psjson

#! /usr/bin/env bash
# Use at own risk.  License CC0 see https://creativecommons.org/publicdomain/mark/1.0/
# alanberezin.com

OPTIND=1         # Reset in case getopts has been used previously in the shell.
opt_stripclasspath=false
while getopts ":hc" opt; do
    case "$1" in
    -h | --help )
        cat <<-EOF
        Usage: ps [any-ps-options] | psjson [-c] [repeat-ps-options-see-below]
           or: docker ps [any-docker-ps-options] | psjson [-c] 
           or: psjson -h

        Reading the stdin, jsonify the output of ps (or docker ps). If you pipe the contents of "ps" 
        directly into psjson, then you do not need to repeat the ps options to psjon.
        However, if you pipe the output of ps though grep, or anything else, that strips the
        header off the output, then you need to repeat the ps options (like, -f) that would 
        have the effect of changing the default columns that are output Also, if you pipe
        though grep or other commands that colorize, use the --color=never option to grep.

        -c    For ps(1) rows that have a "-cp" in the command string, remove the following
              token.  This effectively strips nuesance long java classpaths that clutter the output.
            

        A nice feature of psjon is that it tokenizes the CMD column of ps. this is handy when, for example, the process is a java server with a mile long classpath.

        For docker ps, the parsing of CREATED and STATUS needs work

        Examples:
          Correct:    ps -f | grep foo | psjson -f
          Correct:    ps -ef | grep foo | psjson -f  #dont need to repeat the -e flag 
                                                     #it doesnt effect the header
          Correct:    ps -f | psjson 
          Correct:    ps -f | psjson -f    #final -f is ignored
          Incorrect:  ps -f | grep --color=never foo | psjson  #wrong unless foo was a column header in ps
          marginal:  ps | grep foo | psjson  #should use grep --color=never to be safe

        This cmd requires that jq be installed.  See https://stedolan.github.io/jq/manual/v1.4/
EOF
        exit 0
        ;;
    -c )
        opt_stripclasspath=true
        shift
        ;;
    esac
done

#In case output of ps has no header (like if it was grep'd out)
#Note: test for header is an imperfect test which relies on PID in the header, maybe improve
tmpfile=$(mktemp)
trap 'rm -f $tmpfile' EXIT
cat > $tmpfile
# read -e -p 'debug> ' < /dev/tty
if cat "$tmpfile" | head -1 | grep --color=never ' PID ' > /dev/null ; then
    : 
elif cat "$tmpfile" | head -1 | grep --color=never 'CONTAINER' > /dev/null ; then
  # A docker ps output.  Reqrite header to avoid whitespace.
  #TODO ceated and status fields still have embedded whitespace
    header="CONTAINER_ID IMAGE COMMAND CREATED STATUS PORTS NAMES"
    sed -i '1s/^/'"$header"'\n/
    2d' $tmpfile
else 
    header="$(ps $@ | head -1)"
    sed -i '1s;^;'"$header"'\n;' $tmpfile
fi

arg_separator='_-_-'  #something unique bc we may need to split on it later
# Credit to https://unix.stackexchange.com/a/243485/103690 for the basic idea.
jq='
[sub("\n$";"") | splits("\n") | sub("^ +";"") | [splits(" +")]] 
| .[0] as $orig_header 
| [$orig_header[], (range(1000) | tostring | "arg'$arg_separator'"+.)] as $newheader 
| .[1:] | [.[] | [. as $x | range($newheader | length) 
| {"key": $newheader[.], "value": $x[.]}] | from_entries]
| del(.[][] | select(. == null))
'
if [ "$opt_stripclasspath" = true ] ; then
    jq=$jq'

      | .[] | . as $process | keys[] | select($process[.] == "-cp") 
      | split("_-_-")[1] |  tonumber as $argnumber | $process 
      | del( .["arg_-_-" + (($argnumber+1) | tostring)]) | [.]'
fi
echo $jq > jq.tmp
cat $tmpfile | jq -sR "$jq"

convert ps output to JSON and tokenize the CMD column into separate key/val pairs.
This also works somewhat for docker ps