Line number efficiency

[Hans Aberg]

One should note that under Unicode, there are 2^N character positions, N =
20, 24, or even 32. So in the Flex skeleton segment
  yy_match:
    do {
      register YY_CHAR yy_c = yy_ec[YY_SC_TO_UI(*yy_cp)];
      ...
the yy_ec[...] will in most cases be replaced with a compressed table,
which will add additional overhead.

One may then slip in the line number count here, as a side effect in the
yy_ec() function call. Depending on how yy_ec() looks like, this may not
cause much overhead.

  Hans Aberg