broadway VS livebook

You can actually have "background jobs" in very different ways in Elixir.

> I want background work to live on different compute capacity than http requests, both because they have very different resources usage

In Elixir, because of the way the BEAM works (the unit of parallelism is much cheaper and consume a low amount of memory), "incoming http requests" and related "workers" are not as expensive (a lot less actually) compared to other stacks (for instance Ruby and Python), where it is quite critical to release "http workers" and not hold the connection (which is what lead to the creation of background job tools like Resque, DelayedJob, Sidekiq, Celery...).

This means that you can actually hold incoming HTTP connections a lot longer without troubles.

A consequence of this is that implementing "reverse proxies", or anything calling third party servers _right in the middle_ of your own HTTP call, is usually perfectly acceptable (something I've done more than a couple of times, the latest one powering the reverse proxy behind https://transport.data.gouv.fr - code available at https://github.com/etalab/transport-site/tree/master/apps/un...).

As a consequence, what would be a bad pattern in Python or Ruby (holding the incoming HTTP connection) is not a problem with Elixir.

> because I want to have state or queues in front of background work so there's a well-defined process for retry, error handling, and back-pressure.

Unless you deal with immediate stuff like reverse proxying or cheap "one off async tasks" (like recording a metric), there also are solutions to have more "stateful" background works in Elixir, too.

A popular background job queue is https://github.com/sorentwo/oban (roughly similar to Sidekiq at al), which uses Postgres.

It handles retries, errors etc.

But it's not the only solution, as you have other tools dedicated to processing, such as Broadway (https://github.com/dashbitco/broadway), which handles back-pressure, fault-tolerance, batching etc natively.

You also have more simple options, such as flow (https://github.com/dashbitco/flow), gen_stage (https://github.com/elixir-lang/gen_stage), Task.async_stream (https://hexdocs.pm/elixir/1.12/Task.html#async_stream/5) etc.

It allows to use the "right tool for the job" quite easily.

It is also interesting to note there is no need to "go evented" if you need to fetch data from multiple HTTP servers: it can happen in the exact same process (even: in a background task attached to your HTTP server), as done here https://transport.data.gouv.fr/explore (if you zoom you will see vehicle moving in realtime, and ~80 data sources are being polled every 10 seconds & broadcasted to the visitors via pubsub & websockets).

While in general you are right, I'd strongly say "it depends".

The raw BEAM ecosystem (things that come ootb) is huge in itself, many things that would require additional machinery/libs/infra/... in other tech stacks are simply covered right away with stuff like OTP.

The machine-learning ecosystem is kinda thriving, not fully at python levels yet but catching up rapidly and already outshining most other tech stacks - especially factoring in again the BEAM underpinnings that allow stuff like https://elixir-broadway.org/ for data pipelines (which requires a lot of additional python machinery to even replicate), and I'd argue that iE LiveBook already is a much better story than Jupyter notebooks.

The web framework story is already excellent, as you mentioned with Phoenix/Ecto/Liveview/Oban/... which are kinda best-of-breed in the industry right now. Not only for the first few days into a project (lots of tech stacks are compelling here for one reason or another), but the scaling up capabilities are flat out amazing, you can get _so much_ mileage out of the stack before even looking into anything like k8s or whatever and can focus on iteration features instead of spending time in optimizations/infra/... even when traffic peaks occur.

What may be missing are some adjacent libs or QoL in many smaller places. But its getting better for a while... we now have a great storybook reimplementation that doesn't suck ass with nodejs ecosystem craziness/provisioning/slowness like the real storybook. We have usable solutions for i18n or auth that may not be fancy but do the job. And since being Elixir, many missing things are just a few macros away if you need it. Especially the last 2 years have been quite a ride, and I was a phoenix user since probably 1.2 years back, but recently things are stepping up.

Right now I am eagerly waiting for BeaconCMS to mature enough, thats an absolute pain point to get solved since nearly every web platform at some point needs some CMS style free-style pages and right now I have to always implement an integration to some external system... can't wait to have this as a lib mounted in my app like everything else. Oh, and types of course, one of the few points that keep people from trying elixir and I (coming from Rust, Go, Typescript) learned to love. But it looks like we're getting there.

> In other words, there is a subset of distributed problems that Distributed Erlang solves very well out of the box: homogeneous systems working on ephemeral data. And some of the scenarios above are very common.

Speaking of which, I'm looking forward to using Broadway [1] in a new project here in my company. Here, people are using an enterprise integration engine specialized in the healthcare space [2], with built-in single-branch version control and all actions going through the UI.

As I come from a background of several years with Ruby on Rails, I really hope to convince people to use this great library your company created, since RoR is severely lacking when handling heavy concurrency like when gluing multiple APIs in complex workflows. Software engineers are going to love it, but integration analysts are used to IDEs with GUIs, so we'll need to create a pretty admin dashboard to convince them to switch.

[1] https://elixir-broadway.org/

To listen a message broker the most used library is broadway, this library helps to create GenServer's that listens a specific queue and process message by message (or by chunks).

Outside of that, Elixir can be used for data pipelines, audio-video processing, and it is making inroads on Machine Learning with projects like Livebook, Nx, and Bumblebee.

Like many applications, our infrastructure relies on queues to decouple various components. In our system we use AWS Kinesis as a data stream, consumed by Broadway consumers for some critical parts of our infrastructure. We have found that sometimes our Broadway consumers for AWS Kinesis fail in ways that do not gracefully recover when they crash. For example, each Kinesis shard has its own supervision tree managed by the Kinesis Broadway consumer. We found that if a shard consumer experienced a crash-inducing error, the shard would not restart and the crash would not cascade up to the Broadway producer. While we have worked on contributing to this consumer library, we decided that it would be important to have runtime control over stopping and starting consumers to respond to such failures just in case.

If you are looking for an even higher-level abstraction, Broadway is a good starting point. It is built on top of GenStage and offers several additional features, including consuming data from external queues like Amazon SQS, Apache Kafka, and RabbitMQ.

This was a great excuse to use Elixir's Broadway. A Broadway pipeline consists of one producer and one or more workers. The producer is in charge of producing jobs. The workers consume and work those jobs, each working in parallel. Broadway gives us a few things out of the box:

To get started you need a running instance of Livebook

LiveBook site

I prefer functional programming with Livebook[1] for this type of thing. Once you run a cell, it can be published right into a web component as well.

[1] - https://livebook.dev

If you're a coder, Livebook might be worth a look too. I certainly have my eyes on it.

Would highly recommend looking at Jose's use of livebook to answer these. It makes testing easier. It's old but still relevant. Video link inside

That's hard to say. Racket is a rather complete language, as is F# and Elixir. And F# and Racket are extremely capable multi-paradigm languages, supporting basically any paradigm. Elixir is a bit more restricted in terms of its paradigms, but that's a feature oftentimes, and it also makes up for it with its process framework and deep VM support from the BEAM.

I would say that the key difference is that F# and Elixir are backed by industry whereas Racket is primarily backed via academia. Thus, the incentives and goals are more aligned for F# and Elixir to be used in industrial settings.

Also, both F# and Elixir gain a lot from their host VMs in the CLR and BEAM. Overall, F# is the cleanest language of the three, as it is easy to write concise imperative, functional, or OOP code and has easy asynchronous facilities. Elixir supports macros, and although Racket's macro system is far more advanced, I don't think it really provides any measurable utility over Elixir's. I would also say that F# and Elixir's documentation is better than Racket's. Racket has a lot of documentation, but it can be a little terse at times. And Elixir definitely has the most active, vibrant, and complete ecosystem of all three languages, as well as job market.

The last thing is that F# and Elixir have extremely good notebook implementations in Polyglot Notebooks (https://marketplace.visualstudio.com/items?itemName=ms-dotne...) and Livebook (https://livebook.dev/), respectively. I would say both of these exceed the standard Python Jupyter notebook, and Racket doesn't have anything like Polyglot Notebooks or Livebook. (As an aside, it's possible for someone to implement a Racket kernel for Polyglot Notebooks, so maybe that's a good side project for me.)

So for me, over time, it has slowly whittled down to F# and Elixir being my two languages that I reach for to handle effectively any project. Racket just doesn't pull me in that direction, and I would say that Racket is a bit too locked to DrRacket. I tried doing some GUI stuff in Racket, and despite it having an already built framework, I have actually found it easier to write my own due to bugs found and the poor performance of Racket Draw.

This looks very similar to LiveBook¹. It is purely Elixir/BEAM based, but is quite polished and seems like a perfect workflow tool that is also able to expose these workflows (simply called livebooks) as web apps that some functional, non-technical person can execute on his/her own.

1: https://livebook.dev/