This is an implementation of the intersection of a CF grammar and a FS
automaton, as described by Yehoshua Bar-Hillel et al., "On formal
properties of simple phrase structure grammars", pp. 116-150, 1964.


The input consists of 1+n sections. The first one describes the CFG
and the follwing ones describe FSAs.  Each section is terminated by a
line with a single dot in it.

The input is line-oriented.  A long line can be broken by a backslash
at the end of a line; backslashes in any other position are just
backslashes.  Comments start with a #, and continue to end of line.
There is no way to use # as a symbol.

The CFG section starts with a line specifying a list of one or more
start symbols.  The further lines are production rules of the form

	N -> A b C ...

Symbols may contain anything but layout; the -> is treated as a
symbol.  Terminals and nonterminals are not formally distinguished.
Any symbol defined in the grammar is a nonterminal, any symbol used
but not defined is a terminal.  The program reports which symbols it
has identified as terminals and nonterminals.

A FSA section starts with two lines, the first line specifying one or
more start states and the second line one or more end states, as
unsigned integers.  The further lines each contains a transition of
the form

	state#1 x... state#2

where x is any symbol and x... is a non-empty sequence of such
symbols.  x can be a nonterminal defined in the grammar, in which case
it just means that the input described by the FSA can contain
nonterminals.

A sequence of more than one x implies anonymous states, the ones
between the various xs.  The state numbers chosen for these start at
one above the highest specified state.  This may be unfortunate, and
it is sometimes handy to use a start state 1 and an end state 0, to
remedy this.

The symbol ? has a special meaning in the FSA section and stands for
all terminals in the grammar.  It cannot be used in the grammar.
There is no way to use ? itself as a symbol.

The FSA need not be deterministic but must be epsilon-free.  Modifying
the algorithm to accept epsilon moves would not be not trivial.

The grammar may contain terminals that do not occur in the FSA and vice
versa, but such terminals will (of course) be eliminated from the
intersection grammar.

When the FSA does not restrict the CFG at all, one would probably want
the program to finally yield the original grammar again.  This does
not happen, because the state numbers remain attached to the symbols
in the output grammar.  The -r option tries to remedy this, by
substituting back the generated rules for the start symbol and the
terminal symbols.  The renaming of the other symbols remains,
unfortunately.

The option -v switches on the verbose mode.  

Some variable-sized data is kept in arrays, but the grammar nodes and
the strings are allocated dynamically; all allocation is checked.



make GRAMMAR=gramX test_cfg-fsa
	runs a test on gramX