You may technically be coding in one language, but JS written for Node looks radically different from JS written to run in the browser. In my experience, there’s rarely more than some small chunks of code that you touch that will run in both environments.
If you get some better tooling for multi-language development, it will usually pay off a lot faster than choosing Node for your backend will.
JS in Node and JS in browser is not that different. It's the same language, using the same JavaScript engine (v8 for chrome). The difference is in the APIs provided by Node vs Browser, which is pretty minor. Many popular JS libraries will run in Node just as well as Browser. That being said, JS is a pretty shit language and I don't know why anyone would want to use it for backend work.
My experience is different. I write a decent amount of JS code each year and try to catch up with current best practices where I can, but every year it’s been a bit of a pain to write any code that’s shared between browser and client. Various ES features will have different levels of support in different layers, and this is especially tough once you take into account the various transpilers or polyfills, and the big sticking points for me over the past two years have been ES imports and async (shockingly, with backend support lagging).
These days my main strategy is to have three TypeScript projects—frontend, backend, and common. I then whitelist imports in the common libraries.
> It's the same language, using the same JavaScript engine (v8 for chrome).
I find it completely unacceptable to assume that V8 is running in the browser. In general, I do all of my JS development work in Firefox or Safari, and this saves me a bunch of time checking portability later.
> Various ES features will have different levels of support in different layers, and this is especially tough once you take into account the various transpilers or polyfills, and the big sticking points for me over the past two years have been ES imports and async
New features will have varying support between implementations in any language. You have to take these differences into account even if you're writing frontend-only code.
> I find it completely unacceptable to assume that V8 is running in the browser. In general, I do all of my JS development work in Firefox or Safari, and this saves me a bunch of time checking portability later.
The point is it can be the same javascript engine for frontend and backend. The difference between Node JS and Firefox JS is the same as between Chrome JS and Firefox JS.
> New features will have varying support between implementations in any language. You have to take these differences into account even if you're writing frontend-only code.
For frontend, you use transpilers or polyfills to smooth over the differences between browsers. You then package these up, with rollup or webpack or whatever, and deliver to the client.
My experience is that once you add backend to your list of supported targets, you have to get quite a bit of new tooling in place. Backend code is generally not packaged before running, imports are done at runtime rather than build time, etc. There’s a whole pipeline between your source code and the JavaScript engine, and that pipeline has a different shape for backend and frontend, and typically uses completely different libraries to make it work.
> The point is it can be the same javascript engine for frontend and backend. The difference between Node JS and Firefox JS is the same as between Chrome JS and Firefox JS.
I don’t know what kind of point that is, because it doesn’t matter to me that sometimes the frontend and backend will happen to run on the same engine. I haven’t figured out a way to leverage that fact to give me any additional productivity.
For the projects I’ve worked on, it can end up taking me quite a bit of time figuring out how to make one piece of code work in both frontend and backend, even though I can trivially make it work in either environment as I please.
Maybe other people have already solved this, but I recently went through and made a bunch of PRs to fix a common issue I saw in other people’s codebases and it was super rare to see any code shared between frontend and backend.
> For frontend, you use transpilers or polyfills to smooth over the differences between browsers. You then package these up, with rollup or webpack or whatever, and deliver to the client.
You still have to identify which polyfills you need, add them in, test them, etc. Polyfills are also quite buggy especially for new features from my experience. Also, the fact that you're typically running your build, packaging, linting, testing etc. on Node for your frontend code, says a lot.
> My experience is that once you add backend to your list of supported targets, you have to get quite a bit of new tooling in place.
When is that really even a consideration though? When do you actually need to deploy your frontend app to Node? If you have common model code, or say, input validation/sanitation, business logic, etc - that can easily be identical for both browser and Node.
> Backend code is generally not packaged before running, imports are done at runtime rather than build time, etc.
That really depends on your setup. You can do imports at runtime or build time for both Node and browser. If you're transpiling the setup is pretty much identical.
> I don’t know what kind of point that is, because it doesn’t matter to me that sometimes the frontend and backend will happen to run on the same engine.
How do you run your unit tests, your static code analysis, your packaging and traspiling? Do you run it in the browser or in Node? There is no fundamental difference between JS of the same version in Node vs Browser. Any browser specific or Node specific libraries/features you use are generally not part of any stable JS spec.
> it was super rare to see any code shared between frontend and backend.
Well I'm assuming these are different applications, so that's expected. I don't know why you wouldn't share your model definitions and/or validation/sanitation code though. People do this even for backends/frontends written in different languages.
Depends. It sure is nice to write helper functions that I can import on the serverside js and in the client side. Also if you use next or nuxt or something like that you get universal rendering, which is nice. Totally depends on your use case.
You just have to keep an eye on your client bundle size if these shares functions ref something like underscore, moment or something like that where you’re pulling in the entire thing for one function. I’m aware you can just pick pieces with {} and import, but not everyone is aware and often require the entire thing.
The purpose of next (and perhaps nuxt, I've never used it) is more in the direction of a front-backend. The API-implementing level is usually very different than the one consuming it, even when the consumer is on the server-side. If this wasn't the case, IMHO, projects like Meteor would be immensely more popular.
So coding in one language is a BIT of a red herring.
It only works if the paradigms your app uses are the same in the front and back. And I have nott seen a project where the KIND of problems that need solving (in the front vs back) were close enough for the same language to be a benefit.
I mean at the end of the day, I can build a house with JUST a screwdriver, but man, I'd rather use the right tool fot the specific job.
My company uses Node with Typescript on the front and backend; we’ve strongly typed our APIs and thanks to the excellent io-ts library we’ve also automated the marshaling and unmarshaling data as it crosses the network boundary so that we can continue to use the strongly typed data, and also are constrained by the types to only call our APIs in valid ways.
Some subset of that can be achieved with stuff like GraphQL or Swagger and what not though I haven’t given it a serious try since we’ve never run into hiccups yet with this system, reliant on a very simple library.
We also use the same package manager (NPM) on both ends and can therefore invoke scripts all over our codebase with the same syntax; and although react code and node code is quite different in structure, they all have the same idioms and async syntax and stuff, so the amount of retraining necessary to convert someone from backend to full stack isn’t very large due to the same environment, and keeping code style consistent across the project is also easy.
So I feel like there’s definitely stuff to profit off of with sharing a single language.
Reminds me of yesterday's hn frontpager, the paper about a "programble programming language". they were talking about how domain specific languages within a single language become super distinct and ungrokable to people who are otherwise fluent in the parent language. Think like, an angular dev reading nest.js node backend stuff. With javascript this is sorta everywhere. JS DSLs arent like explicitly uninterpretable to otherwise-experienced JS devs but there is a context shift cost for sure.
Ive not worked with c# much, but is that not as much of a problem with it? Like, is there more of a standardized way of doing things? Python is kinda like that i guess.
