#! /usr/bin/gawk -f
# Last edited on 2026-07-03 17:33:07 by stolfi

# Deletes from an SBJ pinyin file those parts that apparently
# were systematically omitted by the VMS author.
# Assumes each line starts with a locus ID <b...> and
# keywords like [

BEGIN {
  printf "doing n055_delete_omitted_bencao_fields.gawk ...\n" > "/dev/stderr"
}

/^ *([#]|$)/ { next; }

/^[<]/ {
  loc = $1; $1 = ""
  gsub(/\[ *zhǔ +zhì *\]/, "[zhǔ]", $0)
  gsub(/\[ *wèi *\][^][]*[.,]/, " ", $0)
  gsub(/\[ *shēng *\][^][]*[.,]/, " ", $0)
  gsub(/\[ *yī +míng *\] [^][,.]*[,.]/, " ", $0)
  gsub(/[(][^][()]*[)]/, " ", $0)
  
  gsub(/[<][%$][>]/, " ", $0)
  
  gsub(/[][,.:;!*'-]/, " ", $0)
  
  $0 = (loc " " $0)

  gsub(/[>] */, "> ", $0)
  gsub(/[ ][ ]+/, " ", $0)
  gsub(/[ ]*$/, " ", $0)
  print
  next
}

// { data_error("invalid line format") }