debian dependency tree

apt-rdepends-tree

#! /bin/bash
# Description: show dependency tree
# Author: damphat

if [ $# != 1 ]; then
	echo 'Usage: apt-rdepends-tree <package>'
	echo 'Required packages: apt-rdepends'
	exit 1  
fi

# tree template
T1="    ├─"
T2="    │    ├─"
T3="    │    └─"

# tree template for last node
T4="    └─"
T5="         ├─"
T6="         └─"

# mark '1' for parent node, '2' for child node
TEXT="$(apt-rdepends $1 | sed -e 's/^/1 /' -e 's/.*: /2 /'; echo '-')"
TOTAL=$(echo "$TEXT" | grep '^1' | wc -l) # total parent

COUNT=0
echo "$TEXT" | while read line; do
	tmp=$last
	[ "${line:0:1}" != "${last:0:1}" ] && tmp=$(echo $last | sed -e 's/^2/3/')

	[ "${tmp:0:1}" == "1" ] && ((COUNT++))

	if [ "$TOTAL" != "$COUNT" ]; then
		echo $tmp | sed -e "s/^1/$T1/" -e "s/^2/$T2/" -e "s/^3/$T3/"
	else
		echo $tmp | sed -e "s/^1/$T4/" -e "s/^2/$T5/" -e "s/^3/$T6/"
	fi

	last=$line
done

MestreLion Hi, Very nice script, thank you. The only missed things we strongly need are:

Thanks @hiddenman , glad you liked!
Currently I'm unable to dive into this, but I can share a few thoughts about each, it may help:

1. Include build-depends for packages, not only Depends

Is there a such an option in apt-rdepends? If so, this is possible.

Also, I believe it should either print the build-depends OR the Depends (controllable by a new --build-depends flag), as there are very few (if any) use-cases for a combined list including both. And whoever needs both can always run it twice, with and without the flag.

2. Mark somehow those packages which are not available (using apt-cache maybe). For example, if "gcc-8-base" is missed in the local apt repository, then show this something like gcc-8-base (-), those which exist, mark as (+)

This might be feasible, possibly best done in parse_deps(). I suggest the mark be (!) for not available, and no mark if available.

3. Print this graph as a plain-text to be able to iterator over it and build each package from the list

A new --plain flag can easily do that, perhaps using tr -d to remove the UTF (and whitespace?) characters. Indentation can be replaced with TABs, or completely stripped if --no-indent, which would imply --plain.

Will try to implement this by myself. May be you consider implementing this :)

I will. And if you do, it would be amazing if you could open a PR in my repo!!!
(you can also open a new issue there so we can best discuss the above points instead of polluting the this gist)