Hacker News new | past | comments | ask | show | jobs | submit login
Unix Tricks (cfenollosa.com)
361 points by dedalus on July 3, 2014 | hide | past | favorite | 162 comments

Missing trick:

In bash, "ESC" then "." fetches the last parameter of the previous command. It's invaluable (same as typing "!$" but you get to see and edit it)

You can get any parameter of the previous command by using "Esc, <number>, Esc, .". If you repeat pressing "Esc, ." you get the n-th parameter from the command before the previous one, etc.

More details in this Stackoverflow answer: http://stackoverflow.com/a/4010170/578588

It's actually Meta, emulated by an ESC character by the terminal. OSX tip: check "use Option as Meta" in Terminal's preferences, and you can keep Option pressed while you mash ".".

Which is fine if you never need to generate characters using the Option key. For example on a German keyboard you need the Option key to write { or }.

I'm on a french layout so I know :-)

It also has the side effect of "resurrecting" dead keys (which are unrelated to Option), making input of 'ê'and 'ï' impossible.

An alternative is using iTerm2 which has a similar setting, liberally applyable to either Option key.

You can also use the following options to see and edit all expansions before running a command:

    shopt -s histverify histreedit

I prefer `alt+.`, which does the same thing without having to reach for the ESC key.

I use `!:p` to the same effect. Which has the benefit that you can insert a history number between the bang! and the colon: ( set PS1 to include the command number if you do this a lot ).

There are some great tips/tricks in this document. Too bad it's not in Markdown or something easier on the eyes!

Markdown-ified version: http://www.notehub.org/kywfm

Editing password is hackernews if anyone wants to tune up the markdown.

Ctrl+r is fine when you're looking for something recent, but this one is useful when you're looking through a long list of similar commands.

    alias hist='history  | grep'
Use it like this:

    $ hist git
      9543  git add toto
      9544  git commit
      9545  git log
      9546  git log
      9548  git add -A
      9549  git commit
      9550  git log
      9633  git pull
      9955  cd dev/git
      9957  git grep copyof
      9958  git grep copyof
      9959  git pull
By the way, do you know you can call back the 9633rd command in the history with "!9633"?

I use this all the time. In fact it was one of the first aliases i made when i switched to a UNIX system. Though I named mine `greph` (for grep history). Absolutely invaluable

Wow, this is great. All too commonly I do cat ~/.bash_history|grep, but this is way easier. Thanks!

Its implied on the list but not mentioned specifically.

will be substituted with the last command you typed. So if your somebody like me who frequently types

     $ apt-get update
     Could not lock yada yada are you root?
     $ sudo apt-get update
instead you can type

     $ apt-get update
     $ sudo !!

It always feels more logical to me to just type "Ctrl-P Ctrl-A sudo" (previous history line, beginning of line, add missing sudo) than typing "sudo !!" or "sudo !!<tab>" if you need to edit the substitution.

Same thing for the last argument, it's easily edited if you know the right shortcuts: Ctrl-P Alt-B Ctrl-U. (Works for bash, for zsh you need http://stackoverflow.com/q/3483604/414272). That said, I didn't know about Alt-.

What I love about this is that if you up-arrow through your past commands, the "sudo !!" shows up as "sudo apt-get update"

Bash substitutions are done prior to execution so your history stays intact. Older open source projects are just amazing for how complete their feature set is.

is this how people feel about religious texts?

Originally I was gonna write:

>Religious texts don't change over time due to community suggested changes.

Then I caught myself.

You mean my bash fanaticism isn't religious?

I named my cat sudo bangbang

    man hier
is pretty cool. It explains the root directory structure of the system.

Which was updated just yesterday[1].

[1]: https://plus.google.com/+LennartPoetteringTheOneAndOnly/post...

international pun, hier being french for yesterday.

The SEE ALSO for 'man hier' alerted me to 'apropos'. Unbelievable, so much emacs goodness in bash.

"apropos Regular Expressions" brings up wild stuff: re_format(7) -- posix regex docs at last!, treereg(1) (huh????).

Some serious lore in here ...

TIL. As a long-time DOS user that often gets a bit lost on Unix systems I cannot stress how awesome it is to have found out about this!

Wow, DOS users still exist? Which one? FreeDOS?

Back in the late 90s when I first moved to Debian from Windows, growing up with a DOS CLI made the transition far easier to me (and having some FTP exposure didn't hurt either).

There's no good way to refer to the entire DOS/Windows command shell lineage without confusion, but that's what I meant. I don't technically use DOS at all anymore. I think "that system" is easier to understand in it's entirety and find your way around than Unix, although it has been getting progressively more complex for years. I actually use OS X most of the time now (plus Linux on servers and Windows for 3D stuff).

> I think [DOS] is easier to understand in it's entirety and find your way around than Unix

Yikes, I recently had to do a bit of development on a windows box and I found the command-line tools (not to mention CMD itself, which seems to have stopped development in 1993) to be absolutely awful. I could barely survive without Cygwin, git bash, etc giving me some semblance of a functional shell setup. I guess it's different strokes for different folks.

I agree: awful tools and batch language is horrible (and largely non-portable between different versions). Getting complicated stuff to work can be tricky due to the lack of essential commands. For instance, there's no way to reliably get an ISO formatted datestamp without a 3rd party utility (although I think PowerShell supports this). Lots of odd little things like that simply don't work.

But the filesystem layout and set of built-in commands is pretty easy to understand fully. It's generally pretty easy to find things. In that sense Unix is more complex (but also more sensible, usually).

> But the filesystem layout and set of built-in commands is pretty easy to understand fully. It's generally pretty easy to find things.

On the face of it, the basic directories are straightforward (Programs -> \Progra~1, User files -> \Docume~1, System files -> \Window~1, Very System files -> \Window~1\System32), but I don't recall having an easy time finding things (speaking from memories of my XP days).

Where would you look for a config file of some application? Maybe it'll be in \Progra~1. Or $user\LocalSettings. Or ApplicationData. Maybe in the Registry? (Don't get me started about the registry!)

Where would you look for log files? Does the Event Viewer show everything nowadays, or do you still have to hunt for logfiles in the same fashion as config files?

On *nix, I know that my configs are under /etc/ (global) and ~/.{program name}/ (user-specific), and my logs are under /var/log/. I don't know what every single directory is, but I know where to look for things when I need to.

What really gets me is that MS almost fixed that with powershell -- and then made the defaults policies so restrictive that you need to go through hoops to actually be able to use the thing. I do get that it's "really powerful and lets not create another set of visual basic script viruses" -- but the UI is just horrible. Maybe it's better in 8, but last I checked I think you had to go into the policy editor and choose between "safe - no powershell for you", "safe-ish - only signed powershell (still no powershell for you) and: open season on shooting your feet off: powershell always, everywhere, all the time.

Don't get me wrong, I still prefer living with bash and a full GNU system -- it's just sad about what MS did (or didn't) do with their command line tools since 2003ish and on.

Backwards compatibility. PowerShell is modern and powerful.


> There's no good way to refer to the entire DOS/Windows command shell lineage without confusion

cmd.exe? Powershell?

Nice to see something I've gradually learned about by gut feel summarized.

Why do some distributions use /opt a lot while others don't seem to use it at all?

Short answer would be historical reasons.


"Contains locally installed software. Originated in System V, which has a package manager that installs software to this directory (one subdirectory per package)."

Solaris made extensive use of /opt, that flowed into some Linux packagers. It then fell out of favor as everyone put everything in /usr

definitely feel like this one was more widely known in the days of printed manuals.

You know what I hate? My history isn't aggregated in real time across all my Mac OS X terminal windows.

You know what else I hate? Typing in long commands in the Mac OS X terminal and then them wrapping weirdly. Especially when I hit the up arrow to go back in my terminal history.

Here you go, just throw it in your .bashrc:

  shopt -s histappend
  export HISTSIZE=100000
  export HISTFILESIZE=100000
  export HISTCONTROL=ignoredups:erasedups
  export PROMPT_COMMAND="history -a;history -c;history -r;$PROMPT_COMMAND"
I wish I could give credit to where I originally found this but it was ages ago.

I've been using `history -a`, but what do `history -c` and `history -r` do?

It looks like it clears your shell's history, and then re-populates it from the .bash_history file?

   history -a  # append history lines from this session to the history file.
   #History file may contain history from other terminals not in this one so:

   history -c # clear [in-memory] history list deleting all of the entries.
   history -r # read the history file and append the contents to the history list instead.
I've heard that -n can be problematic which is why -c then -r is used.

Many thanks for this! This has been a thorn in my side for a while but never got round to doing anything about it...

Don't you want HISTCONTROL=ignoreboth ?

    > You know what else I hate? Typing in long commands in the Mac OS X terminal and then them wrapping weirdly.
Yeah, keeping ctrl pressed in and pressing x followed by e (CTRL + x e) will open up the current line in $EDITOR and when you edit and save, replaces the current line with what you entered in your editor. Really a killer feature.

In Linux with bash (and probably other shells that are bash-compatible), you can use vi to edit any of the commands in your command history, and then execute the edited version.

Do this at the command prompt, or once in your ~/.bash_profile to make it permanent:

set -o vi

After that, you can search for any of the commands in your history, edit it, and then execute the edited command, by doing this:

At the prompt, type Esc once to get into vi command mode. Then you can press the k key repeatedly to scroll up through the command history, or (often easier) use the ? (search backward) vi command to search for a pattern to find a specific command. Once found, press v to edit it in a temp file. Then when you save and quit, the edited command gets executed.

The same technique works with emacs as the editor instead of vi, if you don't give the 'set -o vi' command, because the default editor for command line history is emacs. Also, if you have run 'set -o vi', you can switch the editor for commands back to emacs with 'set -o emacs'.

Or emacs.

The 'set -o <editor>' bit sets the readline editing environment to be similar to vi. It can be set to emacs.

C-xC-e (edit-and-execute-command) invokes the editor specified by $VISUAL, $EDITOR, or emacs, in that order. You could set it to scrivner if you wanted to (though I'm not sure that would necessarily work on exit).

I've tested with VISUAL set to nedit, from which I then changed it to uptime. Now I get loadavg when I want to edit my command line ;-)

If you like that, you can put:

set editing-mode vi

in your ~/.inputrc (instead) and all GNU libreadline using programs will give you vi keys instead of emacs :-)

Cool :)

Yes, that is also a solution. However, you cannot use vim so if you're used to vim, it might feel limited without certain commands. You're also missing eventual plugins.

It's better than nothing though.

I don't think this is the case, though I may be misunderstanding what you wrote.

At the shell, it's true that you're limited to vi-style controls (so no text objects like ci").

However, once you press 'v' in normal mode you can drop into either vi or vim - this only depends on what $EDITOR is set to[1].

[1] http://blog.sanctum.geek.nz/vi-mode-in-bash/

Yes works with ksh, been using it for decades, very handy.

> My history isn't aggregated in real time across all my Mac OS X terminal windows.

zsh has `setopt inc_append_history` and `setopt share_history`. i am sure bash has something similar.

>You know what else I hate? Typing in long commands in the Mac OS X terminal and then them wrapping weirdly.

Are you escaping your color codes in PS1 correctly?


    PS1="\033[1;32m \w \033[m $"
Good (notice the extra \[ and \]):

    PS1="\[\033[1;32m\] \w \[\033[m\] $"

Mine seems escaped as hell, but still gets wrapped. Anything I'm missing?

edit: couldn't get UTF-8 branch symbol to be properly displayed here, but this is how it looks:


  function fancyPrompt {

   local bgBlue="\[\033[48;5;31m\]"
   local fgBlue="\[\033[38;5;31m\]"

   local fgWhite="\[\033[38;5;231m\]"

   local bgDarkBlue="\[\033[48;5;24m\]"
   local fgDarkBlue="\[\033[38;5;24m\]"

   local bgDarkGray="\[\033[48;5;237m\]"
   local bgLightGray="\[\033[48;5;245m\]"
   local fgLightGray="\[\033[38;5;245m\]"
   local colorClear="\[\033[0m"

   local branch
   local branch_symbol="\[\] "

   if branch=$( { git symbolic-ref --quiet HEAD || git rev-parse --short HEAD; } 2>/dev/null ); then
   	export PS1="${bgBlue}${fgWhite}\h${colorClear}${fgBlue}${bgDarkBlue}\[\] ${fgWhite}\w${bgLightGray}${fgDarkBlue}\[\] ${fgWhite}${branch_symbol}${branch}${fgLightGray}${bgDarkGray}\[\] ${colorClear}"
   	export PS1="${bgBlue}${fgWhite}\h${colorClear}${fgBlue}${bgDarkBlue}\[\] ${fgWhite}\w${bgDarkGray}${fgDarkBlue}\[\] ${colorClear}"


That's quite pretty. How do you get the triangles?

I've been loving a prompt which color codes git branches (which, if I understand right, would be built-in if I used zsh instead of bash) [0], though I have to edit the last line in order to get my history appendation working as well.

0: https://gist.github.com/tobiassjosten/828432

Try not escaping around the triangles and branch symbols.

This. Right Here. For the longest time I could not figure out why my Ctrl-r (history) was so messed up.

> You know what else I hate? Typing in long commands in the Mac OS X terminal and then them wrapping weirdly.

I solved this by adding the following line in my ~/.bashrc

  shopt -s checkwinsize
Never tried on Mac OS X though, so I don't know if it works.

> Typing in long commands in the Mac OS X terminal and then them wrapping weirdly.

For all people who dont use Mac OS X:

This happens because your PS1 is wrongly set and bash cant calculate correctly the length left of your line. Try it out, by going back to default with no colors and crap and see how long it goes.

For mac os x users. The above wont help, dont even try it.

I found the solution to bash problems was to move to zsh. I did that almost 10 years ago and haven't regretted the decision.

I switched to zsh for a few hours once, but it broke TRAMP in Emacs. There is probably a work around...

Can you suggest an intro guide to zsh for former bashers?

I prefer to keep mine separate, but I always use specific terminals / screen sessions for (mostly) the same type of work (admin, specific projects, web dev, etc.) so each have their own history, as well as keeping a per-directory record.

So far over 130000 command lines (with timestamp & cwd) for past 2.5 years, just on my laptop.

You know what I hate? My history isn't aggregated in real time across all my Mac OS X terminal windows.

I used to feel this way. Then one day I figured out how to enable this in Bash. Resulted in a confusing mess. Turns out you most likely want terminal sessions to be distinct until you end them.

You should just move to zsh that does this natively and in a clean manner.

It's amazing how many years I've been using bash and I still have things to discover.

Ctrl-x Ctrl-e nice tip!

I still prefer "sh" on BSD (particularly OpenBSD). It doesn't have as many knobs to turn nor as many surprises.

sh, the Bourne shell mode of pdksh? Isn't that a little too limited? pdksh (/bin/ksh) though is great, yes.

I'm in general a big fan of the OpenBSD userland: straightforward and very, very well-documented. Real manpages, no religious "The full documentation for xyz is maintained as a TeXinfo manual" crap.

A well-conceived featureset and good docs are key to properly learning a platform. OpenBSD taught me a lot about Unix. If you later realize that Bash has some feature that you actually need, you can still install it.

Also worth noting that native borne shell was always slightly slower than korn shell. Though been a long time since I checked that one out, though did on few systems and found it to be so. But there again csh was also about then, distant memory in for many that one.

Also anything that will run script wise on bourne will run as is under korn.

Though you may well find that it is historical, which one a user picks shell wise. People who started or worked a lot with Sun systems will be csh fans. Old vets more inclided to bourne, though very few. As for korn shell, that would be mostly down to AIX systems as that is the default upon them as with the other systems mentions, default wise. Then for Linux you will find a bias towards bash.

Least that is what I have observed.

Though history does give us some intersting trends and the awk, perl, python transition and preference will also mostly be down to when somebody got into unix as a whole. Again old school, awk. Old, perl and not so long ago the python brigade.

But that is just a rule in thumb and more helpful in explaining why there is a solo perl script when everything else done in xyzzy type encounters.

As for OpenBSD, good choice, I prefer it due to the file system layout and more akin to old school unix unlike Linux which is `creative` more than not in choices, so feels less at home.

Still back in the early days we had AT&T and Berkly BSD flavours and with that the ps command, oh the fun and games.

But least thinks a little bit more common across flavours than before and yet still each has there own quirks.

I think shell performance stopped mattering the moment I dumped my 50MHz SPARC LX for a Sun Ultra 2.

Very true, but in production, every nanosecond counts sometimes.

Yes that's the one. It's perfectly fine. I found that if you have to push it past the limit it's probably the wrong tool for the job and it's time to use Python or Perl or pick something off the shelf and think about more structure to whatever you're doing.

Shell scripts really don't scale development-wise over time.

Definitely agree with the rest of the points though.

In bash fc will open your $EDITOR with the last command you just typed. Just like Ctrl-x Ctrl-e but you don't need a command in place to edit.

`man readline` and `man bash` alike blow my mind every time I pop them open.

> - 'sort | uniq' to check for duplicate lines

sort -u

A personal favourite of mine is aliasing common cd operations.

    alias ..="cd .."
    alias ...="cd ../.."
    # etc
and of course `cd -` to jump to previous directory

> In bash, 'ctrl-r' searches your command history

No! In emacs ctrl-r searches your command history. It happens that bash use emacs mode by default. All other emacs basic commands work too: ctrl-p, ctrl-n, ctrl-f etc...

If you are a vim user, and especially if you love venting how much you hate emacs, then please add this to your bashrc:

set -o vi

and now you can use vim command instead of emacs.

'ctrl-r' is reverse-i-search in vi mode as well.

'ctrl-r' is bound to isearch-backward in my Emacs.

'meta-r' is bound to comint-history-isearch-backward-regexp

I actually bookmarked a similar page a little while ago, super useful: http://www.reddit.com/r/linux/comments/21rm3o/what_is_a_usef...

Instead of

    ssh -R 12345:localhost:22 server.com "sleep 1000; exit"

    ssh -R 12345:localhost:22 -n server.com
The -n flag tells ssh to only make a connection, without ever running a shell. This means it even works when you don't have shell access (say, with a command= entry in .authorized_keys).

Also, I cannot recommend envoy enough in lieu of ssh-agent, if you're not using a GUI ssh agent already (e.g OSX keychain or GNOME keyring)

Last, I guarantee you don't want "$@" in those aliases, but either "$*" or $@ (certainly the latter).

I think you meant -N, not -n. From the man page[1]:

     -N      Do not execute a remote command.  This is useful for just for‐
             warding ports (protocol version 2 only).

     -n      Redirects stdin from /dev/null (actually, prevents reading from
             stdin).  This must be used when ssh is run in the background.
[1]: http://www.openbsd.org/cgi-bin/man.cgi?query=ssh&sektion=1

I usually visit these threads knowing that I will pick up something useful to add to the toolbox... thank you for 'envoy' - I think it will streamline my ssh key management.

I wonder if I can convince the author of this to put it up in a Git repo so everyone can contribute. It probably needs the tiny URLs removed (IME, they're more likely to break than the original site)

You might want to try out the more modern fish shell as well: http://fishshell.com/ It makes history searching even easier.

I love fish and have happily been using it for years. Every time it comes up here, though, someone inevitably will complain about compatibility - and I will admit that RVM, for instance, has definitely caused me problems with fish in the past. I guess it depends quite a lot on your particular usage and requirements.

I never got RVM working with fish; unsurprising, seeing as it’s 20k lines of bash. rbenv works well though (with one additional conf line), and chruby was working on support last time I checked.

There's a wrapper to get RVM working with fish, in case it matters to you now; details here: https://rvm.io/integration/fish

Doesn’t matter now :) rbenv does everything I need in a lighter package.

RVM is pretty heavyweight. Why not something like chruby?

I started using RVM when it was the only (possibly well-known?) game in town, and never switched until I started using fish and found that it didn’t work.

I like the autocompletion of fish, isn't there a way to get this in bash?

I found Ctrl-O (execute current command from history and edit the next one) to be invaluable.

Ctrl-V: type next input literally (e.g. Ctrl-V [Tab] if you don't want [Tab] to autocomplete).

Alt-#: Prefix current line with "#" (don't execute it) and put it into history for later use.

I'm sometimes surprised how many people don't know these: Ctrl-U/K: kill line before/after cursor.

Pipe to clipboard with following alias

alias clip='xsel -i --clipboard'

Get tree view of directories

alias lst='tree -L 2 $1'

Also I use za to go up in directory tree

bash https://gist.github.com/chanux/1119556 fish https://gist.github.com/chanux/9411092

"- Compile your own version of 'screen' from the git sources. Most versions have a slow scrolling on a vertical split or even no vertical split at all"

Or you could just use tmux which afaict is superior to screen in almost every way. http://tmux.sourceforge.net/

I suggest bashmarks http://www.huyng.com/projects/bashmarks/ to bookmark directories and jump there with a quick command. I also like to alias ..="cd .." and alias ...="cd ../.." to quickly navigate up.

Ctrl-R is by far the most useful one for me.

I like this in my ~/.inputrc

Type a few characters and hitting the arrow keys will show only commands that match those first few characters.

The article doesn't say, so it's worth mentioning that you can type ctrl-R multiple times to search further back into command history.

It doesn't mention either that you can search to the other direction by pressing ctrl-s (usually it also sends the stop signal which should be removed by "stty -ixon"). Very useful if you go past the entry you are looking for.

I use C-r constantly. But sometimes I need to find a command, and then edit it before running it. What is the proper way to exit "search mode" and go into the normal "edit" mode?

Any movement key will do the trick. I usually use cursor-right or the end key.

It's so ingrained I had to actually do it in a shell to know what it was I did.

ISTR pressing one of those keys would obliterate the part of my line where I was writing, putting in ^B characters.

I can't replicate it now on my test box; it works exactly the way you say it should. It may be a terminal issue; I'm currently on Windows using cygwin to ssh to Linux.

C-g (as in quit https://www.gnu.org/software/emacs/manual/html_node/emacs/Qu...) is the 'proper' way but yeah, not the only one.

That just wipes out the whole line (on my test box).

C-e ?

Ctrl-P autocompletes from your history too. It works like Ctrl-R but after you've already started typing, if you see what I mean.

This might be a zsh thing though.

Ctrl-p in emacs takes you to the previous line. Since readline uses emcs-style notation by default, Ctrl-p takes you to the previous command in history (and Ctrl-n to the next). I don't know about zsh though.

Also, other emacs keys work like back: ctrl-b, meta-b, forward: ctrl-f, meta-f, start of line: ctrl-a, end of line: ctrl-e etc.

I was going to write a shell package/snippet manager until I discovered this:


It would be nice to get all of these .bash_profile enhancements in as packages.

I suppose it could be improved to prompt for a brew/apt install if needed.

A really handy one a colleague showed me yesterday was the /dev/fd/0|1|2 files, which are stdin/out/err respectively. Means you can use that file for utils that expect a file only.

E.g echo "This would be contents of file" | someCommand /dev/fd/0

These are also available as /dev/std{in,out,err}; the names of which are a little more self documenting :)

    $ ls -l /dev/fd /dev/std* | column -t
    lrwxrwxrwx  1  root  root  13  Jun  27  16:32  /dev/fd      ->  /proc/self/fd
    lrwxrwxrwx  1  root  root  15  Jun  27  16:32  /dev/stderr  ->  /proc/self/fd/2
    lrwxrwxrwx  1  root  root  15  Jun  27  16:32  /dev/stdin   ->  /proc/self/fd/0
    lrwxrwxrwx  1  root  root  15  Jun  27  16:32  /dev/stdout  ->  /proc/self/fd/1
You can even access the file descriptors of other processes with /proc/${pid}/fd/; which is handy for (say) un-deleting a file that a process still has open.

Another handy one is "process substitution" which lets you do the same thing with multiple streams by creating new file descriptors, and automatically turning them into /dev/fd/* paths:

    someCommand <(echo "contents of file 1") <(echo "contents of file 2")
which is somewhat explained by:

    $ echo someCommand <(echo "contents of file 1") <(echo "contents of file 2")
    someCommand /dev/fd/63 /dev/fd/62

"process substitution" is one of the new useful things I learned this year (switched to bash from csh over 20 years ago).

/proc/${pid}/fd/ is very useful for forensic purposes when you find some malware running that deleted itself and/or config files but still has a handle open to them.

Now and then I reread a chapter of "Classic Shell Scripting" (http://shop.oreilly.com/product/9780596005955.do) - probably the best book on shell script I know.

Along with the !! function you can reference the last call of a specific command like so:

e.g. sometimes I forget the "ln -s" order of arguments (to which it gives me an error), so I follow it up with:

    ln -s !ln:3 !ln:2

I used to be excited about this, now I found it shallow. Better unix tip => http://www.confreaks.com/videos/615-cascadiaruby2011-the-uni...

I would add:

    * 'tree' instead of 'ls -R'

Which package would that be in?

On Debian/Ubuntu, you can use the `apt-file` package to determine what package a file resides in. For example:

    apt-file search /bin/tree

`brew install tree` on OSX

Also this bash function from Mathias Bynen's famous dotfiles is a fantastic shortcut for colorised (and more) output from tree. https://github.com/mathiasbynens/dotfiles/blob/master/.funct...

tree on ubuntu

Another one:

Use dcfldd instead of dd if you want to see how far your dd operation is progressing...

dcfldd if=[infile] of=[outfile] sizeprobe=if

So. Much. Better.

dcfldd also has many other useful options - read the man pages - and note that you'll have to install it from your linux distro's package repo beforehand.

You can also send SIGUSR1 to dd for progress info. From the manpage:

  Sending a USR1 signal to a running 'dd' process makes it print I/O statistics to standard error and then resume copying.

    $ dd if=/dev/zero of=/dev/null& pid=$!
    $ kill -USR1 $pid; sleep 1; kill $pid

    18335302+0 records in 18335302+0 records out 9387674624 bytes (9.4 GB) copied, 34.6279 seconds, 271 MB/s

Caution: while this works great on Linux, on OSX, a USR1 will kill dd, and a SIGINFO will get you the block counts.

Indeed, looks like GNU coreutils dd responds to SIGUSR1 on linux and SIGINFO on linux. That's good to know. What's the reason for this discrepancy?

OSX inherited SIGINFO from the BSDs, OpenBSD, FreeBSD, Dragonfly, and NetBSD all have it.

Linux doesn't. I don't know why; but Google in the past has told me that patches to add it were rejected about 10 years ago.

On OSX it's SIGINFO (29) in case anybody wondered

Which you can also get by hitting Ctrl-T. This works for FreeBSD too. (Not sure about the other BSDs.)

It works on all of them, including OSX.


Indeed you can :)

But I prefer not to have to manually keep sending that ;)

while ; do kill -s USR1 $(pgrep '^dd$'); sleep 5; done

Just pipe dd to pv

>Don't know where to start? SMB is usually better than NFS for most cases.

Not sure why he feels SMB is better, because in my experience, NFS usually works more smoothly and is easier to setup than SMB. But then, I don't usually use windows.

One to add: Unfreezing terminal after pressing Ctrl+S by accident. Press Ctrl-Q.

Or drop a `stty -ixon` into your .bash_profile to never have that problem again.

One that I needed to use a lot in the old days, and still occasionally need:

If your font is all messed up, echo C-v C-o. C-v puts you into a literal mode, and C-o gets you back into the normal character set. (You probably got there by emitting a C-p at some point.)

I always used 'reset' for that.

Instead of using netcat when tunneling SSH through another machine like this:

    ProxyCommand ssh -T host1 'nc %h %p'
one should use this instead:

    ProxyCommand ssh -W %h:%p host1

The only problem with -W is that it is not available everywhere yet: RHEL6 defaults to OpenSSH 5.3, while -W was introduced in 5.4. Thankfully EL7 comes with OpenSSH 6.4 (and a kernel newer than 2.6.32!).

I have

for cmd in $(compgen -c); do if [[ $cmd =~ ^[0-9a-zA-Z]+$ ]]; then eval "alias $cmd?='man $cmd'"; fi; done

in my .bashrc to alias command?=man command, saves some typing to get to man pages ( which I do very often )

Speaking of man pages, in earlier versions of Unix and Linux, I used to want to redirect the output of many man commands to files for later reading, e.g. if working on C, say I wanted to read 'man ioctl', 'man stdio', 'man signal', etc. But the man output had {n|t}roff formatting characters in it, for print output, which used to mess up vi. So I used to use this script I wrote, called m:

# Put this script in a directory in your PATH.

# m: pipe man output through col and save to file(s).

mkdir -p ~/man # Only creates the dir first time, including all dirs in the path.

for name: # in $* is implied

    man $name | col -bx > ~/man/$name.m

Do a "chmod u+x m" to be able to run it.

Then run it like this:

m ioctl stdio signal # or any other commands


pushd ~/man; view ioctl.m stdio.m signal.m; popd

to read those pages, now stripped of formatting characters.

Wow. I never knew about the `compgen -c` command to show bash command completions. Also from this reference [1] I found these:

    % compgen -b # built-ins
    % compgen -k # keywords
    % compgen -A function # functions
    % compgen -abckA function # everything
[1] http://www.cyberciti.biz/open-source/command-line-hacks/comp...

For zsh:

  for cmd in $(compgen -c); do if [[ $cmd =~ ^[0-9a-zA-Z]+$ ]]; then eval "alias $cmd\?='man $cmd'"    ; fi; done
# had to escape the ? in the alias name

Cool trick :)

zsh has a simple builtin `r` (which is just an alias for `fc -e -`). It lets you edit the previous command with search and replace. I use it for brew and git a lot.

  $ brew info rust
  $ r fo=stal

I notice j.py in that list. I forked that a wee while back, and have been enhancing it ever since:


https://github.com/rupa/z is really nice, too. Pure bash.

stty - sane

Very handy if you do something silly like cat a binary file without piping it into strings -a

This restores the terminal settings and sanity restored. Note you will often find yourself typing without seeing anything, least until you enter and execute the sane.

Also asvise the following usage ctrl+c a couple of times to clear anything out of the input buffer queued up. stty -sane all good from here onwards and good time to play with od and remember the strings -a pipe next time.

I love these, there's too much to process but I always get to learn at least one ridiculously useful thing. Today's was vim's `:set spell`.

Very interesting resource! It's amazing to see how after 10 years of using command line that you can learn new tricks. Every single day!

I call this one sshping.

    while sleep 0.1; do nc -w 1 -z host 22; done
Useful when waiting for systems to boot.

Add this line to (vixie) cron:

    @reboot /bin/date | /usr/bin/mail -s "$(/bin/hostname) booted" root@example.com
If you don't want to configure postfix as a smarthost (which is not that hard and recommended if only to have mail queued for retry), use ssmtp.

      $ until-success ssh ...
If you have my sysadmin-utils installed:


> 'htop' instead of 'top'

meh, I don't like htop. "atop" on the other hand!!!

Send a command to display 1.

`xterm -display :1 -e <command>`

This is useful when running automated tests on top of vnc.

j.py? The original is autojump: https://github.com/joelthelion/autojump/

z is a nice replacement that is purely written in bash.



And for python3:

python3 -m http.server 8080

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact