I get frustrated by people who insist on a complex vi/emacs setup for something Jetbrains IDEs do better out of the box. Or people who insist Sublime is good for editing their complex React project.
Most things can be solved with plugins though.
I find developers who try to force their tools on others annoying. I'm much more productive on scripting languages with an editor. They just feel more responsive.
Vim is a better text editor than anything else. By far. Unfortunately, vim is terrible at basically everything else.
The problem is, once you learn to use vim's text editing capabilities, then editing code using anything else feels like a pain. I'm not talking about variable lookups / etc. I'm talking the pure "move the cursor to a certain place and enter some text" aspect of it. If you don't know vim, then it's hard to understand how slow editing text feels without it - it's kind of like going from not knowing how to type properly to being a solid touch typer. (not that extreme, but that's what it feels like).
Unfortunately, because vim sucks at everything else, my setup has to be complex to actually get it to do the other things a good IDE does out of the box. If I could do it the other way around, I would - I'd love to be able to use vim-like editing inside of an IDE, but every vim-mode I've tried so far just doesn't work properly (except one - evil mode in Emacs - which is why I recently switched to Spacemacs, but that's another story).
And like almost every other emulation I've ever tried, within 1 minute, I found 5 things that didn't work correctly. And I'm not exaggerating here - I literally spent 1, maybe 2 minutes and found 5 things.
I'll have to use it for a while to find out whether it's just a few minor annoyances, or much worse than vim.
1. First thing I did was to fold all the functions in the file - this is something vim does pretty badly, actually, because it doesn't know how to deal with Python syntax without lots of customization. But anyway, the command for "fold everything that can be folded, recursively" is zM. Then the command to open everything back up is zR...
Except that for some reason, this vim emulation doesn't open everything, it kept the imports at the top of the file closed.
Verdict: Probably not a big deal, just inconsistent (but maybe PyCharm's way is better?)
2. I then did a standard thing I like to do - if you want to delete a paragraph, you hit 'dip'. Works well. If you have a few lines of whitespace between paragrpahs, and you want to delete everything except one line, the command in vim is 'dvip'. This is actually somewhat obscure but I use this constantly to clean up code. Didn't work.
Verdict: obscure command that's rarely copied correctly, but I use it all the time. Annoying.
3. I then tried to change some text. Let's simplify and say that I did 'ci(' to change the contents of parenthesis. Worked well. However, I tried to undo the change. In vim, this is a single 'u' since the change command is atomic. However here, this required hitting 'u' three times.
Verdict: Ok this is really annoying. It's unpredictable and screws up my muscle memory, plus it's a very very used feature.
4. This might just be my fault - I tried to navigate my Python file using ']]' and '[['. The second one took me to the end of the file, the first did nothing - neither did what I want, which is to jump through various methods/functions. I also tried ']m' and ']d'. This one is not stock vim though (or not exactly), so it might be possible and I just don't know it.
--
I don't remember the other thing.
And btw, I totally don't mean to diss PyCharm, which is awesome in many ways. And I could be only nitpicking here - though again, finding 4/5 problems in 1 minute is not a good sign. But that's my point - emulating vim is, for whatever reason, either really hard, or not something that people spend a lot of effort on. That's why I have to rely on vim itself (or actually Spacemacs nowadays, which amazingly does have an almost completely working vim emulation.)
The biggest plus is when working with larger projects. The single-threaded autocomplete and jump-to-definition functionality was very slow in Vim, stalling the UI updates. Goland feels so much smoother in these cases.
It would be useful to say what these 5 things are.
Response Speed. Most emulations tend to fall behind my actual Vim command speeds, which really destroys the vim editing flow.
Occasionally emulators will not honor mode switches correctly or on the first try. So I would expect to be in insert mode after a complex set of commands but the emulator isn’t there. This is also likely due to speed issues, but maybe not.
Emulators do not do a good job of handling macros, and vim’s repeatability. The auto stuff they try to do (such as formatting, etc) precisely the reasons one would prefer an emulator, can mess up repeatability.
Many of them don’t handle certain basic commands. A lot of emulators, for example, do not handle commands like HML which shift the viewport, possibly because the IDEs do not really give them access to it, or maybe because they aren’t known well enough that they fall way down the priority list.
It also has two popular vim packages, surround and easymotion you can use too, which I really like.
I then customize all the IDE stuff to vim combos.... super nice.
only PITA is in some drop downs on a number of jetbrains products is that it doesn't play nice with auto hotkey, I have ALT-J and ALT-K bound to up and down arrow keys which makes for a much nicer vim experience when dealing with autocomplete drop downs ( or any drop down )
Been paying for Intellij for nearly a year now and I don't regret a day of it. Well okay, I do regret the one day EAP was totally broken by the VIM plugin, by besides that it has been brilliant.
Its not perfect but its good enough for moving around and copy/paste.
I occationally go back to emacs to macro some data into different formats.
In vscode, I'm usually already looking at the call site of a function I want to edit. So I move the cursor into the function call and hit F12 which takes me to the function definition to edit it. Once I'm done, I hit Alt+LEFT to navigate back to where I was to continue on...
If I didn't start from a call site but I know the file I want to edit - I hit Ctrl+P, type the name of a file, hit ENTER - now I'm in the file, ready to edit it.
To jump to a function, I collapse all functions (Ctrl+K,Ctrl+0/1/2/3/4) and then scroll (or PAGE UP/DN) to find it...then use arrow keys or Ctrl+G to enter the line number. Now I'm ready to edit the function. I might have to hit Ctrl+Shift+] to expand the body of the function.
Alternatively, I Ctrl+F and type the name of the function I want to edit, pressing ENTER to find it...and I'm ready to edit the function.
What do you do in vim that is far easier?
Stock vim is mouseless, and without arrow-keys, which means relying on /? (searching) or other keys to achieve the function of arriving at text. CTRL-F never quite matches that (after you hit enter, the CTRL-F dialog is still open, and needs to be closed with ESC or by clicking away), and moving to the arrow keys/pgup/pgdown/mouse is slow.
Having said that - there are some plugins that match this behaviour - I particularly like the ones where a key (f in vimium) highlights each word with a character, allowing for one or two keypresses to jump anywhere on the screen.
The part about moving the cursor. Vim users typically don't move the cursor with the mouse but rather with the keyboard. Having to move your hands off the keyboard to reach for the mouse then back is the part that starts to feel sluggish if you get used to not doing it.
In a modern IDE you have a multitude of shortcuts and a mouse, and they complement each other.
Longtime vim users who didn't know how to use vim? That's... interesting.
E.g., say you somehow got to a file, and it now has a bunch of functions in it, and you're literally looking at the line you want to edit. To get to that line, there's a bunch of different ways, but you can do this in like 3 key presses. Then, within the line, you can change anything in it really easily.
This sounds weird, so I'll try to give an example. Let's assume you have something like:
Func(self._something, self.something_else).validate, 'Validate exam foundation table')
And let's say you want to change the string that says 'Validate [...]' to say 'Change is good!!'.
To do that, once you've gotten to the line, I'd probably hit two key presses that mean "go to the first open quote in the line", then I'd hit the keys that mean "delete everything inside of the quote character", then I'd type what I want. All in all, I'd be typing these characters before actually writing what I want to write: f'ci' (so, 5 characters).
That probably looks either idiotic or daunting - but the idea of vim is that it's a language - you are literally thinking in your head "go to that quotation mark", then tell vim to do that and it does. Then you're thinking "delete everything inside of these quotations", and it does it. The distance between thinking what you want to do and doing it is almost zero - it's as close as you can get to literally telling someone exactly what to do.
If you're using a non-vim editor, you'd probably have your cursos in the begining of the line, and have to do "cmd+right" several times to skip over words in order to get to the first word of the string, then have to "shift+cmd+right" to get to the end of the string, then write what you want.
There might also be some small annoyances - the last shift+cmd+right might take you over the last quotation mark, and you'd end up selecting the closing paren as well. In vim - that doesn't happen, you're saying exactly what you want to have happen.
Or you might want to undo this change - in vim, it's always hitting "u" once, because in vim, changing the string to something else is an atomic motion. In another editor, you might have to undo a few words at a time, or maybe not - it's not always predictable. And since it's atomic in vim, it's also easy to re-apply the same action to another string - so changing another string to the same thing is a simple matter of hitting "." on it.
And of course, this is just a tiny example. It's easy to dismiss as trivial or unimportant - these are tiny savings of time, after all, and even if you add it up, I doubt the "time saved" in doing things more efficiently is worth much.
But once you know this language for telling an editor exactly what to do, then using an editor that doesn't speak that language just feels super slow - it feels like you are not able to communicate properly, and suddenly, what used to be zero gap between thinking and doing, takes a long time.
Double-click the word to select it, and start typing right away.
> it's as close as you can get to literally telling someone exactly what to do.
I usually tell someone "Change the word 'validate' to 'something else'". SO i grab my mouse/trackpad, and get there in one click (or double-click).
> It's easy to dismiss as trivial or unimportant - these are tiny savings of time, after all
You should observe most of these "savings" in real life. There's rarely a time when I can't do stuff with code faster than a Vim user. Because by the time the vim user has typed the required sequence of letter, or counted which line/symbol/mark to go to, or wrote a macro correctly, I'll have finished all I needed to do using IDEs built-in capabilities and/or a mouse.
For example: say I jump to this function and want to fix a bug (the `search` should strip off the first two characters here, not just one). So I need to change `regex.search(s[1:], ...)` to `regex.search(s[2:], ...)`:
def compile_hg_glob(line):
pat = glob_to_re(line)
# Mercurial ignore globs are quasi-rooted at directory boundaries or the
# beginning of the pattern.
pat = '(^|/)' + pat
# Mercurial globs also have to match to the end of the pattern.
pat = pat + '$'
try:
regex = re.compile(pat)
return lambda s: regex.search(s[1:] if s.startswith('./') else s)
except:
warn("could not parse hgignore pattern '%s'" % line)
return lambda s: True
Or suppose I'm looking at this and want to add a `,` character to the end of each of the type entries, because I forgot Python likes its commas:
result = result | {
'a': TYPES_ALL
'e': TYPES_FILE_REAL
'x': TYPES_FILE_SYMLINK
'c': TYPES_DIR_REAL
'y': TYPES_DIR_SYMLINK
'f': TYPES_FILE
'd': TYPES_DIR
'r': TYPES_REAL
's': TYPES_SYMLINK
}[c.lower()]
These kinds of things happen constantly, and when you've got Vim burned into your fingers trying to edit text without it (actually edit text, not just jumping to something) feels like typing with oven mitts on.
2: select the first :, ctrl+d until all are selected, right arrow, comma
Doesn’t sound any more efficient? Plus, who writes code without an auto-formatter these days? Problem two would have been solved by the linter the moment you wrote that.
Obviously the first part is language-neutral, and is exactly why I think multiple cursors is one of the greatest inventions in text editing! :)
Eh. I'd probably do 'qqA,<esc>j0q', and then '8@q'.
That's also probably not the best example, since that's also trivial if you have an editor supporting multiple cursors.
That said, I think multiple cursors are the greatest addition to text editing since, well, vim. I was an enthusiastic user of Sublime Text 1, right before I learned about vim.
Luckily, vim has multiple cursor support, which is... decent. Not great, but decent. I actually think that the only thing that makes multiple cursors more awesome, is being able to combine them with vim's commands, since multiple cursors rely on you being able to do precise things at each cursor, and vim gives you the language for it.
Yeah, it's hard to come up with a decent example because I don't even think about it any more, it's just burned into my fingers. I just know how excruciating it feels to edit without Vim (I even wrote that last comment in Vim and then copied it into the browser because writing it in a text field is miserable).
It basically gives vim-like keys anywhere you want, when holding down the caps lock key. So e.g. caps+h/j/k/l are movement keys.
It goes a lot farther than that: I have a key to delete a line, a key to select a word, a key to go up/down by 4 lines, a key to delete the last word, etc. I even mapped caps+f to jump forward 20 characters (to badly simulate using 'f [letter]' in vim).
This makes so it I never have to leave home position on the keyboard, because while holding caps I have all the movement keys I want at my disposal. It's a total lifesaver, and the only way I can effectively use a keyboard outside of vim :)
<c-v>8j$A,<esc>
I've seen many Vim/Emacs-only devs being easily beaten in speed by others using modern IDEs that have tooling and functionality (auto formatting, refactoring, templating, etc) focused on actually getting things done.
Moving a little bit faster through the text file to edit characters is rarely the issue and I certainly wouldn't give up IDE features for it.
And I'm not talking about Vim the editor. I'm talking about Vim as an idea. Which is today implemented for pretty much every single [popular] editor and IDE in use. That fact alone is pretty illustrative of awesomeness of the idea of Vim.
And the comment about Emacs is quite disingenuous. I have never been inside a Formula One cockpit, but I can totally get "the job done" with my Camry, because it is totally more efficient for parallel parking.
I have used many IDEs, including InteliJ (which I used for about seven years). Emacs wins solely by its extensibility, no editor/IDE has ever matched its capabilities. For some people that is not an important factor, just like the efficient parallel parking might not be important for Formula One pilots. But when you can automate pretty much anything it feels very empowering and liberating.
Small example: when I'm about the create a new commit, an emacs-lisp script can figure out the current ticket I'm working on (based on currently clocked Org-mode task), generate a branch name based on the ticket description, generate part of the commit message (in the way that it is in compliance with git tool like gommit), I would just have to type the rest of the message.
That is not only more efficient, but also reduces frustration. Compare that with the workflow in any other editor and IDE - most of these steps will have to be manual. Then someone might say: oh there's actually a plugin for Editor X that does some of that stuff for Jira. But what if I join a different company that uses Pivotal Tracker instead? I can extend my emacs-lisp script. Whereas if that's an InteliJ plugin I would either have to go through daunting process of forking and modifying the plugin or submit a feature request and wait for something that may never happen.
- Go-to-definition is `g d` (with the cursor over the function name in question)
- Project-wide file search is `SPC SPC`
- Jumping to a function is `SPC s i` and then just start typing to fuzzy-match the function name
I stick with this setup because it does pretty much everything I could want from IDEA, and does it without churning up most of the resources on my machine. And trust me, I've tried to switch to IDEA, but it's just too big and too slow, and too poor of an editor for me to actually make the switch.
The one related function that I would miss is using vimgrep and the QuickFix list to get a list of the locations where the function name appears which I can use e.g. to do a search and replace in those files (or a subset of them)
What if I told you I read way more code than I edit?
If you really never did anything except read code, then I think vim wouldn't give much of an advantage. Then again, neither would any of the shortcuts that every IDE provides.
you may be more productive editing code in vim but when it comes to engineering text editing is not the most important thing to be solving for
I just know that subjectively, going from speaking a language with powerful concepts, to using the super crude tools available in most editors, just feels sluggish.
I highly doubt it was actually a good investment time wise to master vim, but I salary really enjoy messing with editors so it was partially for fun.
The complexity is usually overstated and a sunk cost besides.
- I can be somewhere, jump around to a bunch of files, copy something, and then jump back to whereever I was last actually editing quickly (the "jump back to last edit" is 2 strokes). - I wrote a custom mode, with tabification and fontification for the query language used by the cms I work with. I expect I could do that in IntelliJ but it would take a lot more effort. - Things like kill-rectangle and the like; some of which may have parallels in IntelliJ, but I don't know then... and decades of experience in emacs means I do know them (or how to find them) there.
I really like IntelliJ, but I also really like emacs. Which to use depends on what I'm doing.
I also really like _make_ and wrap my other commands (mvn, gradle, etc) in Makefiles. I get jokingly mocked a lot at work :)
I do this as well. I put my scripts in a make file. E.g. to ssh into a machine X, I write a Makefile with .PHONY rule "X" so I simply run "make X".
I also like the way your renewal gets cheaper every year, as a kind of loyalty bonus.
I honestly had no idea they were making so much revenue - very pleased to see it though!
A while back JetBrains made Rider more themable (prior to that update, you could only really theme the editor, not the file explorer etc). After that update, it looks great with a dark theme enabled!
I'm on mobile now, but if I remember later I'll mention the name of the theme I'm using - it's pretty close to VS's dark theme.
That's just it for me. The tools have so many good features, remain stable, and keep all those features out of my way until I need them. Like Adam Savage's first order accessibility principle for his workshops, the tool I need from their IDEs is always often in arms reach, yet still out of the way.
It's based on 3 window panes: you get your code on the left, theirs on the right, and the result in the middle, along with tools for accepting changes from either side, and you can edit the result pane at will too.
I've never liked the git merge UIs in VS or VS code (or other IDEs I've tried over the years) - Rider's is perfect for me.
Personally, I prefer vscode/atom style merging, but to each his own
I always found the Edit and Continue function in Visual Studio to be flakey: often it would say it wasn't "allowed" with the type of project I was using, or it took minutes to apply changes when it did work, or it crashed while applying changes... Rider got Edit and Continue 1-2 years ago (I think), and it's been completely reliable in that time!
JetBrains announces that IntelliJ is going Open Source.
I checkout the code, build it (~20ish minutes) inside of the downloaded bits, launch the newly compiled IDE, build the repo source again this time debugging IntelliJ from within itself. All in less than an hour, with no hiccups. Amazing.
I love JetBrains so damn much it hurts.
Also if you take part in a prerelease EAP (before the product is 1.0) (and your feedback is constructive and valuable enough) Jetbrains might gift you a license. That's how I got my first year of PyCharm (well my first year after the EAP ended anyway).
Also if you transpile all your js into a big minified file, exclude that too, right click on it and there must be an option as mark as plain text.
The development process should not be tied to the IDE!
Firefox is absolutely the best browser IMO. Unlike Eclipse, which is acceptable but doesn’t hold a candle to JetBrains.
That's a nice way of saying it's a trash fire.
There are community versions of some of their IDEs, which are fully FLOSS, unlike Photoshop.
I remember there was a time when IBM really invested in it that it was miles better than anything in the Java editing space, or honestly any competitor at the time. For example, instantaneous compiling and error checking was so far ahead it took MS many years to catch up with it.
It used to be a matter of huge excitement when a new language or framework was supported by Eclipse.
But I moved away from Eclipse for a few years (I wasn’t programming then) and I’m back and suddenly it’s absolute terrible. Not just worse than browsers of the time, but possibly worse than what it was when I last used it.
This isn't an issue with the language so much as it is with the build systems, includes, library paths, and tooling.
Compare to Java (set the classpath to exactly the artifacts you want and IntelliJ uses its own build system unless you explicitly tell it otherwise), Python (set up a virtualenv with exactly what you want and PyCharm handles its own builds, insofar as Python builds).
In C on Windows and Unix you have to figure out how to find and address includes and libraries in a cross platform way. CMake is probably the best bet. If they were in the position of, say, TurboC back in DOS days where they provided stdlib and you added everything else you wanted, they could do as good a job as for Java.
Exactly. #include "header.h". But where is it? Depends on compiler flags. Java has the same problem, but luckily has only ~2 build systems that are well-supported by IDEs.
VSCode with CMake, clangd, and C++ code filter seems to work for the little I've used it so far.
Havn't quite taken the leap of faith in VIM, clangd to see if that is suitable.
I guess you can switch all of that off of course, but that's what makes it an IDE and not an editor. I think the memory tradeoff is perfectly fine for this kind of application.
Anyway, I don't really see the point you trying to argue here. It appears you are principally opposed to software that uses more resources than what you consider necessary, which is fine and in the case of something like CLion or VS code justified to some extent, because they are far from lightweight. But if you remove these ideological objections from the discussion, none of this actually matters all that much considering typical workstation hardware. I honestly don't mind trading 4 or 8 of the 16 GB of RAM in my laptop for all the features I get from CLion compared to the alternatives. A full compile and run of the thing I'm building can easily get by with (much) less than 2 GB. And if I would be developing something that uses up 8+ GB by itself, I probably want a beefier workstation with 32 GB or more anyway. It's just not an issue, RAM is plentiful, and it's in there to be used.
A single ordinary, commodity CPU core can sort 100M of totally random ints in 5 seconds. The amount of work any editor legitimately needs to do to provide you editing support is trivial, in comparison, by any measure.
cmake_minimum_required(VERSION 3.0)
project(prog)
include_directories(.)
file(GLOB_RECURSE sources *.h *.cpp *.c *.hpp)
add_executable(prog main.c)