I like the concept of tiling managers, but find the implementation lacking. Touting being entirely keyboard driven and configured by learning what to edit in some text configuration file seems like an anti-feature to me in a GUI management system.
I want to be able to drag the windows around to rearrange things with my mouse, not have to remember or look up keyboard shortcuts. The model of dragging things around with a mouse is intuitive and simple. Sure, have the keyboard shortcuts as well, but give us both! If I wanted to have to look up the documentation every time I need to perform the simplest operation with my software tools, I’d use git.
I would love to see some sort of tiling layout manager you could run on top of KDE. (Heresy, I know.)
>I want to be able to drag the windows around to rearrange things with my mouse
The whole point of a tilling window manager is letting the WM do that for you automatically. I use xmonad, you can use the mouse to select windows and even drag them around (though the combination of floating and tilled windows is ugly and unproductive). You have to configure everything from a text file with comments, but how is that any different from a GUI with labels and input boxes? It shouldn't take you more than a few days to get used to the shortcuts or define your own. Doing everything from the keyboard greatly improves productivity if you also use vim/emacs or other keyboard-driven text editor for work + the shell.
Warren Toomey: "my mail is stored locally, not on someone else's server"
What does this mean exactly? Is his machine a mail server?
I used to know an engineer/scientists who only used a single laptop (linux) and it acted as his mail server. When he travelled he used dyndns (or some such thing) so that delayed mail would be delivered to his laptop. Could never get a straight answer about whether this worked all the time or not (whether he ended up missing emails).
Perhaps it was more 'my mail is stored locally once I download it'?
I use Thunderbird with pop3 settings that ask remote mail server to delete messages once downloaded. Then authenticated SMTP account to send.
No big reason, just what I do.
Edit: I believe that my shell-account provider is deleting those messages once downloaded, and not copying them somewhere else, but I have no guarantee of that as pointed out lower in the tree from the parent post.
If the gentleman is using a mail server on his PC that would be interesting as you say.
Maybe he's using POP3 instead of IMAP? (ie downloading all the email to a local archive) Thunderbird and other clients still support that, and you can still search & tag messages within your local email archives.
It also means if your email server is hacked, they won't get your entire email history - unless your local machine gets hacked, of course.
When the destination server cannot be reached usually the origin server will try again for a number of hours before giving up (and in the meanwhile it may send one or more annoying messages to the sender). I wouldn't do that though. I think that it is better to have a "regular" server (gmail etc.) handle it and have your server fetch the mail from it, without any risk of losing emails. I have a setup with fetchmail+postfix+dovecot that works very nicely.
I don't do a whole lot of software (got caught up on the hardware side of things), but one thing I'm noticing is that nobody makes extensive use of "advanced" IDE features as in eclipse or visual Studio.
Is this personal taste, or has it to do with the scope of the projects these guys are working on?
I did VI for over 10 years, and now I tried IntelliJ Python integration, and my rate of making stupid errors and wasting valuable time has gone down dramatically. Lightning-fast navigation across files, actually working autocomplete, live code style analysis. To be fair, I tried it before and was unimpressed, but in come cases, modern IDE's are becoming superior. And note that it's out of the box - setting up and trying different plugins just seems annoying in the least.
The selection of people certainly makes it less likely that they use IDEs that look like Eclipse or VS. But the editor windows you are seeing may look plain, but I bet most of them have quite fancy features set up, that do a lot of "IDE"-stuff.
For me, IDEs are great for when I have to do boilerplate coding. But when I switch to real business logic I use an editor like vi or sublime. Anything more impedes my thinking and gets in my way.
I'm guessing you're probably not writing unit test or use code coverage plugins or running static code analysis to provide feedback incrementally and quickly?
Thinking the same thing, I use vi for short scripts all the time but can't imagine doing anything more complex than that without the power of an IDE to assist with or even do the grunt work for me.
I appreciate the cross referencing / browsing / documentation capabilities of a good IDE. (especially if it has a vim plugin so my fingers know what to do)
If I uncharitably assume that grunt work means "redundant boilerplate", that sounds like more of a language issue than a tool issue. This said as a Java programmer who has been doing a lot of JavaScript this last year, and REALLY appreciates the brevity. Even if your IDE can generate it, that doesn't mean I want to read it.
Java: the language that has to wear a paper bag over its face while it's screwin' ya. :-)
(think of all those Eclipse plugins to visualize fugly XML bogo-DSLs and other cruft for The Enterprise -- sure, there are worse languages out there, but not in common use now)
Well, it depends on what you mean, I suppose. I've integrated Emacs with tools like those, so that I can (for example) get linting feedback on code in real time, or jump to locations from a unit test failure report. But I've never written unit tests, or used coverage or static analysis tools, that depended on Emacs, or indeed on any other editor. I've never even seen that kind of tight coupling, much less perpetrated it myself. I'm not sure whether that's what you're talking about, or whether it only sounds like that's what you're talking about, but I hope it's the latter.
1. the ability to run everything (a.k.a the whole project) using the preferred SCM (or build tools)
for example: mvn clean install
where your pom.xml already set to have cobertura (code coverage), checkstyle/pmd (linting/style/formatting), and findbugs (static code analysis)
2. the ability to run a very specific unit-test somewhere deep down in your project via your IDE as you're writing code (the IDE compiles your code on the fly...) and get everything I mentioned above :)
for example: I want to run one new unit-test for a new code => I don't have to drop to my console/command-line/terminal to run mvn clean install. I just go to my unit-test, highlight the method name, run it through eclipse/intellij and I get everything.
Oh, is that all? There's nothing special about that. Any decently capable editor should be able to do likewise. I don't know about Vim because I don't use it, but I would assume it can, and I know very well that Emacs can because I do such things in Emacs every time I write code.
Of course it doesn't work exactly the same way; for example, running a single test doesn't involve highlighting it and telling Emacs to do something magical, but rather flagging it in a way that's specific to the test runner and then doing M-x compile RET RET to point the test runner at that file. The first time I do that, I have to tell Emacs what command I want it to run, but that's no great hardship, since it's buffer-local (each file I'm editing has its own value) and, being a simple test harness invocation, need change only when the filename does. If it matters to me, I can put this information in a configuration file, so I don't have to worry about typing it out even the first time. So far, it hasn't mattered to me. (And, yes, when I run a single test this way, I still get coverage information. Linting happens inline, as linting should, so doesn't really figure into things here.)
The mechanism differs somewhat from what you describe, but the result is the same. Just highlighting the desired test and invoking a keybinding might be a little quicker and more convenient, but so far it hasn't seemed enough so for me to go to the effort of writing Lisp code to do it. It wouldn't be much effort, but since invoking a specific unit test alone takes about a second, it also wouldn't save me much time, which is why I haven't bothered doing it.
A major point of interest here is that none of this workflow is specific to a single language or a single toolset. It works the same way across all the languages I use. Since I work in multiple languages on a daily basis, that's very important to me. It sounds like you work primarily, perhaps exclusively, in Java, so that's not as important to you. The problem is, you seem to have assumed that nobody else's tools can do the same awesome things your tools can do. As I hope you've gleaned from this comment thread, that is rather untrue, and proceeding from the assumption that it is makes you come off as both ignorant and arrogant. What you do with this information is, of course, entirely up to you.
Hm, I don't mean to sound arrogant although I can see why my comments perceived to be that way. My apology.
I know most test-runners have the capability to run specific test which suggest that any editor that is capable of executing command-line and bind it to something can support my workflow with some annoyance here and there (as you mentioned, your setup isn't as magical as my description ;)). I mean... someone could've written shell scripts that help them run a bunch of things too...
This is akin of the discussion of using ctags and the need to detect file changes and re-run ctags. Last time I google it there were plenty plugins for each editors and ways to set it up (except there's no clear winner ;)).
The whole idea here is that the IDE helps you not to worry about any of that. IntelliJ supports multiple (mainstream) languages that fulfill my workflow nicely.
Good point that I'm sure Emacs can handle way more languages but the steps to set it up and the limitation (filename change? method name change?) feels more static than dynamic to me.
The autocompletion in both vim and emacs was always significantly worse than any IDE-based ones, no matter the plugin I tried (for C, C++, Python or Go). Most of them are pretty much limited to ctags with no context matching, others simply died on projects beyond trivial.
It's clang-based, so suffers from none of the tags-type issues: it's context-sensitive, and can see straight through gnarly stuff such as identifiers generated by token pasting.
Good points:
- Only mildly fiddly to set up
- Code browsing works well
- Code completion works well when it works
- Cute gcc/g++/etc. wrapper trick means you don't have to change any build settings
Bad points:
- Making it work with code you haven't built yourself on your local system (e.g., 3rd party stuff that's not complete, code for another platform) is less straightforward
- Code completion sometimes doesn't work with certain files (and for no reason that I've ever been able to figure out)
- It's probably my own limitations but I found it extremely inconvenient to debug (internally it's surprisingly complicated: separate client and server, lots of lengthy command lines, client code writes elisp code to stdout and elisp in emacs eats it up, the actual completion popup is a separate package, etc.)
Cider (for Clojure) and Ensime (for Scala) give you excellent support for auto-completion, documentation, and refactoring in Emacs. It's definitely possible, so it probably just depends on how popular Emacs/Vim are in your language community.
For go auto completion in vim, vim-go with youcompleteme works well for me. It's not tags-based, based on the source AST, shows method signatures as well.
Not all auto-completions are equal. I have felt that Intellij's auto-completion is far superior to vim's YouCompleteMe.
There are many other advantages like out of the box support for a lot of common use cases like refactoring, navigation, build tool support, reuse common shortcuts like Control-C Control-V for copy-paste etc.
I would definitely advice new engineers coming into languages like Java to pick up an IDE like Intellij instead of vim.
I like how you can use command keys to quickly toggle on and off the project file and current-file's-components browsers, leaving the "distraction free" * editor window for much of your work.
* yeah, there's a mode with this name that auto-wacks ALL the supplemental panels, AND the window frame. I don't usually go that far, but it does show that they "get it" as far as what many of us IDE resisters disliked.
Pretty much what you expect from such people. Not very customised and quite boring. A long time ago I used to "rice" my dev machine. I customised everything but these days I work with mostly the defaults with a few changes such as font and colour scheme. A terminal, code editor and doc browser (usually a terminal or browser) is all I really need. Oh and a decent music player (Winamp for me, it still kicks the llamas ass).
I'd wager it's super customized behind the scenes, just not in ways that we'd see from a screenshot. Watch any of them work for five minutes and it'd be obvious.
I am not so sure. I have met a few famous developers and I used to be surprised just how "boring" their setups were. Rob Pike was the only person who seemed to really configure his environment. A lot of old school devs just continue to use the command line in whatever form it is available. Even to the point that some just used cmd rather than bother with things like Cygwin.
I was surprised to see Bjarne using a standard Windows XP machine. Not sure if he used Visual Studio or what but he didn't do anything special as far as I could tell.
Then of course you have RMS who still lives in the 1970s :)
Edit: Like I said in my previous post I quite like using whatever the default is. A few changes here and there but on the whole it is much easier to move around if you just learn to work with whatever it comes with.
> Richard Stallman (2002) I don’t know how to make a screenshot, because I normally use my computer in text-mode. I have X and GNOME installed, but I use them only occasionally.
Does anyone else find this quote completely absurd? What a ridiculous disconnect from modern computing, even in 2002, to use text-mode.
Text-mode can help focus on the task at hand without distraction. It also facilitates speedy, transparent work on remote servers.
I tend to use a single maximized app in a GUI environment, often running emacs, though I do like to split the screen in half to allow two full-page, focused views (e.g., code editor and browser). And when helpful, I'll pull up more windows.
If most of what you do on a computer is write, text-mode is great.
I had my low-end laptop boot into text mode for years because I was more productive coding that way, and could always "startx" on a different TTY to get a GUI. Then I discovered tiling window managers and got a better laptop, but I still code mostly in a minimal terminal emulator.
And yet he is more productive than 90% of developers. And it's no surprise.
Walking through a standard tech office you find half the people on random websites like hn instead of working.
The remaining half have dozens of apps open with the devs rapidly switching between them, as if cycling between 5 chat windows, 10 browser tabs, a couple of text editors, a work email client, a personal email client, and an IDE every couple of mins is productive.
"For personal reasons, I do not browse the web from my computer. (I
also have not net connection much of the time.) To look at page I
send mail to a demon which runs wget and mails the page back to me.
It is very efficient use of my time, but it is slow in real time."
At least some reasonable evidence that the habit is harmful. Hard to argue that Richard Stallman suffers (productivity, or otherwise) by doing his computing in text mode...
Text mode, I can understand. No terminal emulator is faster, and if you have lots and lots of text, there's a difference. There are even MP3 players that work in the console, and the windows+left and right keys shortcuts work great to switch between the open terminals.
But the attitude of 'I won't even take a photography of my screen for a semi-noble purpose, something thousands of people do for spurious purposes, because I will not give up an inch of self righteousness, ever'.
I think this is a really strange thing to say. Why do you care? Why would some famous hacker need to live up to your expectations? What you're saying comes down to a teenager complaining "why are you still on Facebook and not on <social network du jour>!?".
I personally can't get used to tiling wm's, but I use Unity's aerosnap-like all the time. It's either full-screen (I have two screens and the more screens you have the more often you can go full screen) or aerosnap. I rarely have random size windows around. I don't have OCD by any measure, but random size windows on top of each other just annoys me.
I used to use tiling managers, but then I found a terminal emulator (Terminator) with the feature of splitting screen and I just run this in maximized mode and everything else in floating mode.
For Stallman's case, surely there is an emacs command to copy the the whole frame, mode line and all, to text for pasting into a mail buffer.
But seriously, it does make me wonder how much of the vast system he retains and uses. Using it for a decade, myself, (certainly not working ON it or understanding the internals) I learn new parts every month but forget a lot, too.
There is [1]. I'd like to find a package that saves the layout in a form that regenerates the screenshot in a terminal when you cat the text file it outputs.
