Re: [DotGNU] dfa2csharp end result

[Adam Ballai]

Gopal V wrote:

> If memory serves me right, Gopal V wrote:
>  
>
> > Does generate working code , though beauty wise .. I could do far
> > better ... And the command line parts segfault when given no params
> > etc... it's a 3:15 AM status snapshot ... be gentle :-)
> >    
>
> I've cleaned it up a bit ... added a couple more of new transition
> modes ... (WHITESPACE, EPSILON) ... cleaned up the code printing routines.
Great!

> Better collision detection -- for the whole unicode range , rather than
> for the 256 byte range ..
>
> Fixed up inlining of single connected nodes ... ie if a node is connected
> from only one point , do not 'goto' but use it there itself , reduced number
> of jumps ... If there  is a loop back into the inlineable node, ignore
> inlining ..
>
> Unconnected nodes are not generated , and neither are their action and
> transition code blocks ...
Thats a tricky thing for xml documents.  Maybe some error reporting on 
bad xml syntax could be generated in this case.

> --debug option which  generates Debug.WriteLine() of the various operations
> taking place ... (will need -lSystem -DTRACE to compile C# o/p ) 
>
> Line numbers are generated for the %{ %} inline code fragments using the
> original file , while those of the other code are generated with respect
> to the output file .. (#line)
>
> Pulled out most of command line handling from pnet , the configure 
> script off treecc/pnet and made it an independent project (since
> Rhys didn't like the idea of this moving into pnet) ..
Well, as rhys has said before just do it, so I'll get something working 
to parse Xml with the tool.

> A Rot13 convertor is provided as a nice example for the stuff ...
> if anybody is interested, I'll write out the language docs, otherwise
> this is slowly going to fade out :-)
>  
I hope this doesn't, and I'm sure this is useful to any developer just 
as yacc or flex is.

-adam