pdfbadwordpct

#!/usr/bin/env bash

#
# 20201108 fwmechanic wrote
#
# stripped down version of %~dp0/pdfcreatesearchable
#

die() { printf %s "${@+$@$'\n'}" 1>&2 ; exit 1 ; }
see() ( { set -x; } 2>/dev/null ; "$@" ) ;

optq=0

chk_command() { command -v "$1" > /dev/null || die "command '$1' needs to be installed" ; }
chk_command "pdftotext"
chk_command "hunspell"

uwords() ( pdftotext -nopgbrk "$1" - | tr ' ' '\n' | sort -u )

# badwordpct kept in lockstep sync with %~dp0/pdfcreatesearchable

badwordpct() ( tgt="$1" base="$2"
   uwct="0" bwct="0" bwpct="0"
   uwfnm="$base.uwords" ; uwords "$tgt" > "$uwfnm"
   if [ -s "$uwfnm" ]; then
      uwct="$(< "$uwfnm" wc -l)"
      h_opts=()
      if [[ "$MY_DICTIONARY" ]]; then  # ref: "burns" COMMENT to https://stackoverflow.com/a/50776529
         h_opts=( "-p" "$MY_DICTIONARY" )
      fi
      # $uwfnm words include punct which hunspell strips (leading to $bwfnm containing dups); `hunspell | sort -u` would make uwct and bwct become apples & oranges ...
      bwfnm="$base.badwords" && hunspell -d en_US "${h_opts[@]}" -l "$uwfnm" > "$bwfnm" &&
      bwct="$(<"$bwfnm" wc -l)" &&
      bwpct=$(( (bwct * 100) / uwct ))
      printf "%5d %5d %3d %s\n" "$uwct" "$bwct" "$bwpct" "$tgt"
   fi
 # rm -f "$uwfnm" "$bwfnm"
   rm -f "$uwfnm" # basis for updating "$MY_DICTIONARY"
   )

do1file() (
   fnm="$1"
   if [[ ! -f "$fnm" ]]; then
      ((optq!=0)) || echo "$fnm does not name a file"
      return
   fi
   # break down $fnm:
   fpath=""
   if [[ $fnm == */* ]]; then
      fpath="${fnm%/*}/"
      cd "$fpath" || die "cd to $fpath failed"
   fi                                   # ; echo "fpath=$fpath'"
   # note that because we cd above, remaining filenames used are sanspath
   sanspath="${fnm##*/}"                # ; echo "sanspath=$sanspath'"    # https://stackoverflow.com/a/965069
   ext="${sanspath##*.}"                # ; echo "ext=$ext'"
   sansext="${sanspath%.*}"             # ; echo "sansext=$sansext'"

   # logical parameters:
   badwordpct "$sanspath" "$sansext"
   )

# parse options
argerr() {  die "usage: $0 [-q] FILE..." ; }
while getopts "hq" opt; do
  case $opt in
    q) optq=1 ;;
    h) argerr;;
    \?) argerr;;
  esac
done
shift "$((OPTIND-1))"  # shift so that $@, $1, etc. refer to the non-option arguments

for fnm in "$@"; do
   do1file "$fnm"
done