#! /usr/bin/gawk
# Last edited on 2026-01-12 17:16:12 by stolfi
# Common functions for scripts of Notes/077.
# To be included in other gawk scripts.
  
function extract_words(raw_lin) { 
  # Removes the locus ID from {raw_lin}, all leading and trailing
  # punctuation chars, and replaces each internal string of one or more
  # punctuation chars with ' '.
  lin = raw_lin;
  gsub(/^[<][0-9A-Za-z.;]+[>][ ]*/, "", lin);
  lin = tolower(lin);
  gsub(/[=.,' ()]+/, " ", lin);
  gsub(/^[ ]+/, "", lin);
  gsub(/[ ]+$/, "", lin);
  gsub(/[ ][ ]+/, " ", lin);
  return lin;
}