Start all of your commands with a comma (2009) | Hacker News Reader

Start all of your commands with a comma (2009)

rhodesmill.org

592 pointsTomte121 comments

I do something a bit special... Instead of naming my script foobarnator (and hence risking that another foobarnator may exist in the future and may clash with my script), I name my script foobarnatorToBeAliased.sh. Then in my list of aliases I alias foobarnator to foobarnatorToBeAliased.sh.

The alias itself looks a bit weird in that to pass arguments to an aliased command you need to use a function (in Bash at least AFACIT) but it all works fine.

As aliases aren't sourced when other scripts are running, I'm sure my scripts/commands ain't ever going to clash with anything.

I've been doing that since years.

It also works fine to add "wrappers" to existing commands: for example to add additional input sanitization catching common mistakes I'm making when calling some commands.

What has always been a blocker for me to follow your example, is the desire to use many such scripts from the command lines provided inside an $editor or $email client. There probably is a way to configure both of them to see the aliases, but it's more straightforward to just have everything in the PATH.

Yes, it seems cleaner to just put "export PATH=$HOME/bin:$PATH" in $HOME/.bashrc. That avoids collisions with /usr/bin and /usr/local/bin. (Or rather, resolves collisions in favor of your personal bin directory.)

Your PATH is inherited by child processes that you start, including scripts that expect the system version of commands that have the same name as your commands.

Porque no los dos? It seems that

    function ,foo() { printf "Fooo\n"; }
    ,foo

actually works! (at least in my bash 5.0.17 shell)

I know this is unimportant, but the phrase is "¿Por qué no los dos?".

Thank you. In case others would wonder too: https://www.rae.es/espanol-al-dia/porque-porque-por-que-por-...

> The alias itself looks a bit weird in that to pass arguments to an aliased command you need to use a function (in Bash at least AFACIT) but it all works fine.

No you don't:

    $ alias say=echo
    $ say beep
    beep

You do need to do something special the other way around (if you wrap a command in a function it needs to explicitly pass the arguments on).

Another advantage of aliases: in zsh at least, autocompletions of the aliased command Just Work (it breaks in wrapper functions)

I used to do this, but it doesn't work with things like "watch".

I actually just learned from a different HN thread yesterday [1] that aliases do get expanded if you add this to your shell's rc file:

    alias watch='watch '

Specifically, if an alias ends with a space, it will cause the subsequent arg to be expanded as well. For example, try running:

    alias beep=echo
    watch beep hello # This will fail
    alias watch='watch '
    watch beep hello # Now it works!

[1]: https://news.ycombinator.com/item?id=31830879

Brilliant feature, but this is the sort of thing that reinforces the idea that CLIs have awful discoverability compared to GUIs.

Which is why CLIs should come with manuals - there is even a dedicated CLI program to view those (well multiple actually, because standards wouldn't be fun if there was only one).

I start my aliases with a period so that I have...

.foo1,.foo2, .foo3, and so on.

If I unknowingly named my command the same as some system command and the custom command ends up earlier in my path it doesn't really cause any problems does it? I wasn't using that system command anyways, and my custom command is only going to take priority if my .bashrc file was run first.

Until you use a different machine and accidentally type your command without realizing it's not there.

It can if you are executing programs that then try to execute those commands and get your commands instead.

I wish it was easier to create "snapshots" of a particular set of environment variables in order to use them later as run environments. I also wish that Unix desktop environments generally made it easier to manage env vars, but that's another complaint. And don't get me started on PAM env vars...

Just env vars?

    env > "$(date +s).env-snapshot"

(And use it again slightly more awkwardly as

    <$file | xargs -I@ echo '"@"' | tr '\n' ' ' | xargs -I@ env @ cmd

or something more convenient to `cmd`.)

Should be

  <$file xargs ...

Oops, you've exposed that it started off as a useless cat I knew I'd be chastised for ;)

If it is any consolation, I can't make any sense of this.

  $ <xyz.txt wc -l
         8
  
  $ <xyz.txt | wc -l
         0

https://direnv.net/

I use Direnv in all of my projects, but it doesn't make it any easier to manage consistent groups of env vars or assign different executables to run in different groups.

> I wish it was easier to create "snapshots" of a particular set of environment variables in order to use them later as run environments.

docker gives you this (but only as part of a whole ecosystem, which for various reasons you might not want to hitch your horse to).

Direnv has been mentioned, and it meshes well with functional package managers like Nix and Guix, each of which include the concept of "Profiles". Each profile can access a specific set of programs via it's environment, but PATH is far from the last variable that you can use Direnv or Profiles to control.

Yes, all of that. And if you later decide to start using the system command, you just rename your custom command.

I use zsh and have the following alias (rather, function):

``` cp () { rsync -avhW --progress --inplace ${@} } ```

If I ever need to use the `cp` binary, I execute `=cp`, which evaluates to `/bin/cp`.

Why use rsync? Only for the progress bar?

I have used a similar windows trick. I wanted to add to the default window system a place for my library of tools. Most are "portable" cmd apps from various sources (sysinternals, nirsoft). I put them all into a directory called ] in the root of a drive.

Because of some common layout of keyboards the \ and ] are near each other, or in reach of left and right pinkies I can type \]\ before the name of the tools. This means I don't even have add a dir to the path.

I am sure this violates someone sensibilities but it has saved time over the years as a consultant when I had to deal with a wonky machine and owner did not want permanent install of diagnostic tools.

I put c:\tools in my path, then put shortcuts to all my tools in there (ie: shortcut to AgentRansack named ar). Now, I pull up the run command (WND+R) type ar and agent ransack opens.

Permissions!

Post-vista, c:\programdata\tools

Set yourself a %TOOLS% environment variable.

Unrelated but I heard it's possible to rename c:\ to /usr/bin or something, depending on your usage. I personally wouldn't mind renaming it to ~/<user-name> and be able to use some of my dotfiles in Windows.

It's possible, but 90% of software will break because of this lol

Clever trick. Too bad there's no good way to namespace commands. It would be interesting if you could have, for instance:

  $ mkdir -p /tmp/bin/my
  $ printf '%s\n%s\n' '#!/bin/sh' 'echo hello' > /tmp/bin/my/hello
  $ chmod 755 /tmp/bin/my/hello
  $ PATH="/tmp/bin:$PATH"
  $ my/hello
  hello

Just add another comma. Of course, this needs some deeper magic to work :)

    ,my,hello

I don't know if I would actually recommend this, but it works:

    $ touch foo::bar
    $ chmod +x foo::bar
    $ ./foo::bar && echo $?

It's an empty file so it does nothing, but most shells' lexers should be smart enough to not try and lex '::'.

Edit: Forgot to mention, this also works with bash functions and aliases. I'm not sure if POSIX sh will allow it. You can also do things like this (again, not sure I'd recommend it):

    $ alias +x='chmod +x'

> mkdir -p /tmp/bin/my

This is not a secure way of using /tmp.

Sure, but that's not the point. It's just an illustrative example, and I wanted a short absolute path that is easily recognizable to the reader. I'll fix it, though, to avoid having an example with bad security practices, though.

edit: Or I would fix it, but I guess I can't edit it. Oh well.

Plan 9 does that, I'm not sure why it never made it into Unix because I always loved it. Your networking-related commands were in `/bin/ip` and you'd run e.g. `ip/ping 1.1.1.1`. If you created the dir `$home/bin/rc/my` and put scripts in it, they would have been runnable as `my/hello` etc like in your example.

Plan 9 also more or less does away with the concept of $PATH in the first place; just union-mount (with `bind`) whatever you want onto `/bin` and call it a day. Yet another thing that would've been great for Unix-likes to pull in (I think some Linuxen can technically do it, but with some limitations).

All linuxen should be able to do that with overlayfs, I think.

no matter which kind of twisted idea someone comes up with related to terminals, there is always a guy who chimes in with "in Plan 9 you can do that...". It makes me smile every time. :)

I've tested it and it works on Linux (bash and zsh), at least for your custom commands

Ahem, technically there is one way (don't know if it is "good", though):

    $ export $MY = /tmp/my
    $ $MY/hello
    hello

You could (ab)use home directories for that. I.e. place your hello into /home/my, and then you can invoke it as ~my/hello. You even get namespace completion for free after the `~`.

In zsh you can do "hash -d my=/path/to/my" and then you can use "~my/hello" from anywhere, and it doesn't need to be in /home.

> and it doesn't need to be in /home.

Neither do home directories.

zsh can do this if you set the PATH_DIRS option, but if I remember right it doesn't autocomplete deeply nested directories, just one level.

I made a little tool that dispatches hierarchical commands in a slightly nicer way, with intelligent autocomplete with command descriptions: https://github.com/ianthehenry/sd

What about:

  my hello

I.e. only my needs to be in the global namespace and all other commands are resolved by my itself.

You’d need to set up tab completion for the `my` command. This is more effort than just naming everything `my-foo` — or `,foo`.

Good point.

I think shells should support that, allowing you to do

  $ my hello

It would mean tools such as git, microk8s and aws wouldn’t have to implement that “look up first argument, check whether it’s a command I know, and if so, run it” on their own, and would have slightly simpler auto complete configs.

It also could mean launching fewer processes (for the case where ‘my’ is a script or executable that looks up the location of ‘hello’ and launches it)

https://ianthehenry.com/posts/sd-my-script-directory/

I tend to give programs longer, more descriptive names and then do the short names as shell aliases. That way there’s little risk of name collision for the original program, scripts using the canonical name remain readable, and the shorthand is free to evolve to fit varying work habits.

My keyboard came with a tab key, so I don't worry about my commands having short names.

Until you have five commands which all use the prefix `sort-incoming-` and then three commands named `sort-incoming-hourly` and `sort-incoming-monthly` etc.

It would be nice to have camel-case completion in the shell, i.e. have “SIH” be completed to “SortIncomingHourly” and “SIM” to “SortIncomingMonthly”, etc. (only for upper-case characters). That way one could get more mileage out of the possible character combinations.

Well, not exactly tab completion, but I have `fzf`[0] tied to my history search.

That allows me to type `soriloco`, punch ctrl+r and be prompted with `some ridiculously long command` (if, of course, I've run it before).

Combined with infinite history this trick has saved me a bunch of typing and even proper remembering. :P

[0]: https://github.com/junegunn/fzf

Yes, fuzzy matching is the way to go. Tab completion isn’t good enough any more.

Why do the names have to be shorter, I never have to type it all out just the first couple of letters and then autocomplete. Dealing with aliases/“shortcuts” is such a pain plus I hate that pattern anyways.

This brought https://github.com/Shopify/comma to mind

Do you know any of the history of why your link to `.../Shopify/...` redirects to `.../nix-community/...`? Is the latter an open-source contribution by Shopify?

I think it was originally developed and maintained by Shopify, but was then handed off to the nix community

Correct. Source: I used the shit out of that command. Thanks Burke

Thanks!

This is for Nix.

I do something similar but 1. for aliases and 2. use period. For example, when working on a Rails project I define the following alias:

    function .routes() {
      if [[ $# -eq 0 ]]; then
        .rails routes
      else
        .rails routes | grep $@
      fi
    }

where .rails is another alias (for running bundle exec rails ...). I can then run ".routes webhook" and see all routes containing "webhook".

Last but not least, I automated the whole thing with bash-ctx (https://github.com/gregnavis/bash-ctx).

I started creating wrapper-scripts or simple scripts that has the commands with my prefered options or a selection menu for said command. I have started calling those scripts my<command>

rsync becomes myrsync mount becomes mymount ssh becoms myssh

No collision and the naming convention empathise that it is indeed used be MYself.

Why people worry so much to create "short, easy to type" (coff coff cryptic) command names, nomenclature is so important in our profession and almost everybody sucks at it, give meaningful names to things even if they are 30+ chars long, is it really that though to press 3 or 4 keys along with maybe 3 or 4 tabs?

I was going to post the same thing. With tab completion long names can be typed quickly anyway. Also, the keyboard is the most fundamental input device for developers... learn how to use it well. It amazes me how many people who are professional developers use copy/paste so much because they can't type well. It slows them down to a crawl.

I don't particularly want to have to type 30-character names all the time without tab completion, but I'm not going to worry about an eight character command vs. a four character one.

Obviously this doesn't apply if you are dyslexic or disabled in some way or something.

I use this method for aliases in my `.*rc` file so I can remember what the heck I called things. Otherwise, I forget whether I have to type something like `COMMANDNAME-log` or `log-COMMANDNAME` or some other weird combo like `print-COMMANDNAME-log`. With a comma, I type `,<tab>` and get all of my custom commands listed easily.

Cool.

I have a script that creates new ~/bin scripts that checks for collisions locally (functions, builtins, and executables) and in linux distros.

For example, I have a ~/bin/up script on every box and platform to install updates. So far so good on collisions.

Commas can also be used in filenames to help you sort things instead of using underscore as a separator, useful for the same reason -- you don't have to hit shift to type it.

Can't you just declare:

    my() {
  script_name="$1"
  script_path="$HOME/.local/bin/my/$script_name"
  shift
  $script_path "$@"
    }

So you have your namespace ? Of course that mean you must put all script in .local/bin, not anywhere in the PATH, and some more work is needed to figure out completion.

But it seems less weird than the comma.

-- EDIT --

With completion:

export MY_SCRIPS_DIR="$HOME/.local/bin/my"

_my_completions() {

   if [ "${#COMP_WORDS[@]}" != "2" ]; then
     return 
   fi

   COMPREPLY=($(compgen -W "$(echo $(find "$MY_SCRIPS_DIR" -printf '%P\n' ))"))
 }


 my() {
  script_name="$1"
  script_path="$MY_SCRIPS_DIR/$script_name"
  shift
  $script_path "$@"
 } 

 complete -F _my_completions my

-- EDIT 2 --

After testing both, the comma win. Kiss, more flexible.

This is a good trick to avoid collisions with system commands. Anytime you add one of your own executables to PATH, prefix it with a comma (,). This way you avoid name collisions and have a handy way of seeing what commands are available on your system using tab completion (, then tab lists all of your personal executables)

Is this comment from a bot? You just reiterated everything from the article.

This is an explanation that the parent comment simply reiterated the point of the article rather than adding content.

This is a link to a story titled "This is the title of this story, which is also found several times in the story itself": https://stuff.mit.edu/people/dpolicar/writing/prose/text/tit...

You are right, but I’m guessing a lot of people don’t read the linked article. It’s a pretty good summation.

Though it doesn’t mention using ~/bin. I use that but don’t add it to my path, so I have to explicitly call my scripts.

The comment history suggests it isn’t a bit though.

Related to GP, here's an interesting article: https://rhodesmill.org/brandon/2009/commands-with-comma/

Meh. I'd prefer to have a path not in my PATH, and then explicitly call the binary with the full path .. and then append a #comment for easy recall with CTRL-R at some point later in the future.

A bit more fiddly perhaps, but future proof.

In the comments, the author argues against the underscore by saying Bash/Zsh already uses it for completion. But it looks like it's a Zsh thing - at least my Bash install is pretty conservative about it.

Not sure what I have installed/sourced, but "_<tab>" says there are 381 possibilities for me. (This is with bash.)

Also in the article itself, the author argues against any prefix char that requires using the shift key.

To mitigate issues you could just append ~/bin/ to $PATH (opposed to prepending). Then system commands have priority over your user home commands, right?

I almost always add "," in cmd.exe to anything I run from Windows Command Prompt? Why? Because cygwin, or mingw's `dir.exe` would be called instead of built-in dir (well, this is further complicated that I don't really use `cmd.exe` but FAR Commander), so if I want to enforce the built-in over external (same name) command I have to prepend it with "," - so I basically started typing "," in front of each of my commands...

I do basically the same thing with any aliases or commands that I have just for me. Except I prefix with _ (an underscore.) Having some sort of prefix is extremely nice as it ensures you never override a built-in on accident, AND it makes it easy to tab complete just the ones that you want.

For example I have ‘_f’ aliased to a fuzzy find command that will launch vim on the selected files, and a different one for launching sql clients to various servers etc.

As a prefix, commas work great!

Unfortunately zsh uses the underscore prefix heavily, _<TAB> results in hundreds of possible completions.

I have my bin folder in ~/usr/bin. The structure of ~/usr matches /usr and /usr/local, so I can build stuff into it with ./configure --prefix=$HOME/usr or similar. (You could of course do ./configure --prefix=$HOME, but then you run the risk of ending up with ~/share and ~/include and so on, something I don't want.)

Similar to the vim <leader> key concept which many people use comma for.

https://tuckerchapman.com/2018/06/16/how-to-use-the-vim-lead...

I use a similar trick for text expansion. All my abbreviations start with a semicolon (;). It’s on the home row, it never occurs at the start of a word in normal typing so I never have a false positive expansion, and it’s easy to remember.

I am surprised to see someone else doing this, but I agree it works surprisingly well and one of my favorite tips.

I've never once had a problem with one of my commands shadowing a system command so I think I will continue on rather than solving this non-problem, thanks.

Is it secure to have ~/bin/ it the PATH?

The issue is that an exploit could write a file in ~/bin/ and trick you in to executing it. But, if something malicious is writing files in your home directory, you've already lost.

It's a default behaviour of at least bash. It's as secure as your home directory is.

If someone can write to your home directory they can also update your bashrc file and add whatever they want to PATH

Cool.

I’m gonna start them with `å` just to be a little different.

That’s be fun when you Exchange scripts with non-danes ;-))

Finally, a weird tip that is really good!

I just add the .sh suffix/extension to them... And I use j- as prefix for the "tab trick"

clever idea, but does it need such a long post to describe?

Two phone sized screenfuls of text is too long now?

I'm genuinely interested to know whether you read longform text at all.

I mostly read books. I don't want to pick on the author, but since you asked I will be more specific: this idea would fit in a tweet.* It's a good idea, but the long build up reminds me of recipe sites.

* I just use "tweet" as a unit of measure, like "the size of two elephants". I do not advocate the use of twitter.

twitter was still fairly new when this was published and only allowed 140 chars which definitely isn't enough.

As I said I wasn’t advocating the use of twitter, merely using it as a unit of measure.

If you're that pressed for time, you probably could get away with just reading the title.

the aesthetics of the comma for daily use are not good

In Zsh, you can used named "directories" (actually it works with normal files too) with tildes for the same effect, except that they must be explicitly defined, so there's no risk of collisions.

    hash -d e=/bin/echo
    ~e hello world

https://ato.pxeger.com/run?1=m724qjhjwYKlpSVpuhY3lTMSizMUdFM...

I do the same thing but with prefix …

I misadvertantly lower cased my functions names when working on .net, but it made it super easy to find my code in the stack traces. (.net idiom is capitalism the first case of a function name). Usually the cause of stack traces was my code so it was an unexpected productivity booster.

haha neat.

The comma is employed in bash and other shell for brace expansion.[1] I can't see how this would conflict with OP's master plan, but I have fundamental objections towards the depths of OP's laziness. Lazy is ok, but there should not be competition to see who can be laziest. At some point, things just won't compress any further, and the attempt at further streamlining efficiency beyond what is useful has a negative effect on efficiency overall. Besides, tab completion is fine as it is and doesn't need what passes for a hack these days. To be clear, contrary to what OP has claimed about not being creative enough to come up with original script names that won't conflict with the flood of new Linux commands every update (wot? LOL), this is only about the OP not being satisfied with tab completion.

[1] https://superuser.com/a/184493

That's a lot of words to say "I don't like this" and insult folks who put time and effort into customizations that improve their own computing experience.

> who put time and effort into customizations

I believe you have made my point quite well once it's understood you are here referring to none other than prefixing script names with a comma. Any words wasted on anything other than the subject is a waste of words, in your comment's case, both ad hominem and straw man fallacies.

Reply on news.ycombinator.com