Determine if .deb package is installed using Bash

I was able to use dpkg-query to determine if a package is installed on Ubuntu, but it’s a bit of a mess because the name used to install the package doesn’t always align with what it’s installed as. For instance, gnutls-dev is installed under libgnutls-dev…

Any suggestions on how I could clean this up a bit?

function isInstalled() {
    if dpkg-query -W -f'${Status}' "$1" 2>/dev/null | grep -q "ok installed" || dpkg-query -W -f'${Status}' "lib$1" 2>/dev/null | grep -q "ok installed"; then
        echo 1;
    fi

    echo 0
}

It seems odd that the function prints 0 in all cases, including when 1 has been printed.

For a utility like this, it’s probably better to communicate via the exit status:

isInstalled() {
    !   dpkg-query -W -f'${Status}' "$1" 2>/dev/null | grep -q "ok installed" 
     || dpkg-query -W -f'${Status}' "lib$1" 2>/dev/null | grep -q "ok installed"
}

I’d be inclined to remove the !, returning zero (success) if the package is found.

If you have aptitude available, it’s possible to ask it to search for any installed package of the given name, or providing the name:

aptitude -q2 search "~i~P?exact-name($1)|~i?exact-name($1)" >/dev/null

• ~i: installed package
• ~P: provides name
• |: alternation
• ^$1$
• ?exact-name: exactly what it says

This is less brittle, as it doesn’t rely on the providing package being formed by prepending lib to the virtual package name. For example, with www-browser, it finds all installed web browsers. Also, it avoids matching superstrings of the supplied name.

If you know that $1 will always be the name of a virtual package, you can of course use just the ~P side of the alternation.

the name used to install the package doesn’t always align with what it’s installed as. For instance, gnutls-dev is installed under libgnutls-dev

That’s because they are different things:

• gnutls-dev is a single command/utility
• libgnutls-dev is a package

Note that one single package can contain various commands. I’d say the most notable cases are coreutils and util-linux.

So to check if a command or package is installed, you’d need to follow different approaches.

For example (assuming a recent version of Bash):

check_command() {
  local 'command' 'not_found'
  for command; do
    if ! type -- "${command}" > '/dev/null' 2>&1; then
      printf '%sn' "Command not found: ${command}"
      (( not_found++ ))
    fi
  done
  if (( not_found > 0 )); then
    printf '%sn' "Missing commands: ${not_found}"
    return '1'
  fi
}

check_package() {
  local 'not_found' 'package' 'packages'
  packages="$( dpkg --get-selections | cut -f '1' | sort )"
  for package; do
    if ! grep -P -e "^${package}$" <<< "${packages}" > '/dev/null' 2>&1; then
      printf '%sn' "Package not found: ${package}"
      (( not_found++ ))
    fi
  done
  if (( not_found > 0 )); then
    printf '%sn' "Missing packages: ${not_found}"
    return '1'
  fi
}

Then:

# cat and lsblk are commands
# coreutils and util-linux are packages

$ check_command cat coreutils lsblk util-linux
Command not found: coreutils
Command not found: util-linux
Missing commands: 2

$ check_package cat coreutils lsblk util-linux
Package not found: cat
Package not found: lsblk
Missing packages: 2