Erlang for boomers, Elixir for millennials, and Gleam for Gen Z? Iunno aboot that. I've never seen those three groups be this friendly to each other. :P
So stoked about Gleam: - Simple like go (small surface area) - Can leverage Elixir and Erlang tools - Scalable and fault tolerant (BEAM) - Type-safe and functional - Familiar to rust users (option/result and pattern matching) - `use` as a solution to callback hell (instead of async/await function coloring)
I used elm once in a production app. Absolutely loved it. Someone came in 3 years later and replaced it because they could not be bothered to learn it.
I have tried to stop me for quite a while but today is the day i can't anymore.. Please stop making these faces on your thumbnails, they never match how you actually respond to the video.
yeah it's this, and as an elixir guy idk if i can manage. i *love* overloading arguments like that, i fucking *love* my left justification, i *hate* nesting
@@lpil down the road - allow the user to decide - method overloading is super convenient - if they want inference in their callbacks etc, then they won't overload. I do realize that overloading usually a burden to implement, so save it for later, but it's convenient for users
Amazing, takes all the features I love of Rust and simplifies it by trading off the fine grained memory layout / control which is fine for many software use cases. Literally been thinking of creating a language like this, I will definitively try and contribute to this! Nice video!
Really fascinating language, I'm all for ML like languages that place a decent emphasis on static type-safe systems. I wish more languages / tools existed like this for building highly dynamic client-side applications!
Gleam looks pretty awesome! It's honestly really funny to see you go through the playground and be super excited about all these language features like type inference, blocks, and pattern matching. I'm already used to all these features from Rust! Gleam seems very similar to Rust overall but the garbage collection and easy JS transpile are finessed as fuck.
I think "rust but trade a bit of performance for some ease of use" is something a lot of people want. Arguably it's why Go is so popular but go is lacking some niceties from Rust. Rust's type system and error handling are amazing but I don't want to think about lifetimes, and its async situation is tricky. Worth it if you need the Perf, but most of the time I don't
@@SeanLazerI agree that it's desirable to have all these great things in a GC language! I do write in Rust and I usually never have to think about lifetimes. 😄 That mostly comes up if you're writing libraries for others to use, especially if your library is full of generics or is highly concurrent. Me, I'm just writing application code. I use Rust because its type system makes modeling logical problems so much easier. Sometimes I'm working on a problem that has to iterate over hundreds of thousands of files and pull data out of them, or reformat them, etc. and I can use as much .clone() etc as I want!
@@SeanLazer tricky is an understatement. it's a mess. I use rest but I think people tend to pretend it's performance is worth all it's complexity. you have to remember that to even get that performance really* you have to write some complex rust. rust compile time is also quite horrendous. I initially like rust but honestly it's development time is quite stagnating
Super pumped to get back into Gleam now it’s hit 1.0. Been following it for years. It’s been a long road, and Louis and the team has iterated and tried out a lot of things, threw out a bunch of stuff that didn’t fit quite right, and ended up in a really nice sweet spot. Elixir is Erlang but better. Gleam could well be Elixir but better.
The thing about pipes. It's very awkward to type "|>" on a non ANSI-keyboard layout. Programming languages are definitely made for American and British layouts, but the rest of the world already have to wrangle semicolon, and curly brackets being hidden behind modifier keys.
I feel that way about certain characters too, and I've got a US keyboard. Here we call () parentheses or parens, {} braces and [] brackets. I prefer brackets for most things because I don't have to hit shift to type them, but I still use braces often in writing code because it just makes sense. However, I did toy with the idea of an all bracket version of LISP. It looked weird, but I think people could get used to it.
I bought a keyboard with ansi layout 3 years ago and put the lost keys (ä,ü,ö,ß) on another layer that I trigger with caps (press -> esc, hold -> layer). Works beautifully. I use keyd on linux and on windows you can use their ancient keyboard layout creator to use altGr key instead of capslock. If I were to use Gleam I would just put |> on that layer and it would become perfectly ergonomic. Still need to look into ZMK/QMK to see if it's possible to have everything on the keyboard itself. I think it pays to invest a bit of time or money into ones tools.
I’ve gotten curious at some point and wrote a typesafe pipeline library for TS. Turns out, it doesn’t solve much, as unlike Elixir, the whole stdlib isn’t designed in a way that puts first argument first, meaning, even if you go crazy with syntax sugar, it still doesnmt feel right and looks messy. Add async/await to it and its impossible to debug and stack traces are useless 😢
Sounds like you'd definitely like Julia. It can use emojis as variables, has block context scoping, implicit _or_ explicit returns, _and_ function overloding is the whole basis of the langauge. Also compiles down to LLVM, so it is as fast at Rust.
The thing I am most curious about is what would happen to Gleam once Elixir's static types system is ready for production. The biggest thing about Gleam is its static typing. Other than that, the language is either very similar to Elixir, or worse in some cases (no function overloading). I guess being able to run on Javascript runtimes might be a benefit, but at that point aren't you better off just using typescript instead?
In systems where you have elixir/gleam on the backend being able to compile to TS/JS is useful as you dont have to go through hoops to call your function in a different lang
@@dandogamer My question is, why would you ever want to compile to JS in the first place? If it is server side, you might as well keep it in the BEAM world. If it is frontend based, Elixir (Phoenix actually) already allows you to just write JS/TS directly.
@@VoidstroyerThe same reason everything is JS nowadays: you don't have to learn a different language, you can reuse type definitions, validation between back/frontend
Also, another thing they have is syntax (i'm not sure about semantics), i'd say it's familiar to Rust/Zig/Go/JS devs, and maybe even C/C++/Java, whereas elixir is similar to Ruby, and erlang is similar to itself. To existing erlang/elixir devs that probably means nothing, but in a world where people don't even want to leave JS for another C-like lang, familiarity is a good way to attract people to the ecosystem. For one, I'll be trying gleam in some toy projects or AoC. I've tried elixir, but the ruby syntax which i'm not familiar with is a deterrent. And erlang is erlang.
@@araozu I understand the argument of "keeping it in the same language" but that is also one of the biggest problems with web dev nowadays. People using JS for everything. Just because you can, doesn't mean you should. To be honest, if you are already really adept at writing JS, learning a different backend language such as Go (or any OOP language) is so trivial. At this point it is just stuborness of JS devs to not want to use a different language for the backend. I didn't list elixir because I did have some issues getting used to Functional programming. But given some time it is really not an issue. I like Elixir & Phoenix because it allows you to use most of Elixir for your webapp, and if you need heavy client side stuff you can just write JS directly. It enables you to use "the right tool for the job". And since I don't see JS devs leaving it all behind for Gleam, I also don't see Gleam getting that much adoption. But hey, I could be completely wrong. But I do believe that once Elixir gets its static type system ready, it is definitely going to negatively impact Gleam.
A reason I like implicit return is that it makes it very obvious whenever there is a code path that branches out of the function, besides the main (often the happy) path
I am not sure about this one. What is its strength? Seems like there are better languages around for various use cases, that do it better in those situations (eg Rust) The sales pitch sounds nice but overall it still is somewhat in an uncanny valley. Doesn’t even have OTP (and iirc won’t ever fully get there?) I would rather use Elixir and NIFs.
This is a cool project but honestly I don't think it's at the point where I would want to abandon elixir for it. It might be useful to work with both languages, but I've been doing a lot of rust plus elixir work recently (they pair so well together), and I just don't see where gleam fits into all of that. It's definitely a cool language and it will bring more people into the ecosystem which is nice, I plan to dive into it to see what the 1.0 releases really like. Probably the most contentious issue I have with the language is the fact that OTP is not included automatically, and I have to wonder if that has something to do with the multiple back ends that they provide. That and I also wish that it had modular level pattern matching/function overloading like erlang or elixir. I knew that they wouldn't have it because they weren't including the argument amount for each function, which is kind of important if you going to do function overloading in that way. It's easily run my favorite features of elixir / erlang. Edit: I spent some time writing a distributed cache system in gleam. It's the kind of project that is somewhat trivial in elixir if you leverage otp. I found that gleam had a lot of really rough edges especially when it came to interop. The ETS library is deprecated, it's from 0.23 or something of the language and so I had to write my own wrappers. It was relatively easy to do this but I noticed that it was very easy to ignore the static type system by using generics and dynamic types (makes sense given that elixir / erlang are dynamic). One of the reasons why I really like using rust with elixir is because of rust's result and option monads, they make it easy to do error handling on the elixir side by simply passing atoms back to the system which minimizes the downside of using NIFs. On the other hand, when you are wrapping elixir with gleam, because gleam is the language that has results, you kind of have to work around the potential to get a nil or error atom. It definitely works but it's not as intuitive. I also really don't like the actor abstraction, it's just not as intuitive as genserver. I basically ended up writing my own wrapper around genserver. It's definitely rough to try to implement Genserver without function overloading but I was able to make my own pattern within the genserver behavior by passing calls, cast, and infos off to an elixir function. I was able to build a basic supervision tree and implement most of the stuff that I wanted to create, but it did take me a lot more time than it would have taken in elixir because I had to write all of these wrappers. That being said, if the community starts to build more libraries, I don't think this should be as much of a problem in the future. It just kind of sucks that a lot of the OTP functionality hasn't been exposed yet and so you kind of have to go and get it yourself if you want it. It also really doesn't play well with static typing, which is probably why they are trying to build a set-based type system in elixir.
I don't underestand why languages, especially new ones, don't have named imports, like in JavaScript. That's the thing I appreciate the most about JavaScript. It's such an easy win, but of the languages I know only Python and Zig do that.
gleam does have named imports, you use the `as` keyword (if you know rust, it's just how rust does it) so `import gleam/string as str` or a more complex example `import gleam/string.{reverse as rev, append as app}`
Gleam is a very exciting language for BEAM, but just keep in mind that OTP is the true engine behind what makes coding on BEAM unique, and the Gleam standard library currently wraps very little of OTP. This is not a criticism, I'm sure it will be easier to make progress on this with a stable language, and the language author loves the runtime.
I predict that pretty soon we'll see new programming languages and libraries released alongside LLM AIs to help developers convert their existing code bases and reduce onboarding
It's sad that it doesn't try to compete Go's async/multithreading standards with goroutines. Like, where does gleam stand now in this whole "Production languages" list from C++, Rust to Typescript, Python to Go being in the middle? I'm placing it after Go (in terms of not covering this humongous market) of software engineers wanting languages that are simple but also very performant. And Gleam does not seem to touch that market as much as Go does. It just seems to "Gleam" .... i don't know who will use it instead of Go, and why.
They seem to prefer elixir concurrent model and more functional style while go is more procedural Go is pretty good, but in my opinion is not the favorite of many devs because of err != nil and not having a strict null checker This language seems to have strict null checker like rust which reduces drastically bugs caused by skill issue They said that the language does not have null and that I doubt, they probably have null with another name
The main reason to use TS/JS is having the ability to not switch languages if you write frontend / complex web apps. I do not see this advantage challanged, where are many good backend languages out where -> but they are all not good enough if you want to do frontend.
The fact that JS got imports so wrong tells me how good the people that design the language are. Like they are absolute geniuses and luck and timing had nothing to do with their success.
it's funny when I go into FP via Haskell & Closure just as a hobby, I don't code for living or for love, it seemed like it was the answer to so many of the things I hated bout programming. now everybody who codes for a living is all over FP, in some partial way at least. it I never would have guessed JS would adopt TS as a default by 2024.
I keep seeing all these new languages coming out but they keep trying to approach but falling short of CL or even Clojure. The Pipelines syntax is just the threading macro from Clojure, which is available as a macro in Common Lisp too (cl-arrows). Programmers are putting themselves through so much pain and effort just to avoid a few extra parentheses and prefix notation
C# doesn't have pipelining so you're either forced to do what JS does (which is what a lot of people in my team do) or you're forced to make fluent interfaces. Either way, Gleam's pipelining seems powerful. Gleam looks really nice. I wouldn't mind swapping out one of my pet API's to Gleam to give it a test.
I think it is reasonable to prefer the implicit returns for the block expressions. So I could understand if explicit return wasn't considered in favor of implicit, so there is only one way to return values.🤔
Theo, I appreciate your video. I remember I gave Prisma a try because of one of your videos. But you get too excited sometimes, I don't want to make the same Prisma mistake with Gleam, by jumping on it. I'll wait a little, just like I should have done with Prisma.
Prisma is fine, better than anything that came before it, and easy to move off of with tools like Drizzle and Kysely. The mindset it teaches is the value. I'm sorry if you feel burned by adopting it early since we've collectively "moved away". There's a significant gap between my excitement for Gleam and my production use and endorsement of Prisma. Prisma was a tool I built multiple businesses with, and have had a great experience working with at scale. Gleam is a brand new language I'm excited about. I'm not endorsing it. I'm not telling people to go rewrite their stuff in it. I'm just excited. If you can't see the difference between my excitement and my endorsements, might be best to avoid my videos for awhile.
@@t3dotgg Thank you for responding to my comment. Based on your response, I wish you hadn't taken it like an attack on what you do. I apologize if my comment came off like a complaint because it wasn't. To be honest with you, thanks to the issues I had with Prisma, I became a drizzle advocate and a huge contributor to the codebase. Before I used prisma I didn't use any "type-safe" library, just mysql2, node-pg and the like. I admire you as a developer and tend to pay attention to your takes because they make me think in a possibly different way. Hopefully next time I comment on one of your videos I'll express myself in a more positive way.
16:25 onward looks almost identical to Rust. They even followed Rust in rejecting function overloading antipattern. Just last week I was talking with a friend the best replacement for the go niche would be simplified Rust with a garbage collector with the good Rust type-system. I just wish it had traits. Pipes look like a useless gimmick to me, especially when you have to use that awkward _
I found Gleam is kinda not so good. It feels more like a new syntax than a new language. There are too many differences between the JS and BEAM runtime: integers, floats, async is mildly/very different between the runtimes, to the point where it's hard to write code that works on both runtimes.
Wow. I almost dismissed Gleam because of its name and icon. I know that's a bad heuristic, but you gotta prioritize where to put your attention, you know. But boy am I glad that I watched this video. Gleam looks awesome. The design and syntax just makes me want to use it.
Implicit returns turns me off so hard. Was interested until I saw that. And not just implicit returns, but FORCED implicit returns meaning no bailing out of a function early. Hope you like nested code!
@6:33 - I believe that python was the first that pushed for homogeneous code formatting. I mean, indentation is part of the syntax. And there is also PEP 8 since 2001
Although I get what you're saying, I don't think the Whatsapp founders' ability to keep the team small was soley due to the language. I don't even think it was a leading factor. I'm sure it helped, but that same team probably could have made JS or Python work. It might have cost a bit more money to run, required more discipline and testing, but I think at the end of the day, it's the team, not the tools. That said, defensive tools do help and I find Gleam to be really interesting. I really do share their values. Personally, I'm a huge Typescript fan, but a lot of that has to do being a full-stack engineer, and although I haven't looked too much into WASM, for a long time Javascript was really the only option on the FE, and with a small team there's a lot of value in having the same language and ecosystem for everything.
As someone really familiar with Rust, I can't help but laugh a little when you point at Rust features and say they're really nice ergonomics. Aside from the pipe operator (which is excellent don't get me wrong!) this entire demo read identically to Rust code for me. Regardless, very cool to see development in the language space. JS, TS, C++, etc. need to get into gear. They're rapidly being outpaced by much nicer languages to work in.
The main difference is that Gleam is a smaller language, thus making it easier to learn. Having all of the features and more is not always a good thing. I like rust but it sometimes feels like C++ if C++ was invented 10 years ago (I mean in the large amount of features and the general complexity of the language)
@@defalur That definitely is the biggest distinction in philosophy between the two languages. Having said that, Gleam seems to lean more heavily into custom operators like >. for int vs float disambiguation. While that does make it easier to infer the type signatures for things like functions, it does also remove the possibility of a library defining types which also use those operators. Aside from that, it looks like you can just use Gleam code in Rust with only minor changes to the semantics. Where Rust definitely is bigger is in concepts like Async, traits, and generics. I completely agree these things do add complexity to Rust, and there are examples of languages that don't need them (Go and C being the biggest examples). An important distinction in my mind tho is that Rust doesn't have bloat in things that you shouldn't use. While there exists many different Async environments and strategies, none of them are wrong. Whereas in C++ and C, there are definitely wrong things you can reach for (e.g., non-owning pointers, etc.) One thing I find interesting is Gleam, being a more functional language, completely prohibits mutability. That's definitely a viable strategy, but is certainly more complex and restrictive for certain algorithms than Rust's borrow checker, a frequent pain point for new adopters.
@@zactron1997 I disagree - C definitely needs traits, generics, or something of the sort. Relying on macros and void * is awful. C might be "simple". But because it is so simple it creates complex code. Basic paradigms have to be implemented in round-about ways... which is why C++ was invented in the first place. I think Go, to a lesser extent now, suffers that same problem.
@@lucass8119 personally, I agree. I think Go and C are too simplistic for the kind of complex problems an average programmer is expected to work with. Go mostly avoids this problem by having such an exhaustive standard library, but it still has issues. That's why I personally think Rust is a better bet on future software, since it accepts that some concepts just are complex, and is attempting to meet that complexity head on.
Even though it's still in private beta, Jai is still my dream/favorite language. It was initially made for making video games, so it's made to be performant, both in the code being fast, and the compiler being fast, but it's also useful for a lot of other use cases to, as it's also a replacement for C/C++
Sounds like a cool language to use inside an Elixir project for the business logic. I don't think it makes any sense to drop Elixir existing ecosystem (ExUnit for tests, Phoenix for the web, Ecto for interacting with db...)
Great to know, Your content has been so helpful, As far as this Glean, I like it, it’s sweet, at the end of the day is this just like another Typescript ? I’m coming up with something soon :)
"we know what a float is", then proceeds to skip over the part where it explains how BEAM floats are different than floats in other languages (no [-]Infinity or NaN, and division by zero equals 0). Being an elixir fanboy Theo himself probably knows this already, but I'd guess most of his subscribers are JS devs who are not aware of BEAM's float rules, so I thought I'd mention