Extract all text between two directory entries in a file - regex

Question

I have a file that documents the structure of several directories. I am trying to print the text for each directory individually. My input file looks like this:
$ cat file.txt
/bin:
file_1
file_2
file_3
/sbin:
file_a
file_b
file_c
/usr/local/bin:
doc_a
doc_b
doc_c
What I'm trying to do is print a specific section of the file based on user selection:
#!/bin/bash
PS3=$'\nMake a selection '
select dir in $(grep ':' file.txt;) do
case $REPLY in
[0-9]) echo $dir
# Need something here. Maybe a pcregrep regex?
# pcregrep '(<= $dir)*(some_fancy_regex)' file.txt
break;;
esac
done
The user is presented with menu options:
1) /bin:
2) /sbin:
3) /usr/local/bin:
Make a selection
Suppose the user chooses 2. Currently, this just prints the chosen directory on the screen. I would like to display the directory as well as the files it contains.
/sbin:
file_a
file_b
file_c
From what I've read it seems like a pcre regex would work here. I barely understand non-pcre style regex. I'm trying to wrap my brain around positive and negative lookahead & lookbehind but I really don't know what I'm doing yet. If someone could help me figure this out I would appreciate it.
All directories begin with a / and end with :
File names listed under each directory may contain:
[a-z], [A-Z],[0-9]
Literal characters . _ - [
All directory / file structures end with a blank empty line

Answer 1

It can be done purely in bash 4 in a single pass without using any external tool. Here is the script to solve this problem:
#!/bin/bash
# declare an associative array
declare -A dirs=()
# loop thru all lines and populate our associate array
# with dir as key and \n separated file names as value
while read -r; do
[[ -z $REPLY ]] && continue
if [[ $REPLY == *: ]]; then
d="$REPLY"
else
dirs["$d"]+=$'\n'"$REPLY"
fi
done < file.txt
# present a menu to customer and print selected dir name with file names
select dir in "${!dirs[#]}"; do
if [[ -n $dir ]]; then
printf '%s%s\n' "$dir" "${dirs[$dir]}"
break
fi
done
Output:
1) /usr/local/bin:
2) /bin:
3) /sbin:
#? 1
/usr/local/bin:
doc_a
doc_b
doc_c
and this:
1) /usr/local/bin:
2) /bin:
3) /sbin:
#? 3
/sbin:
file_a
file_b
file_c

Answer 2

With GNU sed and bash:
dir="/usr/local/bin:"
sed -n "/${dir//\//\/}/,/^$/{/^$/d;p}" file
With bash:
dir="/usr/local/bin:"
while IFS= read -r line; do
[[ $line == $dir ]] && switch=1
[[ $line == "" ]] && switch=0
[[ $switch == 1 ]] && echo "$line"
done < file
Output in both cases:
/usr/local/bin:
doc_a
doc_b
doc_c

Answer 3

Don't mistake shell for a text-processing tool, that's what awk is for. All you need are these 4 lines:
$ cat tst.sh
awk -v RS= -F'\n' -v OFS=') ' '{print NR, $1}' file.txt >&2
printf '\nMake a selection: ' >&2
IFS= read -r rsp
awk -v RS= -v nr="$rsp" 'NR==nr' file.txt
$ ./tst.sh
1) /bin:
2) /sbin:
3) /usr/local/bin:
Make a selection: 2
/sbin:
file_a
file_b
file_c

Answer 4

Grep is not to best tool to do this because it is line-oriented; you can't really have grep look at expressions that span multiple lines, except with some contortion – and that -z option is not specified by POSIX.
You could do it like this:
#!/bin/bash
PS3=$'\nMake a selection '
mapfile -t opts < <(grep ':' file.txt)
select dir in "${opts[#]}"; do
sed -n "\#$dir#,/^$/{/^$/q;p}" file.txt
break
done
First, I've changed your menu creation. Notice that you have a spare semicolon within the command substitution and a missing one after it; using grep like this would also break if there are spaces in the directory names. I've thus used mapfile to get all the lines containing : into an array.
Then, once I know about the directory, I use sed to print "from the directory name on until the next empty line". That would simply be
sed -n "/$dir/,/^$/p"
but this falls short on multiple fronts. First of all, the directory name can contain slashes, which trips up the / delimited addressing. We can use \%regexp% instead of /regexp/, where % can be any character; I've chosen #.
Now, we have
sed -n "\#$dir#,/^$/p"
That's almost there, but prints trailing blank lines; we suppress that by using {/^$/q;p} instead of just p, which says "if the line is blank, quit, else print it".
Sample output (edited to use a directory name with a space):
1) /bin blah:
2) /sbin:
3) /usr/local/bin:
Make a selection 1
/bin blah:
file_1
file_2
file_3
Remark: Non-GNU seds (like the one found in macOS) might complain about the two commands in curly braces; using {/^$/q;p;} instead (extra semicolon) might help.