Deno 2 with Ryan Dahl

Рет қаралды 20,436

Күн бұрын

Пікірлер: 68

@Skylla54 2 ай бұрын

I was not aware of deno jupyter. I was just looking for something like that a week a go :D 2 minutes later, and it runs and works, sweet :D

@kirillvoloshin2065 2 ай бұрын

loved my experience with deno! especially not having to browse stackoverflow for every error or config mismatch

@FrankHale 2 ай бұрын

This was such a good interview. I love Deno!

@JeffHykin 2 ай бұрын

Ryan nailed the part about notebooks. I do AI research (5th year PhD) everything is in python and Jupyter Notebooks are such a pain because of the external env. Matplotlib is trash and also the standard, and everything is unreliable. I already use deno for all my scripts, but sadly there's still a long way to go before machine learning in JS is practical. I love to see the progress though! I'll be the first on the ML-in-deno train as soon as its practical! (Aka pytorch)

@JeffHykin 2 ай бұрын

Dang :/ that isn't great. I really wish they allowed it for proxy objects.

@GolderiQ 2 ай бұрын

@@RustIsWinningwho they?

@Danielo515 2 ай бұрын

It is not that big of a deal when you can just have functions

@kinsondigital Ай бұрын

I have been using deno now for EVERYTHING js/ts for the past 2 or so years and I am in love with it. I will avoid node at all costs. But now with deno 2.0, avoiding node is that much easier. I think deno and JSR is the future for sure.

@emilbonnek 2 ай бұрын

I didnt know Deno matches Bun performance, but that actually makes me a lot more excited about Deno.

@danvilela 2 ай бұрын

zig vs rust

@pietraderdetective8953 2 ай бұрын

Are there some benchmark numbers we can see? I know people are saying bench numbers are BS..though it's usually said by the losing party / slower bench numbers.

@offroaders123 2 ай бұрын

This was outstanding, so glad to hear all the things you covered here!

@SDAravind 2 ай бұрын

Im planning to using deno2 with svelte5 and sveltekit. What are your thoughts? Both are in RC

@eddie3800 2 ай бұрын

I know it's very easy to say, but it seems to me that, simplifying it a lot, for one Javascript runtime to surpass another, it has to go through guaranteeing backward compatibility and adding features that surpass the previous one, but starting from what there is. If you create a parallel project, even if it has objectively better functionalities, as happened with Deno at the beginning, I think it is more complicated to make the leap. The difficult thing is knowing when to make the cut to be backward compatible, but I think it has to be gradual based on what was there before. I think they have identified this well from the beginning in bun js and starting from what already existed, they try to build something that integrates it and surpasses it. I think it would be something similar to what happens with philosophical systems, for one to be more powerful than the other it has to be able to wrap it around and provide, let's say, more powerful explanations than the philosophical system that one wants to overcome or in the sciences when the theory of Enstein's relativity integrates Newton's theory of gravitation and provides a more powerful explanation. That is, bun js, if I'm not mistaken, wants to be able to run a large part of the node.js ecosystem within bun.js and once the entire node.js ecosystem is sufficiently wrapped in bun.js, overcome it with the objective advantages it provides. That is to say, the new runtime has to start from the present in progress to have more possibilities, with the difficulty that this entails. I don't understand programming languages like Rust or Zig, but from what I understand, and from what Sumner said, he started the bun.js project with Rust and saw that it was more productive with Zig and decided to move to Zig. Additionally, Zig, if I understand correctly, makes it very easy to compile C and C++ code, which is what Node.js is written in. So I think, in part, it has more advantage over Deno and Rust.

@eddie3800 2 ай бұрын

@@RustIsWinning Thanks for your comment. I was referring more to how Deno initially approached it. I agree that probably the best option is to start a project from scratch, but I was more referring to people deciding to take the leap. It is clear that they are two different projects and from what I have heard from Dahl on other occasions, he created Deno to adapt to the things that had advanced and to correct some decisions he had made when creating Node. From certain reports from last year it seems that at the moment, Deno js is used more than bun js. I was referring rather to the strategy carried out from the beginning by bun to ensure that people implement it in real, day-to-day life. I don't know if it makes sense to compare Rust and Zig, but I was referring to what Sumner had said, that with Zig he advanced much faster, I don't know if that's the case in general or if it's just his thing. I also don't know if, even if you advance faster, that will lead to other problems in the long run for not using Rust, but from what I understand, there is no compelling reason to believe that.

@eddie3800 2 ай бұрын

@@RustIsWinning I agree with what you say. You know better what you're talking about. I believe you if you tell me that they are causing serious long-term problems by using Zig instead of Rust or that they are simply making questionable decisions as you say. As far as I could understand, I thought that there were no serious problems when deciding to use Zig instead of Rust. What I wanted to say is that it seemed to me that with the bun js strategy they had better understood from the beginning how Node js had to be overcome. Also, it must be said that they have been able to see the problems of Deno js to transfer users from Node and learn from that. I think what you say about PHP goes in the sense that I mentioned, that even though PHP is slower, people continue using it. That's why it's not just the fact that it's better on paper, but you have to know how to get people to use it, among other factors. I argued in favor of Zig because I didn't know that it could cause problems in the long run, I was simply saying what Sumner had commented, that he was advancing faster, I didn't know that there were problems with Zig. Exactly what serious problems are there with Zig?

@eddie3800 2 ай бұрын

@@RustIsWinning Yes, I have heard Sumner himself say that they have to make it more stable. It is true that Zig is not yet in version 1.0, I don't know if that is a significant problem.

@itsthesteve 2 ай бұрын

I gotta say after years of listening to the podcast, it's like meeting a blind date seeing y'all on video.

@cotyhamilton 2 ай бұрын

Omg, best crossover episode

@cotyhamilton 2 ай бұрын

Laughing at ever having to use Python 😂 same experience

@cotyhamilton 2 ай бұрын

“Matplotlib is absolute trash” 😂

@StingSting844 2 ай бұрын

Why isn't deno supporting something like go channels for concurrency? Is it impossible to use v8 isolates or something else for this?

@thejaredwilcurt 2 ай бұрын

Eventually Shadow Realms (or something equivalent) will be added into JavaScript and once that makes it's way into V8, Node will have access to it. There are already many libraries banking on this approach to executing isolated parallel JavaScript. For example, once available, Vitest should be able to get a massive performance boost. Of course nothing actually works in Deno, so this won't be of any help to it. And Bun is not based on V8, so it will need to wait for Apple to drag their feet on adopting Shadow Realms, which will likely delay Bun's ability to use this feature for years.

@trejohnson7677 2 ай бұрын

@@thejaredwilcurthow honestly did u read the explainer 4 shadowrealms and think *this is very much like a go channel*. or r u just yapping?

@MommysGoodPuppy 2 ай бұрын

webworkers work great, i think node workers are supported too

@seanknowles9985 2 ай бұрын

Workspaces doesn't work at the moment when trying to deploy to Deno deploy. Very frustrating.

@wKich 2 ай бұрын

Yep, that's right. And when you decide to build npm package of your project for node users from deno monorepo you'll get broken npm packages, with all glossiness, but broken :)

@brunodearaujoalves698 19 күн бұрын

What's the lib that Ryan Dahl recommends on 38:00?

@syntaxfm 19 күн бұрын

It is called observable plot: observablehq.com/plot/ www.npmjs.com/package/@observablehq/plot You can run it inside of Jupyter notebooks with Deno: docs.deno.com/runtime/reference/cli/jupyter/

@sanderd17 2 ай бұрын

The thing that irks me about typescript up to now, is that it's a script. It should be able to run without being compiled (or compile on the fly). Deno does support running ts, but the browser definitely needs it soon.

@alikin 2 ай бұрын

Thank you for doing this, Ryan is … Thank you Ryan!

@tech-daddy Ай бұрын

Now we need a guide to make this work with Remix and recommendation of how to really handle the dependency jungle.

@nathantoups 2 ай бұрын

Amazing episode. Deno is so cool. 7 years already!?

@andrewzuo86 2 ай бұрын

"We're not working in Javascript because it's a beautiful language. We're working in it because it's very familiar." 😆

@BboyKeny 2 ай бұрын

Makes me wonder if the majority of js devs are monolingual.

@deleted-u5g 2 ай бұрын

@@BboyKeny javascript sucks in its syntax and everything, not to mention type checking which is non-existant.

@mohamedyamani8502 2 ай бұрын

nooo, no chapters :(

@syntaxfm 2 ай бұрын

Fixed.

@mohamedyamani8502 2 ай бұрын

@@syntaxfm NOICE!

@SDAravind 2 ай бұрын

42:53 will deno have library like pydantic?

@dzmitryluhauskoi6363 Ай бұрын

Zod and TypeBox should work just fine.

@feldinho 2 ай бұрын

The real missing story in JS for AI in my opinion is Pandas. I wouldn't say Pandas is easy but it's the best tool out there to wrangle the data so it can be used for training. Dedicated languages like R are better at it, but the fact that Pandas is in the same language and ecosystem as every other AI tool out there makes it pretty much unbeatable. (for now)

@nikilk 2 ай бұрын

The problem when you say Deno is for the noobs starting out is you loose the ones that know stuff don’t you think ? Second question: What if I wanna use Biome instead of Prettier ? Someone once told me never pick a piece of software because it’s easy to use, but instead pick one that makes the overall system your building easy to build. Easy to use tooling many times end up becoming harder to build systems, because it abstracts away too much you lose control.

@megamind452 2 ай бұрын

Please we need monorepo support of deno with next, angular and other frameworks. Maybe it has support for nextjs if we run nextjs with deno. But still it will be a game changer. I have always loved deno the dinosaur ❤.

@siya.abc123 2 ай бұрын

Before I use Deno, I'll try Go. That means I won't use Dental

@_vicary 2 ай бұрын

Fresh 2.0 when?

@khashayarjamsahar8510 2 ай бұрын

perfect

@erikslorenz 2 ай бұрын

The problem with Deno is that when I want better node, I just use go.

@belkocik 2 ай бұрын

@JeffHykin 2 ай бұрын

I don't think I'll be making interactive plots with Go in a Jupyter notebook anytime soon

@crosfitjesus 2 ай бұрын

Real

@itsthesteve 2 ай бұрын

Doesn't really track 1:1. If there's something that you're used to (node) and want something better, use Deno/bun. This analogy is like saying "if you want better sandwiches, make your own bread". Yes, it's not "hard", but outside the comfort zone of "take bread out of bag and make sandwich". I've switched to go for some apis I've made. It's a fine language. But not a blanket statement in my opinion.

@reze_dev 2 ай бұрын

@RustIsWinning Go is enough. Just saw fiber vs actix benchmark in kubernetes and fiber is better

@cdtwebdev 2 ай бұрын

I think Ryan is great and incredibly talented and visionary, but contrary to popular belief, I do want to write Rust or C++ performance applications using wasm to run on the browser.

@wKich 2 ай бұрын

Ha, you invented a new tool with all your previous experience to avoid many mistakes you had made in the past and... gotcha... you didn't add compatibility level to existing environment and all your potential newcomers can't use your glossy duplo-thing, because many serious projects simply don't work in it. And then you decided spent a few years to add node support to your tool. Genius...