I have next:
$ export A=1
$ eval "echo $A; echo $A; lll; echo $?"
The output is:
1
1
-bash: lll: command not found
0
I don't know why I cannot get the exit code of lll, and I happen to try next:
$ export A=1
$ eval "echo $A; echo $A; lll; echo \$?"
1
1
-bash: lll: command not found
127
You can see above works, meanwhile next also work:
$ export A=1
$ eval "echo \$A; echo \$A; lll; echo \$?"
1
1
-bash: lll: command not found
127
I wonder, why I had to add one \ before $?? Also, why \ before $A is not a must?
Remember that, because you have embraced the whole expression to be evaluated, in double quotes. This means that all parameters inside it will be expanded, before eval is invoked. If you don't escape $?, the parameters to be expanded are A and ?. You get the value of A before eval is run (which is what you want), but you would also get the value of ? before eval is run (which is not what you want). The backslash causes a literal $ to be passed to eval, hence defering the time of calculating the status code.
$ eval "echo $A; echo $A; lll; echo $?"
The variables enclosed in double quotes are expanded. The line above is the same as:
$ eval "echo 1; echo 1; lll; echo 0"
In fact, this string is exactly what eval receives as argument. You don't even need to export A for this.
In order to achieve what you want to should enclose the string in single quotes. This way the variables are not expanded any more and eval receives as argument the string exactly as you typed it.
Try these two sets of commands
$ export A=1
$ B="echo $A; echo $A; lll; echo $?"
$ eval $B
$ A=2
$ eval $B
vs.
$ export A=1
$ B='echo $A; echo $A; lll; echo $?'
$ eval $B
$ A=2
$ eval $B
That's a difference, isn't it?
Related
Consider the following:
#!/bin/tcsh
set thing = 'marker:echo "quoted argument"'
set a = `echo "$thing" | sed 's/\([^:]*\):\(.*\)/\1/'`
set b = `echo "$thing" | sed 's/\([^:]*\):\(.*\)/\2/'`
echo $a
echo $b
$b
echo "quoted argument"
This gives
marker
echo "quoted argument"
"quoted argument"
quoted argument
If $b is echo "quoted argument", why does evaluating $b give a different result from echo "quoted argument"?
Since I know tcsh is awful (but it's what I have to use for work), here is the same problem in Bash:
thing='marker:echo "quoted argument"'
a=`echo "$thing" | sed 's/\(.*\):\([^:]*\)/\1/'`
b=`echo "$thing" | sed 's/\(.*\):\([^:]*\)/\2/'`
echo $a
echo $b
$b
echo "quoted argument"
The output is the same. Note that, were I doing this in Bash, I would certainly use a map. I do not have that luxury :). The solution must work in tcsh.
Desired Output
I would like $b to behave just as if I typed the command in myself as I see it:
marker
echo "quoted argument"
quoted argument
quoted argument
This is a follow-up question to Accessing array elements with spaces in TCSH.
Yeah, eval is the "solution" here (well the solution is not to have a command in a string in the first place see http://mywiki.wooledge.org/BashFAQ/050 for more).
The reason you see the quotes when you run $b is because of the order of evaluation of a shell command. The very last thing that the shell does, after all other expansions, is to remote quotes (however it doesn't remove quotes that resulted from any of the expansions).
So when you have b='echo "quoted arguments"' and run $b as the command line what happens is that the variable is expanded so you get echo "quoted arguments" and then that is run as-is.
$ c ()
{
printf 'argc: %s\n' "$#";
printf 'argv: %s\n' "$#"
}
$ b='echo "quoted arguments"'
$ c "quoted arguments"
argc: 1
argv: quoted arguments
$ c $b
argc: 3
argv: echo
argv: "quoted
argv: arguments"
$ c "$b"
argc: 1
argv: echo "quoted arguments"
$ eval c $b
argc: 2
argv: echo
argv: quoted arguments
$ eval c "$b"
argc: 2
argv: echo
argv: quoted arguments
One thing you must keep in mind with command substitution is that each pipe and each command you string together executes within its own subshell. Each time that happens, the shell process the command or string you provide:
If $b is echo "quoted argument", why does evaluating $b give a
different result from echo "quoted argument"?
set thing = 'marker:echo "quoted argument"'
set b = `echo "$thing" | sed 's/\([^:]*\):\(.*\)/\2/'`
echo $b
echo "quoted argument"
In the case of b, you are assigning the return from sed exactly as it is returned to b including the quotes. They become part of b. So echo $b is equivalent to echo '"quoted argument"'. Whereas, your echo "quoted argument" prints the string as the characters contained within the quotes, the shell removing the literal quotes.
Sorry for the initial confusion.
I explained my question in the comments:
VAR=
INS="Installing $VAR"
echo $INS
. # In each echo command I want to dynamically substitute
. # the $VAR variable in the $INS variable. I want to do
echo $INS # the substitution of the variable on echo command.
Is this possible?
You need a function to do the job gracefully.
say() {
echo "Installing $INS"
}
INS=hello
say
INS=world
say
Or just this:
say() {
echo "Installing $#"
}
say hello
say world
For example:
[ghoti#pc ~]$ cat varins
#!/bin/bash
msg='Installing "$VAR"'
for VAR in foo bar baz; do
eval echo "$msg"
done
[ghoti#pc ~]$ ./varins
Installing foo
Installing bar
Installing baz
[ghoti#pc ~]$
This relies on the fact that $VAR won't be expanded inside single quotes. The eval command will expand the $msg variable, in which the shell will find $VAR to replace.
Parameter substitution can do it also:
ins='aaa $var aaa'
var='xxx'
echo "'${ins//\$var/$var}'"
result:
'aaa xxx aaa'
INS='Installing $VAR'
result=`eval "echo $INS"`
echo $result
you have to use single quotes to not substitute $VAR on creation of INS
eval evaluates $INS at runtime and with echo and backticks it is returned as substituted string
As Hachi said, you need single quotes to prevent $VAR from premature evaluation. Graham suggests in the comments to avoid the subshell call like this:
#!/bin/bash
VAR=17
INS='Installing $VAR'
eval "echo \"$INS\""
VAR=42
eval "echo \"$INS\""
If you don't do it:
VAR=23
INS="Installing $VAR"
echo $(eval "echo $INS")
VAR=8
echo $(eval "echo $INS")
23 is bound immediately, and the assignment VAR=8 isn't noticed.
Result:
Installing 17
Installing 42
Installing 23
Installing 23
$ OPTION="-n"
$ echo $OPTION
$
Nothing happens. I expected this.
$ OPTION="-n"
$ echo $OPTION
-n
$
Why is this?
-n is a parameter to echo, which means the trailing newline is suppressed. The fact that there's no empty line between $ echo $OPTION and the following $ means that $OPTION is properly set to -n.
If you put something else in front of $OPTION, the echo will work as you expect it to. echo only interprets words at the beginning of the command as options. As soon as it finds a non-option word ("OPTION", in this case), all words that follow are treated as literals, and not parsed as options to echo.
$ echo OPTION is set to $OPTION
OPTION is set to -n
$
Remember that the shell expands environment variables before the command is executed. Thus:
option="-n"
echo $option
becomes
echo -n ""
With the value of $option being interpreted as a parameter for the echo command. If you were using Kornshell (which is 95% similar to BASH), you could have used the builtin print command instead:
option="-n"
print -- "$option"
Unfortunately, BASH doesn't have the print command, and using the double dash in the BASH echo command will print out a double dash -- not what you want.
Instead, you'll have to use the printf command which is a bit slower than echo:
option="-n"
printf -- "$option\n" #Must include the \n to make a new line!
Of course, if you had this, you'd be in trouble:
option="%d"
printf -- "$option\n"
To get around that:
option="%d"
printf "%s\n", "$option"
By the way, you have the same problems with test:
option="-n"
if [ "$option" -eq "-n" ] #Won't work!
This is why you'll see people do this:
if [ "x$option" -eq "x-n" ] #Will work
To get the desired result, you could do this:
$ OUTPUT='-n'
$ echo -en ${OUTPUT}\\n
I'm having a problem with a shell script (POSIX shell under HP-UX, FWIW). I have a function called print_arg into which I'm passing the name of a parameter as $1. Given the name of the parameter, I then want to print the name and the value of that parameter. However, I keep getting an error. Here's an example of what I'm trying to do:
#!/usr/bin/sh
function print_arg
{
# $1 holds the name of the argument to be shown
arg=$1
# The following line errors off with
# ./test_print.sh[9]: argval=${"$arg"}: The specified substitution is not valid for this command.
argval=${"$arg"}
if [[ $argval != '' ]] ; then
printf "ftp_func: $arg='$argval'\n"
fi
}
COMMAND="XYZ"
print_arg "COMMAND"
I've tried re-writing the offending line every way I can think of. I've consulted the local oracles. I've checked the online "BASH Scripting Guide". And I sharpened up the ol' wavy-bladed knife and scrubbed the altar until it gleamed, but then I discovered that our local supply of virgins has been cut down to, like, nothin'. Drat!
Any advice regarding how to get the value of a parameter whose name is passed into a function as a parameter will be received appreciatively.
You could use eval, though using direct indirection as suggested by SiegeX is probably nicer if you can use bash.
#!/bin/sh
foo=bar
print_arg () {
arg=$1
eval argval=\"\$$arg\"
echo "$argval"
}
print_arg foo
In bash (but not in other sh implementations), indirection is done by: ${!arg}
Input
foo=bar
bar=baz
echo $foo
echo ${!foo}
Output
bar
baz
This worked surprisingly well:
#!/bin/sh
foo=bar
print_arg () {
local line name value
set | \
while read line; do
name=${line%=*} value=${line#*=\'}
if [ "$name" = "$1" ]; then
echo ${value%\'}
fi
done
}
print_arg foo
It has all the POSIX clunkiness, in Bash would be much sorter, but then again, you won't need it because you have ${!}. This -in case it proves solid- would have the advantage of using only builtins and no eval. If I were to construct this function using an external command, it would have to be sed. Would obviate the need for the read loop and the substitutions. Mind that asking for indirections in POSIX without eval, has to be paid with clunkiness! So don't beat me!
Even though the answer's already accepted, here's another method for those who need to preserve newlines and special characters like Escape ( \033 ): Storing the variable in base64.
You need: bc, wc, echo, tail, tr, uuencode, uudecode
Example
#!/bin/sh
#====== Definition =======#
varA="a
b
c"
# uuencode the variable
varB="`echo "$varA" | uuencode -m -`"
# Skip the first line of the uuencode output.
varB="`NUM=\`(echo "$varB"|wc -l|tr -d "\n"; echo -1)|bc \`; echo "$varB" | tail -n $NUM)`"
#====== Access =======#
namevar1=varB
namevar2=varA
echo simple eval:
eval "echo \$$namevar2"
echo simple echo:
echo $varB
echo precise echo:
echo "$varB"
echo echo of base64
eval "echo \$$namevar1"
echo echo of base64 - with updated newlines
eval "echo \$$namevar1 | tr ' ' '\n'"
echo echo of un-based, using sh instead of eval (but could be made with eval, too)
export $namevar1
sh -c "(echo 'begin-base64 644 -'; echo \$$namevar1 | tr ' ' '\n' )|uudecode"
Result
simple eval:
a b c
simple echo:
YQpiCmMK ====
precise echo:
YQpiCmMK
====
echo of base64
YQpiCmMK ====
echo of base64 - with updated newlines
YQpiCmMK
====
echo of un-based, using sh instead of eval (but could be made with eval, too)
a
b
c
Alternative
You also could use the set command and parse it's output; with that, you don't need to treat the variable in a special way before it's accessed.
A safer solution with eval:
v=1
valid_var_name='[[:alpha:]_][[:alnum:]_]*$'
print_arg() {
local arg=$1
if ! expr "$arg" : "$valid_var_name" >/dev/null; then
echo "$0: invalid variable name ($arg)" >&2
exit 1
fi
local argval
eval argval=\$$arg
echo "$argval"
}
print_arg v
print_arg 'v; echo test'
Inspired by the following answer.
I am confused by the usage of brackets, parentheses, curly braces in Bash, as well as the difference between their double or single forms. Is there a clear explanation?
In Bash, test and [ are shell builtins.
The double bracket, which is a shell keyword, enables additional functionality. For example, you can use && and || instead of -a and -o and there's a regular expression matching operator =~.
Also, in a simple test, double square brackets seem to evaluate quite a lot quicker than single ones.
$ time for ((i=0; i<10000000; i++)); do [[ "$i" = 1000 ]]; done
real 0m24.548s
user 0m24.337s
sys 0m0.036s
$ time for ((i=0; i<10000000; i++)); do [ "$i" = 1000 ]; done
real 0m33.478s
user 0m33.478s
sys 0m0.000s
The braces, in addition to delimiting a variable name are used for parameter expansion so you can do things like:
Truncate the contents of a variable
$ var="abcde"; echo ${var%d*}
abc
Make substitutions similar to sed
$ var="abcde"; echo ${var/de/12}
abc12
Use a default value
$ default="hello"; unset var; echo ${var:-$default}
hello
and several more
Also, brace expansions create lists of strings which are typically iterated over in loops:
$ echo f{oo,ee,a}d
food feed fad
$ mv error.log{,.OLD}
(error.log is renamed to error.log.OLD because the brace expression
expands to "mv error.log error.log.OLD")
$ for num in {000..2}; do echo "$num"; done
000
001
002
$ echo {00..8..2}
00 02 04 06 08
$ echo {D..T..4}
D H L P T
Note that the leading zero and increment features weren't available before Bash 4.
Thanks to gboffi for reminding me about brace expansions.
Double parentheses are used for arithmetic operations:
((a++))
((meaning = 42))
for ((i=0; i<10; i++))
echo $((a + b + (14 * c)))
and they enable you to omit the dollar signs on integer and array variables and include spaces around operators for readability.
Single brackets are also used for array indices:
array[4]="hello"
element=${array[index]}
Curly brace are required for (most/all?) array references on the right hand side.
ephemient's comment reminded me that parentheses are also used for subshells. And that they are used to create arrays.
array=(1 2 3)
echo ${array[1]}
2
A single bracket ([) usually actually calls a program named [; man test or man [ for more info. Example:
$ VARIABLE=abcdef
$ if [ $VARIABLE == abcdef ] ; then echo yes ; else echo no ; fi
yes
The double bracket ([[) does the same thing (basically) as a single bracket, but is a bash builtin.
$ VARIABLE=abcdef
$ if [[ $VARIABLE == 123456 ]] ; then echo yes ; else echo no ; fi
no
Parentheses (()) are used to create a subshell. For example:
$ pwd
/home/user
$ (cd /tmp; pwd)
/tmp
$ pwd
/home/user
As you can see, the subshell allowed you to perform operations without affecting the environment of the current shell.
(a) Braces ({}) are used to unambiguously identify variables. Example:
$ VARIABLE=abcdef
$ echo Variable: $VARIABLE
Variable: abcdef
$ echo Variable: $VARIABLE123456
Variable:
$ echo Variable: ${VARIABLE}123456
Variable: abcdef123456
(b) Braces are also used to execute a sequence of commands in the current shell context, e.g.
$ { date; top -b -n1 | head ; } >logfile
# 'date' and 'top' output are concatenated,
# could be useful sometimes to hunt for a top loader )
$ { date; make 2>&1; date; } | tee logfile
# now we can calculate the duration of a build from the logfile
There is a subtle syntactic difference with ( ), though (see bash reference) ; essentially, a semicolon ; after the last command within braces is a must, and the braces {, } must be surrounded by spaces.
Brackets
if [ CONDITION ] Test construct
if [[ CONDITION ]] Extended test construct
Array[1]=element1 Array initialization
[a-z] Range of characters within a Regular Expression
$[ expression ] A non-standard & obsolete version of $(( expression )) [1]
[1] http://wiki.bash-hackers.org/scripting/obsolete
Curly Braces
${variable} Parameter substitution
${!variable} Indirect variable reference
{ command1; command2; . . . commandN; } Block of code
{string1,string2,string3,...} Brace expansion
{a..z} Extended brace expansion
{} Text replacement, after find and xargs
Parentheses
( command1; command2 ) Command group executed within a subshell
Array=(element1 element2 element3) Array initialization
result=$(COMMAND) Command substitution, new style
>(COMMAND) Process substitution
<(COMMAND) Process substitution
Double Parentheses
(( var = 78 )) Integer arithmetic
var=$(( 20 + 5 )) Integer arithmetic, with variable assignment
(( var++ )) C-style variable increment
(( var-- )) C-style variable decrement
(( var0 = var1<98?9:21 )) C-style ternary operation
I just wanted to add these from TLDP:
~:$ echo $SHELL
/bin/bash
~:$ echo ${#SHELL}
9
~:$ ARRAY=(one two three)
~:$ echo ${#ARRAY}
3
~:$ echo ${TEST:-test}
test
~:$ echo $TEST
~:$ export TEST=a_string
~:$ echo ${TEST:-test}
a_string
~:$ echo ${TEST2:-$TEST}
a_string
~:$ echo $TEST2
~:$ echo ${TEST2:=$TEST}
a_string
~:$ echo $TEST2
a_string
~:$ export STRING="thisisaverylongname"
~:$ echo ${STRING:4}
isaverylongname
~:$ echo ${STRING:6:5}
avery
~:$ echo ${ARRAY[*]}
one two one three one four
~:$ echo ${ARRAY[*]#one}
two three four
~:$ echo ${ARRAY[*]#t}
one wo one hree one four
~:$ echo ${ARRAY[*]#t*}
one wo one hree one four
~:$ echo ${ARRAY[*]##t*}
one one one four
~:$ echo $STRING
thisisaverylongname
~:$ echo ${STRING%name}
thisisaverylong
~:$ echo ${STRING/name/string}
thisisaverylongstring
The difference between test, [ and [[ is explained in great details in the BashFAQ.
(Note: The link shows many examples for comparison)
To cut a long story short: test implements the old, portable syntax of
the command. In almost all shells (the oldest Bourne shells are the
exception), [ is a synonym for test (but requires a final argument of
]). Although all modern shells have built-in implementations of [,
there usually still is an external executable of that name, e.g.
/bin/[.
[[ is a new, improved version of it, and it is a keyword, not a program.
This has beneficial effects on the ease of use, as shown below. [[ is
understood by KornShell and BASH (e.g. 2.03), but not by the older
POSIX or BourneShell.
And the conclusion:
When should the new test command [[ be used, and when the old one [?
If portability/conformance to POSIX or the BourneShell is a concern, the old syntax should
be used. If on the other hand the script requires BASH, Zsh, or KornShell,
the new syntax is usually more flexible.
Parentheses in function definition
Parentheses () are being used in function definition:
function_name () { command1 ; command2 ; }
That is the reason you have to escape parentheses even in command parameters:
$ echo (
bash: syntax error near unexpected token `newline'
$ echo \(
(
$ echo () { command echo The command echo was redefined. ; }
$ echo anything
The command echo was redefined.
Some common and handy uses for brackets, parenthesis, and braces
As mentioned above, sometimes you want a message displayed without losing the return value. This is a handy snippet:
$ [ -f go.mod ] || { echo 'File not found' && false; }
This produced no output and a 0 (true) return value if the file go.mod exists in the current directory. Test the result:
$ echo $?
0
If the file does not exist, you get the message but also a return value of 1 (false), which can also be tested:
$ [ -f fake_file ] || { echo 'File not found'; false; }
File not found
$ echo $?
1
You can also simply create a function to check if a file exists:
fileexists() { [ -f "$1" ]; }
or if a file is readable (not corrupted, have permissions, etc.):
canread() { [ -r "$1" ]; }
or if it is a directory:
isdir() { [ -d "$1" ]; }
or is writable for the current user:
canwrite() { [ -w "$1" ]; }
or if a file exists and is not empty (like a log file with content...)
isempty() { [ -s "$1" ]; }
There are more details at: TLDP
You can also see if a program exists and is available on the path:
exists () { command -v $1 > /dev/null 2>&1; }
This is useful in scripts, for example:
# gitit does an autosave commit to the current
# if Git is installed and available.
# If git is not available, it will use brew
# (on macOS) to install it.
#
# The first argument passed, if any, is used as
# the commit message; otherwise the default is used.
gitit() {
$(exists git) && {
git add --all;
git commit -m "${1:-'GitBot: dev progress autosave'}";
git push;
} || brew install git;
}
Additional info about How to use parentheses to group and expand expressions:
(it is listed on the link syntax-brackets)
Some main points in there:
Group commands in a sub-shell: ( )
(list)
Group commands in the current shell: { }
{ list; }
Test - return the binary result of an expression: [[ ]]
[[ expression ]]
Arithmetic expansion
The format for Arithmetic expansion is:
$(( expression ))
The format for a simple Arithmetic Evaluation is:
(( expression ))
Combine multiple expressions
( expression )
(( expr1 && expr2 ))
Truncate the contents of a variable
$ var="abcde"; echo ${var%d*}
abc
Make substitutions similar to sed
$ var="abcde"; echo ${var/de/12}
abc12
Use a default value
$ default="hello"; unset var; echo ${var:-$default}
hello