mjhjmtu

Hello i am trying to parse the result of du command but i don't know how to do it without looping
i execute the below command

du -shm /var/vmail/mailboxes/domain/*/mail

my result is a list e.g.:

80 /var/vmail/mailboxes/domain/USER1/mail 
150 /var/vmail/mailboxes/domain/USER2/mail 
220 /var/vmail/mailboxes/domain/USER3/mail 

The above list i want to transform it to json object with the below:

USER1: 80, USER2:150, USER3:220 

or at least to have something:

80 USER1
150 USER2
220 USER3

I tried with sed -e but without success, any ideas?

du -shm /var/vmail/mailboxes/domain/*/mail | awk -F '[ /]' ' print $1, $7 '

With the input from du in the format that you give, this would produce

80 USER1
150 USER2
220 USER3

The awk program simply picks out the appropriate fields from the output of du. It interprets each line as fields delimited by either spaces or slashes.

awk -F '[ /]' ' print $1, $(NF - 1) ' would also work.

To get this into JSON, using jq:

du -shm /var/vmail/mailboxes/domain/*/mail |
awk -F '[ /]' ' print $1, $7 ' |
jq -sR 'split("n")[0:-1] | map(split(" ")) | map((.[1]):.[0]) | add'

This would produce

 "USER1": "80",
 "USER2": "150",
 "USER3": "220"

The jq expression first splits the input into separate array elements (one per line of input), and then splits these again on spaces into subarrays. At this point we have

[
 [
 "80",
 "USER1"
 ],
 [
 "150",
 "USER2"
 ],
 [
 "220",
 "USER3"
 ]
]

After that, objects are created with the second element in each subarray as the key and the first as the value:

[
 
 "USER1": "80"
 ,
 
 "USER2": "150"
 ,
 
 "USER3": "220"
 
]

The add at the end gives us the final result.

try

du -sm /var/vmail/mailboxes/domain/*/mail |
 sed 's:/var/vmail/mailboxes/domain/::; s:/mail::'

where

• 
  s:/mail:: will replace /mail by nothing

to go to json

du -sm /var/vmail/mailboxes/domain/*/mail |
 sed 's:/var/vmail/mailboxes/domain/::; s:/mail::' |
 awk 'BEGIN sep="" ; printf "%s "%s":%s",sep,$2,$1; sep="," END printf "n"'

You can use awk to convert to 2-column format:

$ du -shm | awk -F/ 'print $1" "$(NF - 1)'
80 USER1
150 USER2
220 USER3

For JSON, try:

$ du -shm | awk -F/ 'print """$(NF - 1)"":"$1' | paste -sd, | printf "$(cat)" | jq

 "USER1": 80,
 "USER2": 150,
 "USER3": 220

jq does pretty well with raw-text, i.e. the -R flag, e.g.:

du -sm /var/vmail/mailboxes/domain/*/mail |
jq -R '
 split(" ") |
 .[0] as $size | # Remember size 
 .[1] | split("/") | .[-2] as $name | # Extract username
 ($name) : ($size) # Compose object
' 

Gives you:

 "USER1": "80"


 "USER2": "150"


 "USER3": "220"

To merge the objects, use a second jq invocation:

... | jq -s add

Output:

 "USER1": "80",
 "USER2": "150",
 "USER3": "220"

Or in compact mode:

... | jq -sc add

Output:

"USER1":"80","USER2":"150","USER3":"220"

This is a bit tricky to do with sed. However, here is a commented suggestion that works with GNU sed:

parse.sed

s|([^ ]+) .*/([^/]+)/.*| "2" : 1,| # Extract size and name in json-format
1h; 1!H # Correctly add them to hold-space
$ 
 z # Clear pattern-space
 s/^// # Prepend start-curly-brace
 G # Fetch formatted json from hold-space
 s/,$// # Remove last comma
 s/$/n/ # Append end-curly-brace
 p # Print

Run it like this:

du -shm /var/vmail/mailboxes/domain/*/mail | sed -Enf parse.sed

Or as a one-liner:

du -shm /var/vmail/mailboxes/domain/*/mail |
sed -Ene 's|([^ ]+) .*/([^/]+)/.*| "2" : 1,|;1h;1!H;$z;s/^//;G;s/,$//;s/$/n/;p'

Output in both cases:

 "USER1" : 80,
 "USER2" : 150,
 "USER3" : 220