Node 5.7.0 (stable) released | Hacker News Reader

Node 5.7.0 (stable) released

nodejs.org

161 points56k46 comments

One of the performance improvements [0] is pretty crazy! Up to an 18,000% performance increase.

[0] https://github.com/nodejs/node/pull/5123

Great news! Node.js could greatly benefit from more reimplementations of its native modules. I had implemented the `fs.readdir` module in C and saw some good performance gains (it was a hackish experiment I did some time ago and can't remember numbers).

The `http` module is said to be outperformed by `eshttp` [1] by a factor of 2 times, which is written in pure JavaScript.

There is also `node-lwan` [2], an experiment in using `lwan` [3] in place of the built in `http` module. Since its the most performant web server in JSON serialization, according to techempower.com benchmarks [4], performance gains are to be expected.

[1] https://github.com/iefserge/eshttp [2] https://github.com/raadad/node-lwan [3] http://lwan.ws/ [4] https://www.techempower.com/benchmarks/#section=data-r12&hw=...

Yes, but it breaks some cases. I'd recommend to wait 5.7.1 (https://github.com/nodejs/node/issues/5400)

Probably an ignorant question but this makes me wonder: how can they release when their tests don't pass?

I haven't been following, but it looks like the tests were passing and they just didn't have tests to prevent these issues that came up.

Break early, break often :-).

Maybe psychology...

People won't use x.y.0 so they build in a bug to push to x.y.1 afterwards

This is exactly why I've stayed away from this stack wherever possible!

I hope you stay away from GCC as well, which for example broke C++ in 4.7.0 and fixed it in 4.7.2.

Any new major release is bound to break something for someone. Prudence suggests adopting a major release only after a few months. This applies to all platforms.

You're right; I was a bit harsh. I'm just really conservative platform-wise - Node is just starting to get there.

The Node/npm ecosystem does still feel a bit wobbly as an application platform, though, kind of like Ruby was five years ago...

This is not true. Some platforms, such as Clojure, are usually very stable once released, and there is rarely the need for point releases, as the code was stabilized during the alpha/rc periods.

I am happy that there is Node.

I'm still mind blown every single day about the Node community—for me it's the fastest evolving dev ecosystem which is at the same time high performant and robust.

It is amazing isn't it? I'm so happy I switched. I love this community and the things they are doing.

Still waiting for async/await. Also doing multi-precision numbers by default would be nice.

I started to use async/await because react-native and now I can't live without it. In nodejs you can have it:

npm install --save babel-core babel-loader babel-polyfill babel-preset-es2015 babel-preset-stage-0

Adding this to entry point file

  require("babel-core/register");
  require("babel-polyfill");

and having a .babelrc with these presets:

  {
    "presets": ["es2015", "stage-0"]
  }

Then you can transpile with babel-cli ( sudo npm install -g babel-cli if you don't have it yet)

  babel app.js -o transpiled_app.js

Make sure you have npm version 3, or it become too slow to initialise.

as @vdaniuk mention Chackra supports it, but did anyone tried this https://github.com/nodejs/node-chakracore?

Is there any other way to get async/await today ?

Worth considering the es2015-node5 preset instead. It will only transpile the features that are not already in node 5.x. The standard es2015 preset transpiles lots of APIs that are already natively available in node 5.

coroutines with generators are more powerful (cancellation, automatic resource cleanup when the generator terminates etc) and have worked for a long time out of the box. The only difference is that you write `yield` instead of `await` and `decorator(function*` instead of `async function`

Babel can transform async functions into Bluebird coroutines, so the best of both worlds - https://www.npmjs.com/package/babel-plugin-transform-async-t...

Yes but with this you lose the powerful features I just mentioned: automatic resource cleanup[1], cancellation, yield handlers.. don't you?

1. https://github.com/petkaantonov/bluebird/issues/1014

No I think those should work with async functions as it is literally just translating `async` -> `yield`, but worth investigating.

Thanks for commenting on that, I didn't think about those issues. I'll research on how much impact could have on performance. In the other hand I believe syntax is not the only difference. I'm not full time js dev, but after a quick search, I found Domenic pointing 4 differences here https://esdiscuss.org/topic/does-async-await-solve-a-real-pr...

These seem to be mostly about undecorated generators. 3. Is a legit difference but it is actually a syntax error to do yield 1 + yield 2. So that is fixable in a backward compatible way.

Could you put a reference for the decorator/yield usage? Thanks

http://bluebirdjs.com/docs/api/promise.coroutine.html

Thanks, I will review it.

https://github.com/llambda/koa-boiler/blob/master/app.js#L75 is an example of Bluebird coroutines paired with Koa.

async/await is not in the ECMA spec. It a proposal, like Object.observe(which was discarded) And remember one thing, Nodejs maintainers DO NOT control V8, the javascript engine powering it, they are entirely dependent on what the V8 team at Google does.

So you'll have to wait a long time and even so there is absolutely no guarantee async/away will ever make it to the spec. Same deal for decorators, developers shouldn't depend on non ECMA features. Even worse, the final proposal might be totally different from the current Babel/whatever implementation.

V8 doesn't even implement 100% of ES6, yet.

Stage 4 requires two working implementations. V8 could be the second one. No feature can reach stage 4 if it would be this strict as you describe.

Microsoft's Chakra engine is available as an alternative backend for node (windows only at the moment) and has native support of async/await.

https://github.com/nodejs/node-chakracore

Does it still only run on Windows?

AFAIK linux and mac release is coming later this year and even then it won't have JIT support.

For now. The Linux branch of ChakraCore is under active development.

Disclaimer: working at Msft on ChakraCore

Yeah this is a long way off and lots of things need to happen before node can even consider this… 1. TC39 needs to get async functions to stage 4: https://github.com/tc39/ecma262/blob/master/README.md 2. V8 needs to implement: https://github.com/nodejs/promises/issues/4 3. Node then needs to use this updated V8, and to be able to use this reliably, this V8 will need to reach an LTS release: https://github.com/nodejs/LTS/blob/master/schedule.png 4. Also Node API needs to become promises-compatible, which is god-knows how long away: https://github.com/nodejs/promises

I would love an async/await feature. Though generators have been added, and they are currently optimization killers in V8: https://github.com/petkaantonov/bluebird/wiki/Optimization-k...

That's always the case for new language features. ES5 stuff was slow in the beginning and it took a while for v8 to catch up. This shouldn't be a reason for not using these new features.

This article is also years old. While it seems to be updated recently I wonder how accurate it still is.

async/await isn't part of any standard yet. It's a stage 3 proposal at this time, which means it may not even land in ES 2016: https://github.com/tc39/ecma262

The last thing we want is native support for pre-standard features. As long as it's not in a spec (or draft at least), it's sci-fi.

Besides, node doesn't implement language features, V8 does. Here's the relevant issue: https://bugs.chromium.org/p/v8/issues/detail?id=4483

ES2016 is already finished, it only has array.includes and the exponentiation operator[0].

We'll have to wait for ES2017 at the earliest for async to arrive.

[0]http://www.2ality.com/2016/01/ecmascript-2016.html

Thanks for pointing that out. I haven't been following the releases that closely. But it's good to see that after the kitchen sink nature of ES2015 the yearly releases following it are a lot more incremental than revolutionary.

Multi precision numbers are going to be a long wait

It will probably take years for async/await to become part of Node.

If you are eager to start using it look at Typescript 1.7 - which transpiles to ES6. I have started trying it out on a side project.

> child_process: spawn() and spawnSync() now support a 'shell' option to allow for optional execution of the given command inside a shell. If set to true, cmd.exe will be used on Windows and /bin/sh elsewhere. A path to a custom shell can also be passed to override these defaults. On Windows, this option allows .bat. and .cmd files to be executed with spawn() and spawnSync(). (Colin Ihrig) #4598

Does this mean that it can now open a real shell (like iTerm) and execute a script ? or I misunderstood ?

Sadly not. When calling spawn[Sync]() in prior versions the command would be executed directly as a child of the node process. If you wanted to use any features like piping, automatic environment variable interpolation or backgrounding etc; you had to manually wrap your command with a system shell like bash(sh).

This addition when set to true, will do the wrapping for you in a cross-platform manor.

Thanks for the clarification

> Does this mean that it can now open a real shell (like iTerm)

iTerm is not a shell. bash, zsh, fish, etc are shells. iTerm is a terminal program that opens a shell.

This actually works even today:

child_process.exec('open -a Terminal.app')

I'm pretty sure you can pass a script arg to the application too, but I haven't done it in a while.

Reply on news.ycombinator.com