Process substitution in bash

shruggy · 04-27-2021, 09:26 AM

Quote:

Originally Posted by igadoter

Don't get it

Code:

$(< /tmp/file)

is command output substitution - but where's command ? < /tmp/file is just redirection.

From the GNU Bash manual:

Quote:

The command substitution $(cat file) can be replaced by the equivalent but faster $(< file).

pan64 · 04-27-2021, 09:35 AM

process substitution is when the < or > is put before (. Also in that case there is no $.
$( ) is a different thing, that is called subshell.
So

Code:

<( echo wobble )  # this is process substitution
$(< /etc/file)    # this is a simple redirection in a subshell

Do not mix them.

igadoter · 04-27-2021, 09:42 AM

Quote:

Originally Posted by shruggy

Here are a couple of examples of how I used the output variant when doing Kakoune golf:

You got my attention.

shruggy · 04-27-2021, 09:51 AM

Well, the problem was that the -f option of kak required a file argument. I could use /dev/stdout, but in the end choose >(cat) instead.

igadoter · 04-27-2021, 10:01 AM

Hello world interactive programm

Code:

% echo hello $(< <(read World; echo $World) )
Mars
hello Mars

GazL · 04-27-2021, 10:38 AM

Quote:

Originally Posted by pan64

$( ) is a different thing, that is called subshell.

Have to take issue with that. Though most of these substitutions do run in sub-shells.

$(< /tmp/file) is a special case of command-substitution that is a more efficient replacement for $(< /tmp/file cat ).

(< /tmp/file cat ) would be an example of a subshell with internal redirection.

pan64 · 04-27-2021, 12:09 PM

There are 3 cases:

Code:

$(< file)

in this case the subshell will do a redirection on its stdin and after that will exit, there is nothing [else] to do. The "result" is that the content of stdin will arrive to stdout without any modification (finally you will get the content of the file).

Code:

$(cat file)

in this case the subshell will fork/exec an additional command, a cat which will read the file and send the content to the stdout of the subshell. The result will be the same, just in this case an additional binary (cat) was involved too.
At the end they produce exactly the same result, just the former one is much more efficient.

Code:

$(<file cat)

Now the file will be redirected to stdin by the subshell, but the shell will execute an additional cat command which will process that input (and will do nothing, just pass it to stdout). Again, functionally identical, but inefficient as the second example. (the only difference is: the subshell will create the file descriptor - open the file and pass it to cat or cat will open the file)

igadoter · 04-27-2021, 12:21 PM

Thanks. Construction with <file even works at command line

Code:

 
% </tmp/file cat
world

and

Code:

% cat /tmp/file
world

of course one can put whatever wants instead of cat say

Code:

% </tmp/file dd

MadeInGermany · 04-27-2021, 02:05 PM

Code:

$(< /tmp/file)

is specific to bash and zsh:
in this (sub shell) context, if the command is missing then stdin is connected to stdout, behaving like cat
Other shells need a command there:

Code:

$(cat < /tmp/file)

zsh does it even in a the main context:

Code:

< /tmp/file

reads the file, where other shells need

Code:

cat < /tmp/file

Consequently zsh writes the file (from stdin) with

Code:

> /tmp/file

where other shells need

Code:

cat > /tmp/file

Just seeing your last post - do you have zsh?

mimorek · 04-27-2021, 02:58 PM

From the bash man page:

Quote:

The command substitution $(cat file) can be replaced by the equivalent but faster $(< file).

igadoter · 04-27-2021, 08:13 PM

Quote:

Originally Posted by MadeInGermany

Just seeing your last post - do you have zsh?

Yes. But never used.