I completely switched to VS Code once I discovered the remote development feature [0]. It allows you to run VS Code locally but work on a project in a different environment (via SSH, Docker, WSL). The integration is seamless - search, debugger, terminal, extensions - everything looks and behaves as if it was running locally but is delegated to the configured remote. You can even have different remotes opened at the same time.
I use this setup primarily to have at least some sort of a barrier between my system and the gigabytes of NPM packages that get downloaded as dependencies. Moving between systems easy as well - I just copy the VM images. It also makes it easy to experiment a little bit. If I want to e.g. upgrade an important package and something goes wrong, I just revert the VM and I'm back in business. Having Arch as the distro is a nice bonus.
There's a chance that IntelliJ IDEs will get the same remote functionality as well but the timeline is unclear [1].
This is such a hugely undervalued feature, and no other editors seem to provide anything remotely close. I found it particularly useful for the COVID work from home - I yeeted my work laptop into a corner and just SSH into it to continue working with the ergonomics of my home desktop machine.
As someone who used tramp and switched to VS Code, it's absolutely not the same thing. The difference is plugins work on the remote side of the connection without having to be aware they're in a remote connection. VS Code makes this work by having an agent running on the remote host. For the first time I actually got autocomplete and code jumping working without a horrendous amount of effort.
This is also true with TRAMP. Anything in core Emacs, and any package outside it that isn't aggressively poorly written, works transparently with TRAMP. The core utilities for interacting with the OS such as start-file-process are all TRAMP-aware. (Or more specifically, TRAMP provides file handlers which customize start-file-process)
One area I have found where TRAMP really suffers is when you have a relatively high-latency connection. I'm living on the east coast right now and connect to a dev box in Washington to work, and tramp is basically unusable. Every time you enter a folder in a project, emacs will stall for 1-2 seconds (unless the folder has been entered and cached before: unfortunately every time you restart emacs you'll have to go through it all again). If you've forgotten to disable autocomplete in a remote shell buffer, the first character you type will cause emacs itself to hang forever. There's unacceptable lag when opening and even navigating files. Good luck ever running grep or magit.
With VSCode on the other hand, I have no issues at all: couldn't even tell the project was remote.
I still use emacs (emacs -nw via ssh), but it's nowhere near VSCode's responsiveness and now I cant use GUI emacs. It's definitely a source of frustration :/
That comment doesn't contradict the post at all. If anything it supports that statement and then goes on to mentions that high latency connections are problematic.
High latency connections are a fact of life. If Emacs Tramp, for which development was started in 1998, when high latency connection were a lot more common, still has issues with high latency connections, that's not a glowing reference for its architecture.
Especially since it seems VSCode manages those connections better.
It's ok to admit your favorite tech has flaws, everything has them.
Not disagreeing with that. I am just saying that the previous post was that plugins don't have to be modified with Tramp and work transparently. The post referenced above does not disagree with that. Instead it says that tramp does not deal well with high latency connections. I think is important to be clear where the issue lies if one wants to get it fixed.
> The difference is plugins work on the remote side of the connection without having to be aware they're in a remote connection.
This is cool when the machine you are connecting to is more powerful than the one you are sitting on. This is far less cool when the machine you are connecting too is less powerful than your desktop (RPi or Beaglebone, for example).
Remote has been a game changer for my development as well. Work forces me on a Windows machine with the full suite of corporate spyware. Said spyware loses its mind when compiling, debugging, language server inspection, git actions on large repo, etc. Performance loss is somewhere in the region 2-10x.
With remote, I am able to do all coding on a non-infested Linux server with almost seamless usability.
Your company is equally cool with you just having all of your code/IP on a unsecured (i.e. not running crazy large antimalware, not actually insecure) Linux box?
sounds like my company: any windows machine is infested with a dozen or so McAffee things that make it unworkably slow. But none of the drones that role out McAffee has ever touched linux or osx, they prefer to leave that alone. So even though I actually prefer windows as a desktop OS, I mostly worked on a MacBook (until the pandemic, now I WFH on a my personal windows machine). But since WSL2 is finally backported to whichever outdated windows my corporate laptop runs, it is actually quite usable with VSCOde/remote containers + docker
I read about this fact elsewhere also. Could you expand on this a bit? My impression was that VSCode and its associated extensions are open source, so I'm curious how Microsoft could make certain parts exclusive to its own build of VSCode.
Because VSCode is under a permissive license (MIT) rather than a copyleft license (GPL), Microsoft can add any proprietary extensions it wants. "Permissive" licenses are only permissive for the creator and other developers (only sometimes), not for the end user.
In order to contribute you have to sign away your rights, so it's only permissive to MS. It's a pervasive view on open source, cheered by sockpuppets, making people believe MS is now a open source company.
The situation is similar to Google Chrome/Chromium: Most of it is open-source, but the release binaries contain some proprietary bits for telemetry, marketplace access, etc.
Eclipse foundation has a new project, Theia[1], which forks VSCode and one of the reasons they state is that the extension ecosystem is not opensource.
They want Theia to be a tool to build your own editor and it can be built as a hosted solution (like code-server) or a standalone editor
Edit: I see that the article also mentions Theia but this point about Theia vs VSCode is not mentioned.
I use JetBrain's remote python interpreter functionality[0] with great success and convenience. For anyone whose development environment is appropriate for its use, I honestly cannot recommend it highly enough. It's only available under the full "Professional" license version, however.
This uses SSH and SCP to automatically deploy the code to the remote environment and run it with the interpreter on that machine, rather than your own local laptop/desktop. One advantage is that it does not require any special software to be installed on the remote server. I believe VS Code's solution requires a pretty hefty installation consisting of basically an entire copy of VSCode onto the remote server.
The downside of JetBrain's approach... I believe that it requires a local copy of the files in order to perform the "LSP"-like syntax error highlighting / suggestion functionality that JetBrains software is so lauded for. This could be difficult/inconvenient if your codebase is particularly enormous, or if you aren't allowed to mirror it locally.
However, given the amount of support that JetBrains has for this kind of setup, I wouldn't be shocked to see an additional option for something more in the style of "VS Code Remote Development"
I spoke to the team in KubeCon last year and they said “our research team is looking into it” with a hand wave. I doubt IntelliJ will actually implement this anytime soon, sadly.
It is also great for string all dependencies for development together in a docker-compose file. On any machine you just have to git clone and open vscode in the container and you are good to go. Even if you require a particular version of postgresql, some compiler version or a complete Kafka cluster; everything can be configured in a Dockerfile/docker-compose file and stored with the code. No more pages long readme's of how to get all dependencies set up just right. Especially when often switching between projects and machines this is a godsend. And for onboarding it is so easy to get someone a fully configured working environment.
Absolutely. I build Android images (AOSP/kernel/lk customization) and at the beginning I was using Eclipse over a shared network folder (I have dedicated remote hardware to host and build these monstrous images), but everything was a chore. Then I switched to VSCode + Remote Development over SSH.
Imagine placing the workspace at the root directory of the entire Android source tree, pressing ctrl + shift + F and finding anything in less than a couple of seconds. For C/C++ code, press F12 just ANYWHERE on the code to go to the definition, no matter if it's the entire kernel code, or vendor or AOSP code: it just finds the definition at lightspeed.
For building I use the integrated console. If there is an error or a warning anywhere, you can just click on the console an it takes you to the file with the error/warning. Also it integrates perfectly with Git and has extensions to parse DeviceTree files.
And everything was just much more appreciated while working at home during pandemic.
It sounds like a testimonial advertising, but it's the way it is.
> I use this setup primarily to have at least some sort of a barrier between my system and the gigabytes of NPM packages that get downloaded as dependencies.
You can do something similar with IntelliJ by running npm inside a Docker container with your working directory mounted into it. IntelliJ will happily index the resulting node_modules folder without executing any code on the host.
The only thing that's been preventing me from fully adopting this is that I don't have my local shell environment (autocomplete, syntax highlighting, etc.) anymore. Maybe I'll have to look at getting that setup in a container sometime, and then copy that over to every container I use.
Maybe it works for literal remote development, but the jetbrains support for remote dev in a docker container is a complete afterthought. There's tons of critical, relevant feature requests that have been stuck in the pipeline for years. In vscode, it is well thought out and polished. My team desperately wanted development out of containers to ease on-boarding and standardization of development environments. Trying to make it work with CLion was a nightmare that was practically dead before it started, and we thought we'd never get it. Then we found vscode's support and have never looked back. I sometimes laugh looking back at what we were having to do to get it to even slightly work, and think about all the troubles we would have had ahead.
I use this setup primarily to have at least some sort of a barrier between my system and the gigabytes of NPM packages that get downloaded as dependencies. Moving between systems easy as well - I just copy the VM images. It also makes it easy to experiment a little bit. If I want to e.g. upgrade an important package and something goes wrong, I just revert the VM and I'm back in business. Having Arch as the distro is a nice bonus.
There's a chance that IntelliJ IDEs will get the same remote functionality as well but the timeline is unclear [1].
[0] https://code.visualstudio.com/docs/remote/remote-overview
[1] https://youtrack.jetbrains.com/issue/IDEA-226455