#! /usr/bin/gawk
# Last edited on 2026-01-12 17:00:33 by stolfi
# Common functions for scripts of Notes/076.
# To be included in other gawk scripts.
  
function extract_words(raw_lin) { 
  # Removes the locus ID such as "<f...>" from {raw_lin}, all leading and trailing
  # punctuation chars, and replaces each internal string of one or more
  # punctuation chars with ' '.
  lin = raw_lin;
  # Remove locus ID and leading blanks:
  gsub(/^[<][0-9A-Za-z.]+[>][ ]*/, "", lin);
  if (match(lin, /^[<]/)){ data_error("malformed locus ID"); }
  # Remove inline comments:
  gsub(/[<][!][^<>]*[>]/, "", lin);
  # Remove special IVTFF markers:
  gsub(/[<][-~$%][>]/, "", lin);
  if (match(lin, /[<>]/)){ data_error("malformed comment or '<...>' marker"); }
  if (match(lin, /[@*%$()]/)){ data_error("invalid char in text"); }
  # Map to lowercase:
  lin = tolower(lin);
  # Replace strings of word delimiters by space:
  gsub(/[-<>=«».,' ()]+/, " ", lin);
  # Remove leading, trailing, and superfluous spaces:
  gsub(/^[ ]+/, "", lin);
  gsub(/[ ]+$/, "", lin);
  gsub(/[ ][ ]+/, " ", lin);
  return lin;
}