Installation instructions:

Make a copy of the most recent bleedperl.
You can find that at:
        rsync://ftp.linux.activestate.com/perl-current/ 
Apply the enlosed patch file on it.  If the patch doesn't apply, let
me know at once.

Unpack the Rx-0.50.tar.gz distribution.  

Run the Makefile.PL.  This will set up the symlinks for you.

cd into RxPerl and apply the patch.  Do 'make perl'.  If Perl doesn't
build, let me know at once.

cd back to Rx-0.50 and run 
        make
If this fails, let me know at once.  Ignore the following warnings:
        Dump.c: In function `start':
        Dump.c:162: warning: assignment from incompatible pointer type
        Dump.c: In function `dump_compiled_regex':
        Dump.c:372: warning: passing arg 1 of `Perl_newSVpv' from incompatible pointer type

Now try the test program.  Run

        ./rxperl  -I RxPerl/lib ./test.pl

Other utilities to try: try.pl dump_regexes.pl bytecode.pl annotate.pl

----------------------------------------------------------------

How to interpret offset and length information:

The dumped regex structure is a hash.  Hash keys are node ID numbers;
each corresponds to a node in the original compiled regex.  If the
hash key is an ordinary number (like '3') then the node was present in
the original regex; if it contains letters also (like '3e' or '3n')
then it was added as part of the instrumentation process, somewhere
after the real node 3 and before whatever node followed the real node
3.

The offset and length information tells you what part of the original
regex each node corresponds to.  What part of the regex corresponded
to node 3?  Look at $regex->{OFFSETS}[3].  The is the offset into the
original regex string at which the node starts.  If it's a 17, it
means that node 3 corresponds to part of the regex starting at
character #17.  Where does the corresponding part of the regex end?
Look at $regex->{LENGTHS}[3].  If it's a 12, that means that the part
of the regex that corresponds to node 3 is 12 characters long.  Node 3
is derived from the partof the regex at characters 17-28, inclusive.

Example:  The original regex is /abc*defg[hijk]lmnop/

The contains the following nodes:

        Node  0: EXACT "ab"
        Node  2: STAR
        Node  3: EXACT "c"
        Node  5: EXACT "defg"
        Node  7: ANYOF "hijk"
        Node 18: EXACT "lmnop"
        Node 21: END

Here is the OFFSETS array:

          'OFFSETS' => [ 1, undef, 4, 3, undef, 5, undef, 9, undef,
                         undef, undef, undef, undef, undef, undef,
                         undef, undef, undef, 15, undef, undef, 20 ],


Element #7 has the value 9; this is because the part of the regex that
corresponds to the ANYOF node begins at character position 9.
That's the '[' character.

Here is the LENGTHS array:

          'LENGTHS' => [ 2, undef, 1, 1, undef, 4, undef, 6, undef,
                         undef, undef, undef, undef, undef, undef,
                         undef, undef, undef, 5, undef, undef, 0 ],


Element #7 has the value 6; that's because the part of the regex that
corresponds to the ANYOF node is 6 characters long.  ("[hijk]")

Element #8 of both arrays is undef.  That is because there is no node 8.

Element #21 of the offsets array is 20, even though the regex is only
19 characters long.  That's because node 21 is the END node.
Similarly, element #21 of the lengths array is 0.

If you don't like the arrays, it is easy for me to write a little bit
of Perl code that reorganizes them in whatever way you would prefer.
For exmaple, rather than two seprate arrays, you might like the offset
and length data to be incorporated directly into the node structures
themselves.  Let me know.