Syntax Question for Bash using pipes

Question

I am trying to set a variable to a value returned by a program (lynx). I cannot get past which parenthesis or bracket I use to accomplish my goal.

I have this:

DEBFILE=${(lynx -listonly -dump https://gridtracker.org/release.html) | grep .deb | grep nwjs | awk '{ print $2 }'}

Where am I going off the rails. Be easy on me. I haven't messed with this in a while.

I simplified the statement to include the URL. Ultimately I would like to use a variable for the URL in that code block as well.

such as...

DEBFILE=${(lynx -listonly -dump ${GT_DOWNLOAD_LINK} ) | grep .deb | grep nwjs | awk '{ print $2 }'}

as it is defined already near the beginning of my script.

I know it is a syntax problem because I broke it down into a few more lines, echo'd the command to a file, and read it back into a variable that way.

THIS works

echo $(lynx -listonly -dump https://gridtracker.org/release.html | grep .deb | grep nwjs) | awk '{ print $2 }' >> temp.txt
        DEBFILE=$(cat temp.txt)
        echo $DEBFILE

Answer 1

You indeed have nested syntax errors, probabily arising from a confusion about variable expansion and command substitution.

What you want to achieve is command substitution, i.e. you want to assign to the variable DEBFILE the output of a command. This is done as follows:

DEBFILE=$( your command )

The syntax ${ ... } you used in your original attempt is instead parameter/variable expansion, where you "dereference" the shell variable whose name is placed in the curly braces. Note that ${VARIABLE} is basically the same as $VARIABLE, so for the first command you showed, your shell would try to look up a variable whose name is (lynx -listonly -dump https://gridtracker.org/release.html) | grep .deb | grep nwjs | awk '{ print $2 }' and assign its value to DEBFILE.⁽¹⁾

The correct way to achieve your goal should simply be

DEBFILE=$(lynx -listonly -dump https://gridtracker.org/release.html | grep .deb | grep nwjs | awk '{ print $2 }')

and you could probably replace the grep-grep-awk pipeline by a single awk command, which would probably look like

DEBFILE=$(lynx -listonly -dump https://gridtracker.org/release.html | awk '/\.grep/ && /nwjs/ { print $2 }')

or

DEBFILE=$(lynx -listonly -dump https://gridtracker.org/release.html | awk '(index($0,".deb") && index($0,"nwjs")) { print $2 }')

Note that the code you labelled as "This works" basically does exactly what I proposed, just with an unnecessary echo and the intermediate step of writing the output to a file from which it is read back to the variable $DEBFILE:

• echo $(lynx ... | grep .deb | grep nwjs) has the same effect as lynx ... | grep .deb | grep nwjs (but could possibly introduce unwanted reformatting of whitespace) because it basically makes the console output of the lynx ... | grep ... | grep ... pipeline an argument to echo, which will simply print it to the console (again).
• So, echo $(lynx ... | grep .deb | grep nwjs) | awk '{ print $2 }' does the same as lynx ... | grep .deb | grep nwjs | awk '{ print $2 }'

---

⁽¹⁾ As noted by Stéphane Chazelas, ksh93 actually does support command substitution in a ${ cmd; } form, but it requires the command to end in a ;, it is not POSIX-compatible, and Bash doesn't support it.