Hacker News Reader

Comment by lostcolony | Hacker News Reader

'Avoid using language extensions and libraries that do not play well with your IDE'

I'm of the opinion this should be extended to "does not play well without an IDE". Because even in projects that said "everyone, use Eclipse(/IntelliJ/whatever)", and tried to share project files, there was constant pain in ensuring that everyone had the same development environment ("Oh, yeah, I made a local change to my project file, but then I also made a change that needed to be shared, and whoops, I broke everyone" and the like). Not to mention trying to work with code deployed onto boxes without an IDE. I've gotten to where "if I can't make sense of it and be productive in it with just vim, grep, and find, your code is too complex".

View on HN

Well, sure, but that's why you don't share project files.

I think that if you do things correctly, your concern is mostly mitigated. My last job was Java/Spring. As someone noted above, you really don't want to be working with Java/Spring without IntelliJ.

But we never had issues with anyone breaking builds due to screwing up project files, because those weren't shared and the project would run/build/test immediately after import into the IDE without any additional tweaks.

I think it's sort of odd that many folks complain about IDEs, but then use vim with a million plugins. Isn't vim + plugins = IDE? Or, if the overall point is "java sucks, don't use java" then fine, but if you have to use it, I couldn't even imagine using plain old vim without any language support...

Sure. Does it require a specific IDE then? Does it cleanly import, from scratch, with working tests and etc, regardless of whether they're using Eclipse, IntelliJ, Netbeans, etc? Or is there One True IDE? Is that IDE available everywhere you have to touch code? Is it straightforward to integrate it with your CI/CD tools, despite conforming to that IDE convention? Will it continue to cleanly import after you've left the code to rot for a couple of years in production, and then have to come back in to maintain it? Etc. I've been bitten by all of these.

Whereas I've gone to projects that were written without all of this, and even without getting the code to even compile locally (because of dependency hell that I didn't want to suffer through if I didn't have to), was able to diagnose and fix issues, because the code was written to be understood with only a text editor.

The idea is that you still conform to the convention of the project (Rails, Spring, etc.), not the IDE, and the FOSS CLI tools etc. can still understand the project, but so does the IDE (because it is made to understand the convention, e.g. how the annotations/XML files work, how to get a list of all your Rake tasks).

By not sharing IDE project files, not having any IDE-specific machinery in the project, but still expecting CI to be able to run it, being able to theoretically hack on the project and get results (build, test) even with just bash and grep and nano, you get the best of both worlds.

Modern IDEs are better at this than they ever used to be. I was really anti-IDE for a long time, but JetBrains' products made it make sense finally.

Someone wrote a post a while back about this topic and called it the grep test. See: http://jamie-wong.com/2013/07/12/grep-test/

It's funny that you mention the Java ecosystem as one of the worst offenders, since the nature of the language itself and the culture around best practices in the early days should have put it in a particularly favorable position here. This is covered partly in another great post titled "Java for Everything": http://www.teamten.com/lawrence/writings/java-for-everything...

Unfortunately, I too find most Java projects are developed in a way that make them completely unapproachable without the assistance of an IDE. I'm secretly hoping that Microsoft announces first-class support for Android app development in VSCode, given that the team's overall gestalt is on producing a nimble, code-oriented editor. They may not otherwise have any incentive to make that kind of thing happen, except that I think it would be a real boon for VSCode adoption. Anxiously waiting to see if there will be any interesting announcements to come out of their developer conference this weekend, although I'm not counting on it.

Even more amusing, I didn't mention Java, just IDEs popular with Java (but which support other languages too). I was certainly thinking it, though. As were you, as soon as you read the description of the problem. Yay, validation.

I agree completely and totally. A good programmer's text editor (e.g. vi or emacs) is far more general than an IDE, more easily extensible and rather more future-proof (folks will still be using emacs & vim in twenty years; will anyone be using Eclipse?).

Requiring an IDE is, to my mind, a symptom of a far-too-complex environment which will lead to breakage.

People have been using eclipse for almost 15 years, and if they stop using it, it will be because there is a better IDE (see: IntelliJ). Things like smart auto-complete, click-to-definition, history/diff merge views and integrated build support/dependency management are a huge productivity win.

If something requires an IDE in order to run it, edit it, etc, that's kind of a code smell. The IDE might be masking complexity for you, but it is creating complexity on the server side or for developers who prefer to use something like vim. Getting something to build for the first time in Eclipse is a pain. So many settings that keep changing which menu they are under in different version and you have no idea what you are doing wrong. If I have to click through dozens of menus and type in values in order to get it to build for the first time that is another code smell. The ideal scenario is to check it out, run a single command for package bundling; setup; etc, and then run.

That reminds me when I first came across autocomplete, I thought, "Wow, this is really nice, I like it." Shortly thereafter, I had a horrifying worry that someday I would come across projects that couldn't be understood without autocomplete.

And soon enough, my next job the code was impossible to understand without autocomplete and a debugger. I am still sad about that to this day.

Reply on news.ycombinator.com