In Advanced Bash-Scripting Guide, in example 27-4, 7-th line from the bottom, I've read this:
A function runs as a sub-process.
I did a test in Bash, and it seems that the above statement is wrong.
Searches on this site, Bash Man, and my search engine don't bring any light.
Do you have the answer and would like to explain?
The Advanced Bash-Scripting Guide is not always reliable and its example scripts contain out-dated practices such as using the effectively deprecated backticks for command substitution, i.e., `command` rather than $(command).
In this particular case, it’s blatantly incorrect.
The section on Shell Functions in the (canonical) Bash manual definitively states that
Shell functions are executed in the current shell context; no new process is created to interpret them.
Curly brace functions will run within the calling shell process, unless they need their own subshell which is:
&Redirections or extra env. variables won't force a new subshell:
hw(){
echo hello world from $BASHPID
echo var=$var
}
var=42 hw >&2
echo $BASHPID #unexports var=42 and restores stdout here
If you define the function with parentheses instead of curlies:
hw()(
echo hello world from $BASHPID
)
hw
echo $BASHPID
it will always run in a new process.
Command substitution $() also always creates processes in bash (but not in ksh if you run builtins inside it).
f() (...) is allowed. Are there any other definitions apart from {...} and (...)? In Bash, I'm not into others yet.
–
Aug 23 '16 at 23:43
function hw { echo hello world; } syntax (no need for () if you type out function and you can specify redirections right after the final } or ) as in hw(){ echo error; } >&2. That's about it.
– Petr Skocik
Aug 23 '16 at 23:54
f()(...) always execute an own shell, whilst f(){...} does not.
– rexkogitans
Aug 24 '16 at 08:22
foo() [[ x = x ]] is a valid function definition, too. However, if you look at the function with type foo you'll see that this is still syntactic sugar for foo() { [[ x = x ]]; }. The same is true of subshell functions: bar() ( : ) becomes bar() { ( : ); }.
– kojiro
Aug 24 '16 at 11:54
The command in question from that example looks like:
echo ${arrayZ[@]/%e/$(replacement)}
The example later states:
# $( ... ) is command substitution.
# A function runs as a sub-process.
Being charitable to ABS Guide, what they apparently meant to write is that the function runs inside a command substitution and the command inside a command substitution runs in a subshell.
I'd note, that if executed in the background, a bash function will execute in a subshell. I only note this because I stumbled onto this question while checking for that fact.
[192.168.28.1 (1ae7468c) ~ 19:17:50]# function foo () { echo $BASHPID; }
[192.168.28.1 (1ae7468c) ~ 19:17:59]# foo
807477
[192.168.28.1 (1ae7468c) ~ 19:18:02]# foo
807477
[192.168.28.1 (1ae7468c) ~ 19:18:03]# foo &
[1] 1996783
1996783
[1]+ Done foo
[192.168.28.1 (1ae7468c) ~ 19:18:06]#
***if you want a proof, I can't give you. I've derived it from webpages, comments... I've read that statement somewhere sometime ago.
– Daniel Bandeira Jul 23 '20 at 15:06