Re: [pdf-devel] Bug in LZW filter?

[Georg Gottleuber]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

I ended up with this encoder structure:

/**************** code fragment ***************************/
while (!pdf_buffer_eob_p (in) && !pdf_buffer_full_p (out))
  {
    /* accumulate input data */
    st->string.suffix = in->data [in->rp++];
    if (!lzw_dict_find (&st->dict, &st->string))
      {
        /* emit output code of longest input sequence in dict */
        lzw_buffer_put_code (&st->buffer, st->string.prefix);

        /* is current bitsize enough? */
        if (st->dict.size - 1 == st->buffer.maxval - st->early_change)
          {
            if (!lzw_buffer_inc_bitsize (&st->buffer))
              {
                lzw_buffer_put_code (&st->buffer, LZW_RESET_CODE);
                lzw_buffer_set_bitsize (&st->buffer, LZW_MIN_BITSIZE);
                lzw_dict_reset (&st->dict);
                dict_reset = PDF_TRUE;
              }
          }
        /* create new dict entry */
        if(!dict_reset)
          lzw_dict_create_entry (&st->dict, &st->string);
        else
          dict_reset = PDF_FALSE;

        st->string.prefix = st->string.suffix
      }
  }
/**************** code fragment ***************************/

Bitsize increase should happen as late as possible (as mentioned in
en.wikipedia.org). This is why it is tested right before creating a
new table entry.

Unit test results are positive for this code (see attached patch).

If you agree with this encoder version I will refactor the decoder in
the same way.

Regards,
Georg

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk7XhEUACgkQ5sLITM1qIaINcACePiqwSZ1iodSIYmuCJBtLDaAK
bsMAoIhZHj6I+IItrmCugFfj/ksBWUcy
=HQFB
-----END PGP SIGNATURE-----

lzw_enc_refactor.patch
Description: Text Data