#! /usr/bin/gawk -f
# Last edited on 2025-11-06 14:28:45 by stolfi

# Reads from stdin a file in the "main.wds" format of the langbank project.
# Writes to stdout one word per line, omitting all punctuation 
# except "=".

BEGIN { skip = 1; nread = 0; nwords = 0; npuncts = 0}

// { nread++ }

{hb}{h3}{tx}{sD}{txs}

/^[$] {hb}{h[0-9]*}{tx}{s[DV]}{txs} *$/ { skip = 0; next }
/^[$] / { skip = 1; next; }
(skip == 1){ next }

/^a / { print tolower($2); nwords++; next; }

/^p =$/ { print $2; npuncts++; next; }

END { 
  printf "read %d lines wrote %d words and %d punctuation\n", \
    nread, nwords, npuncts > "/dev/stderr"
}