The beauty of cat is that streams are the universal interface.
Program A might accept a file as the last positional arg. Program B might accept it as a named arg, where the name/flag could be anything from --input or -f or --file etc.
But a program will read from STDIN, which all good unix programs do, then piping cat into it works every time. I can write the cat foo.txt part before I even know what command I'm piping it into.
This. Sometimes I want to see what I'm looking at and then (using that dump as a reference) follow up with a corresponding filter (| jq .key, or | tail -n 30). Sure, I could use less, but then I context switch on exit; no support from the scrollback buffer.
I've probably lost 10ms * 1E5 of my life from the extra PID. But, probably would lose more in the context switch.
Unless you're executing these commands in a loop over a large number of items, or the item itself is gargantuan, it's almost always harmless.
Personally, when I'm exploring, I build a command line iteratively. Cat the file to see the content, pipe to grep to get the lines I want, sed/awk/cut/etc to finagle from there.
if this wanton abuse of cat(1) doesn't stop, we're on track to run out of PIDs by 2031! Just because Unix makes it cheap and easy to fork doesn't mean you have to!
same I just like monads lol. cat + pipe feels purer and has lower mental load for me, which dominates the efficiency of spawning an extra process for, typically, a few microseconds.
Admittedly its taken me a long time to remember that the file is the last argument to grep, when so many other commands its the first. I'd guess common abuse is due to being easier to type cat x | than to dig up the man page
And also typing cat x to get a quick look at the file, hitting up, then piping that into another command and taking a look, hitting up, piping that result into a third command etc.
It's that way so that you can grep multiple files with a single pattern. It would be odd for the pattern to come after the file arguments. It also allows the files to be optional so that it can grep stdin.
Probably, but knowing that redirection operators can be freely moved within normal arguments [EDIT: thank
ButlerianJihad for pursuing me to make this more accurate] is useful.
`2>&1` redirects FD2 to the current contents of FD1 (stdout), then `> /dev/null` redirects FD1 to /dev/null. That results in your errors going into stdout, and discarding regular output altogether:
When you flip the order, `> /dev/null 2>&1` moves FD1 to /dev/null first, and then FD2 to the contents FD1 (/dev/null again), so you discard both errors and standard output:
All of this depends on your specific shell and its parser. Fish doesn't let you put redirections at the beginning like that (though I wish it did), while GNU Bash does.
fish is not POSIX-compatible, and not Bourne-compatible, so I don't see how that really matters at all. I used the rc shell from plan9 for quite a while, and I wouldn't expect its syntax rules to match, either!
It does not yield the "same output", and here is why: if you cause your command to actually produce output on stderr (fd 2) it will appear as terminal output, because you have actually succeeded in "redirecting" stderr to wherever stdout (fd 1) was pointing initially.
So what we've gathered from this interaction is that the relative order of redirection operators matters, but it doesn't matter where in the command they are placed apart from that.
Presumably written by someone without much interactive shell experience.
When you're building a pipeline, putting cat first can often be quite convenient. Essentially, it's more composable: it defines the input to the pipeline without committing to a specific tool. For example, you can up-arrow in the shell and change the part after the pipe without having to skip back past the filename.
In fact if you don't start with cat, it's possible you're more of a script kiddie than a software developer.
I raised eyebrows recently when I was working with someone and we needed to create a file and instead of starting an editor I did:
cat > filename
...
Ctrl-D
The beauty of cat is that streams are the universal interface.
Program A might accept a file as the last positional arg. Program B might accept it as a named arg, where the name/flag could be anything from --input or -f or --file etc.
But a program will read from STDIN, which all good unix programs do, then piping cat into it works every time. I can write the cat foo.txt part before I even know what command I'm piping it into.
This. Sometimes I want to see what I'm looking at and then (using that dump as a reference) follow up with a corresponding filter (| jq .key, or | tail -n 30). Sure, I could use less, but then I context switch on exit; no support from the scrollback buffer.
I've probably lost 10ms * 1E5 of my life from the extra PID. But, probably would lose more in the context switch.
> Piping a single file through cat spawns an entire process whose only job is to copy bytes to a program that already knew how to read them.
Chrome probably spawned two processes when I cmd+clicked this into a new tab. It really doesn't matter.
Unless you're executing these commands in a loop over a large number of items, or the item itself is gargantuan, it's almost always harmless.
Personally, when I'm exploring, I build a command line iteratively. Cat the file to see the content, pipe to grep to get the lines I want, sed/awk/cut/etc to finagle from there.
if this wanton abuse of cat(1) doesn't stop, we're on track to run out of PIDs by 2031! Just because Unix makes it cheap and easy to fork doesn't mean you have to!
(who gives even a single shit, my god)
Is this a dig at IPv6?
I like piping the output of cat and the mental image of one process feeding another. It's inconsequential, but it brings an epsilon of joy.
same I just like monads lol. cat + pipe feels purer and has lower mental load for me, which dominates the efficiency of spawning an extra process for, typically, a few microseconds.
> Since 1995, occasional awards for UUOC have been given out, usually by Perl luminary Randal L. Schwartz
http://catb.org/jargon/html/U/UUOC.html
Admittedly its taken me a long time to remember that the file is the last argument to grep, when so many other commands its the first. I'd guess common abuse is due to being easier to type cat x | than to dig up the man page
And also typing cat x to get a quick look at the file, hitting up, then piping that into another command and taking a look, hitting up, piping that result into a third command etc.
It's that way so that you can grep multiple files with a single pattern. It would be odd for the pattern to come after the file arguments. It also allows the files to be optional so that it can grep stdin.
The redirection operator is consistent and requires less typing though.
I guess the file is usually the last argument because it's the one that can be omitted.
I'll make a note of it in my AGENTS.md file.
Don't do this:
Do this instead: The front-cat abuse is all about the order. The effective solution needs to keep the relative order of arguments.Or just use cat and spend your brainpower on interesting, useful, and/or worthwhile topics. It boggles my mind that anyone cares about this.
Probably, but knowing that redirection operators can be freely moved within normal arguments [EDIT: thank ButlerianJihad for pursuing me to make this more accurate] is useful.
They are actually not “order-independent”, and their L-R parsing/processing is why constructs such as
work as intended.funny enough,
appears to yield the same output. So i wonder where the not "order-independent" chimes in.`2>&1` redirects FD2 to the current contents of FD1 (stdout), then `> /dev/null` redirects FD1 to /dev/null. That results in your errors going into stdout, and discarding regular output altogether:
When you flip the order, `> /dev/null 2>&1` moves FD1 to /dev/null first, and then FD2 to the contents FD1 (/dev/null again), so you discard both errors and standard output: In your example, `cat file` is unlikely to produce any errors, which is why you're not seeing a difference.All of this depends on your specific shell and its parser. Fish doesn't let you put redirections at the beginning like that (though I wish it did), while GNU Bash does.
fish is not POSIX-compatible, and not Bourne-compatible, so I don't see how that really matters at all. I used the rc shell from plan9 for quite a while, and I wouldn't expect its syntax rules to match, either!
You're absolutely wrong!
It does not yield the "same output", and here is why: if you cause your command to actually produce output on stderr (fd 2) it will appear as terminal output, because you have actually succeeded in "redirecting" stderr to wherever stdout (fd 1) was pointing initially.
So what we've gathered from this interaction is that the relative order of redirection operators matters, but it doesn't matter where in the command they are placed apart from that.
Presumably written by someone without much interactive shell experience.
When you're building a pipeline, putting cat first can often be quite convenient. Essentially, it's more composable: it defines the input to the pipeline without committing to a specific tool. For example, you can up-arrow in the shell and change the part after the pipe without having to skip back past the filename.
In fact if you don't start with cat, it's possible you're more of a script kiddie than a software developer.
"Don't be a catgrepper"
- various HostGator employees, c. 2011
In this day an age this is still making rounds ? So this is the memory usage of cat on my system:
To me there are far more things to worry about than cat. How about your multi-gig browser for one ?Now for firefox:
Maybe people should be looking at that ? I will not even get into modern Linux Desktops :)I’m going to keep doing it but wouldn’t mind it if my shell auto replaced it for me.
I like putting the stdin before the command
< file grep abc
I raised eyebrows recently when I was working with someone and we needed to create a file and instead of starting an editor I did: cat > filename ... Ctrl-D
Why not touch or echo? No reason for an editor or cat
For a one line file sure, but I was creating multiple lines.
You can type the intended file contents as-is.