#! /bin/gawk -f
# Last edited on 2025-04-29 20:45:02 by stolfi

BEGIN {
  abort = -1;
  usage = ( \
    "cat RAW.evt \\\n" \
    "  | evt-to-wds \\\n" \
    "  | fix-raw-words \\\n" \
    "      -f FIXFNS.gawk \\\n" \
    "      -v sample=SAMPLE \\\n" \
    "      [ -v field=FIELD ] \\\n" \
    "      [ -v table=TABLE ] \\\n" \
    "  > GUD.wds" \
  );
  
  # Performs any adjustments to words extracted from text sample
  # SAMPLE needed for proper statistical analysis. 
  # This may include a change of encoding, removing capitalization,
  # elimination of selected words, etc.
  # 
  # The output is the "raw" word list: it should still include all
  # pronounceable words, numerals, and symbols, but should exclude
  # punctuation and other silent marks.
  
  if (sample == "") { arg_error("must define \"sample\""); }
  if (table == "") 
    { split("", wmap); }
  else
    { # Read word-remapping table, if present.
      # To discard a word from the "raw" susbset, 
      # map it to the string "*DELETE*".
      load_remapping_table(table);
    }
  if (field == "") { field = 0; }
  
  nread = 0;  # Number of words read
  nwrite = 0; # Number of words written
}

(abort >= 0) { exit abort; }

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

/./ {
  nread++;
  word = $(field);
  if (word in wmap) 
    { word = wmap[word];
      if ((word == "*DELETE*") || (word == "*delete*")) { next; }
    }
  word = fix_raw_word(word);
  if ((word == "*DELETE*") || (word == "*delete*")) { next; }
  nwds = split(word, wds, "\n");
  for(i = 1; i <= nwds; i++)
    { $(field) = wds[i]; print; nwrite++; }
  next;
}

END {
  if (abort >= 0) { exit abort; }
  printf "%s: %7d words read, %7d written\n", sample, nread, nwrite > "/dev/stderr";
}

function load_remapping_table(file,    nMap,lin,fld,nfld)
{
  # Reads a word mapping table from "file", containing pairs 
  # of the form ORGINAL NEW. 
  # Stores the table in "wmap[ORIGINAL] = NEW".
  
  nMap=0;
  split("", wmap)
  ERRNO = ""
  while((getline lin < file) > 0) { 
    gsub(/^[ ]*/, "", lin);
    if (! match(lin, /^([#]|$)/))
      { gsub(/[ ]*[#].*$/, "", lin);
        nfld = split(lin, fld, " ");
        if (nfld != 2) tbl_error(file, ("bad table entry = \"" lin "\""));
        if (fld[1] in wmap) tbl_error(file, ("repeated key = \"" lin "\""));
        wmap[fld[1]] = fld[2];
        nMap++;
      }
  }
  if (ERRNO != "") { arg_error((file ": " ERRNO)); }
  close (file);
  if (nMap == 0) 
    { printf "warning: file \"" file "\" empty or missing\n" > "/dev/stderr"; }
  else
    { printf "loaded %6d map pairs\n", nMap > "/dev/stderr"; }
}

function arg_error(msg)
{
  printf "%s\n", msg > "/dev/stderr";
  printf "usage: %s\n", usage > "/dev/stderr";
  abort=1;  exit 1;
}

function data_error(msg)
{
  printf "line %d: %s\n", FNR, msg > "/dev/stderr";
  abort = 1; exit 1;
}

function tbl_error(file, msg)
{
  printf "file %s, line %s: %s\n", file, FNR, msg > "/dev/stderr";
  abort = 1; exit 1;
}