there an equivalent maven plugin?
Also what's the licence for elide?
Not yet, but we are working on a Maven plugin :) if you join our Discord or file an issue, I can update you when its out.
Elide is licensed under MIT https://github.com/elide-dev/elide
interesting. I wonder how compares against mavend (mvnd , aka maven as a Daemon service) and his aggressive parallelization.
Do you have benchmarks? How much of a difference does it make?
Benchmarking a toy case (hello world) shows a 20x performance difference; this is true for both kotlinc and javac, when benchmarked through hyperfine. YMMV, but benchmarks of different build styles show native converging with JIT at about 10k classes[1]
So, if your project is 10k classes or less, you are looking at "up to" a 20x improvement in compile time vs. stock javac.
On the dependency fetching side, we haven't benchmarked yet. It definitely feels like 100x+ though on a wall-clock basis.
[1]: https://github.com/simonis/LeydenVsGraalNative