I’m implementing builtins env and export in a project recreating bash shell. I have inside my shell copy of envp, t_env **envp Then I modify it.
What is the conceptual difference between
export a and export a=?
Why in first case I can see it with command export and not env?
How to implement this behavior?
At the moment my implementation is to use flag, to distinguish between env and shell variables, but maybe there is a better way?
And why even exists this option, set variable without value?
9
This exists for historical reasons.
I believe in the original Bourne Shell, you could not assign the variable at the same time as exporting it. So you had to do it in two steps:
variable=value
export variable
Later on these were combined, so you could write
export variable=value
but the original export without assignment is still allowed for compatibility
1
export a makes it possible for a to be added to the environment of a child process but nothing is actually added unless the variable has been assigned a (possibly null) value.
testenv(){
bash -c '
declare -p myvar0 myvar1 myvar2 myvar3
unset_or_null(){
declare -n var=$1
case ${var-1}/${var:-1}/${var:+1} in
1/*/*) echo $1 is unset ;;
*/1/*) echo $1 is null ;;
*/*/1) echo $1 is set ;;
esac
}
unset_or_null myvar0
unset_or_null myvar1
unset_or_null myvar2
unset_or_null myvar3
'
}
export myvar1
export myvar2=
export myvar3=ok
testenv
echo -----
myvar0=
myvar1=
testenv
echo -----
myvar0=ok
myvar1=ok
testenv
bash: line 2: declare: myvar0: not found
bash: line 2: declare: myvar1: not found
declare -x myvar2=""
declare -x myvar3="ok"
myvar0 is unset
myvar1 is unset
myvar2 is null
myvar3 is set
-----
bash: line 2: declare: myvar0: not found
declare -x myvar1=""
declare -x myvar2=""
declare -x myvar3="ok"
myvar0 is unset
myvar1 is null
myvar2 is null
myvar3 is set
-----
bash: line 2: declare: myvar0: not found
declare -x myvar1="ok"
declare -x myvar2=""
declare -x myvar3="ok"
myvar0 is unset
myvar1 is set
myvar2 is null
myvar3 is set