Hi All,

I've just completed the merging of the New Parser branch work with the
HEAD branch.  I did this manually as it was easier for me to do, but
CVS history of the changes I did remain in the new parser branch only.

Some notes:

-To compile the new parser you'll need Flex and Bison.  By default,
SConstruct will not build the new parser unless "buildNewParser=1" is
given.

-To use the new parser at runtime, you'll have to pass the
"--new-parser" flag to Csound, otherwise by default the old parser is
used.

-A test suite I was using of CSD files and a python test harness
script have been checked into CVS as the tests directory.  The test
script defaults to outputting to stdout but I did a write a (not very
good) TK UI also that can optionally show up after running by giving
the "--show-ui" flag to the test.py script.

-I was adding notes on the new parser to the To-fix-and-do file in the
new parser branch and I copied that over to HEAD.  I also tried to
comment the new code as I was working.

-I did do a check to make things C89 compliant, but I haven't looked
at the code in the past month and a half so if there's any
non-compliance, please feel free to correct and checkin.

-The current missing language features I know of are:

  if-then's
  UDO's
  some opcode polymorphism cases
  macro's
  some expression operators

-The optimization run needs to be implemented(stub is in); certain
code is not as efficient and current parser (particularly not
generating unnecessary temp variables)

-The grammar reports ambiguities; last I remember when working, I
thought that the grammar and lexer may need to be worked more tightly
together and become context-dependent instead of context-free.  I
don't know what's the way to go exactly.

-The file Engine/symbtab.c may need to be reworked on how it builds
the symbol table.

-Csound will fail miserably on invalid input; I have not touched
failure at all, and the parser doesn't report anything to Csound at
the moment, so if invalid input is given, csound just keeps going
along in a probably uninitalized state.

Last I checked, using the --new-parser I was able to render "Trapped
in Convert" and also "Xanadu".

When building with the new parser, the following files are generated:

Engine/csound_orclex.c
Engine/csound_orcparse.c
Engine/csound_orcparse.h

PLEASE do not check them in!  (Add them to your individual .cvsignore file).

(I've attached copies here for Anthony to try out at his request.)

To reiterate, the new parser does not compile by default and is not
used at runtime by default, so for the lay person who may be compiling
and using Csound, things should just run as it used to  without any
problems.

If there's any questions, please feel free to ask away here.  It'd be
great to get as many eyes on the code and help in implementing as
possible.  I don't feel like this is "my" code or anything so if you
want to change it please go ahead, but just let everyone here know
you're working on it so we can all coordinate to some degree.

Thanks!
steven