I get a little offended by people who think stackoverflow is a character flaw. I go there because it will tell me things like if you get the arguments to ‘dd’ wrong you’ll zero out the drive you were trying to back up. That the function breaks if you pass in null for the second parameter. Or that it simply doesn’t work properly.
To expand on this, people needlessly cargo-cult arcane dd commands, but on many (most?) systems, "cp foo.iso /dev/sdb" will do the same thing, but with possibly even better performance!
That’s a restore not a backup. I’m not actually defending dd. It’s bonkers how long it was considered acceptable to offer that as a solution for anything except zeroing out a disk. I shed a single proverbial tear and contemplated the sad state of technology every time I used it, and felt dirty for having done so.
It is The Worst.
It’s just the most egregious example. The difference in excess argument processing between commands is rather broad example. Is the last arg special or the first one? Grep and tar are in the majority, but cp works differently. They only make sense if you think of the power user.
What I meant was that using dd to take format/backup disks is just way too dangerous for human beings. The prospect of losing your precious data even once in your lifetime is too expensive a cost for the questionable benefit of feeling cool about doing backups by running a command on the terminal. Just use a gui tool for writing to physical disks. The additional visual feedback that a gui can provide is absolutely essential for human beings performing such a dangerous operation.
At this point recommending dd really is kind of telling the other person to go fuck themselves. It doesn't even follow the arg format of every other unix command. I had to double check to see if anyone fixed that in the interim. Nope, still a=b syntax, rather than -i device1 -o device2.
That really should be a giant clue that it shouldn't be used by anyone, for any reason. "It could be that the purpose of your life is only to serve as a warning to others."
What would you recommend as an alternative for disk imaging in Linux? I'm in the process of migrating a herd of PCs from HDDs to SSDs and would appreciate suggestions.
I'd say the easiest way to remember the argument order is that it's conceptually the same as for mv and cp: ln -s x y is the closest possible symlink-analogue of cp x y.
Ah, I remember it is the non-intuitive one, but then when I use it a while, I keep double-guessing which one is the intuitive one.
Kind of like when my wife tells me I'm doing something wrong and she wants it to be the other way.. I know she thinks this thing is important, but can't work out which way she wants it done.
I remember this as "ln=long -s=short $long $short" as in "ln -s $long $short" creates a $short file pointing to $longfile. But to your point, $short is defaulted to basename($long).
i've always considered manpages to be sort of part of the unix command line experience. since they're fast, it's no matter.
but yea, it took me many years to wire down ln's target link_name semantics for some reason.
what helped for me was to reason about the single argument form. ln -s ~/opt/junk/bin/thinger creates a link to thinger in the current directory. this single argument form is easy to remember, if you want to create a link of course you have to specify the name of the target (from which the link name will be inferred) and since it's the only argument it has to be the first one. now if you want to give it a different name, put the name in obvious still open place, the second argument.
Funny, I always reason about the 3+ argument form... For example, what does ln -s foo bar baz do? Well it can't create a foo symlink that points to bar AND baz, but it could create multiple symlinks - bar and baz - that point to the file foo. Therefore the first arg is the file you want to point to, and the rest of the arguments are symlinks you want to create (of which there happens to usually only be one).
Edit: this line of reasoning works for common usage of a lot of other utils too, like zip/tar etc. Even grep - is it grep FILE PATTERN [PATTERN ...] or grep PATTERN FILE [FILE ...] ?
I learned to behave as if there is no second argument for ln. Instead of doing `ln -s path/to/foo bar` I do `ln -s path/to/foo` then `mv foo bar`. Of course it doesn't always work, but covers most use cases for me.
I generally only need `ln -s <src> <dst>`. I know the -s means symbolic link, but in my head I read it as "source", since that's how I remembered the order long ago.
This is confusing, because when you picture the link as an arrow (as in `ls` output), the link name is the source and the link target is the destination.
I think what a lot of people don’t get about the Unix command line is that learning to use the tool is part of the experience. Sure I forget the precise flags to get various tools doing what I need, but browsing the man page and rediscovering the breadth of the tool is half the beauty.
The only feeling I've gotten looking at a man page is, "well, I don't have to time to dig through all of this. Guess I'll just look at stack overflow which will have my exact use case and required parameters."
I remember the times before I learned Bash and the Unix userland. Those where dark times. I was stuck on Windows 98, which I really didn’t like. It just felt so needlessly crippled. When I discovered Bash, Debian, the Unix way, it felt like a breath of fresh air.
These days I’m on macOS. And one of the best things about it is that it’s a great desktop operating system, with a Unix under the hood. The userland Apple ships is quite dated, but that’s easily solvable by installing GNU Coreutils etc via Homebrew.
nah. learning unix was fun. having all the documentation online and in an easily called up and consistent format made it possible. microcomputer operating systems didn't have things like "man -k" and while we can complain all day here till we're blue in the face about small quirks, it was light-years beyond the crap that was going on in microcomputer operating systems.
even when doing windows or embedded development with tools like visual studio or wrs tornado, i've always insisted on having mkl or cygwin for command line tasks.
This is why imo powershell (when done right) is more powerful than bash (no matter how correctly it’s done). Powershell cmdlet names are descriptive and generally make sense as do the parameter names, additionally the guidelines for how to make powershell commands forces a certain style that once you grok it makes sense for most powershell commands.
If bash has something like that I’m still missing it.
Obviously the brevity of bash can be its own power.
I do feel like we are all a little beholden to bash and it is tiring. While I will put simple work into simple bash scripts I always use shellcheck and do ‘set -euo pipefail’. For the life of me I can’t remember what that is or why I need it, I just know that I do. (I’d be comfortable looking it up if I actually cared enough to.)
Anything more than trivial I do in a high level programming language like Go (previously I would use Ruby or Perl).
From everything I’ve seen PS is a nice direction to go in.
I wish there was a sudden upheaval and everybody used a shell that wasn’t so legacy-cruft-laden.
PowerShell’s learning curve was too steep for me (or maybe I didn’t try hard enough, idk). I used it for a couple of years, but once WSL turned up I went back to my trusty old Linux skills.
My biggest problem with regex is each language/program/whatever has their own take on it. You don't have to learn it over and over, you have to learn multiple slightly different versions of it over and over.
I don't mind the slightly different versions, that's what documentation and SO is for. I mind the language being a garbage kludge of organically-grown syntax and semantics, with not the slightest hint of a single principle or a unifying idea. And I mind that the vast majority of interfaces to it from a general purpose PL is fucking around with strings like a caveman, instead of a typed, IDE-assisted, first-class representation as a full-blown language construct or mix of constructs
been recently learning Python in earnest. Coming from Perl, Ruby, PHP, JavaScript, etc. I'm just dumbfounded. How on earth did Python's regex library get to be so bad? It's like it was designed by people that were figuring out how regexes worked as they went along but must have skipped the part about modifiers, how anchoring works, etc. No wonder so many people hate regexes. The ironic part is Python's "one obvious way to do it" zen thing. It has like 5 different ways to do the same thing with a regex whereas Perl, the "more than one way to do it" camp has simple and obvious ways to do everything with regex. I love regex in Perl. In Python I don't want to touch it ever again. Even JavaScript is miles better.
I agree. I actually have a pretty strong understanding of regexes and use them regularly (for editing, not so much in code), but I despise how every editor/language does them subtly different.
I’ve never understood that one. I realize it’s just a joke, but xvf/cfv (extract/compress, view, uh,. Ok I did have to look up f, which I think is file) and the z for gzip is one of the few commands I have no trouble remembering.
Though if you include the dash to use those as flags, the second one is wrong (it treats the "v" as the file name for "f"). Without the dash, the next positional argument is the file name.
Examples: These do the same thing:
tar cfv foo.tar a b c
tar cvf foo.tar a b c
tar -cvf foo.tar a b c
Reminds me of a peeve I had with the Apple docs when they described Bash like riding a bike (but in a different metaphor - it's good to get around somewhere easily but won't take you as far as an application language)