I've managed to avoid blowing up my brain and created a noweb aspect for the latex listings package, with accompanying lyx module.
As a result I don't need to use noweave, but I get noweave looking output with the listings generated by the listings package with all it's pretty printing. The lyx module (called "newfangle") treats each \lstlisting as a noweb style chunk, and uses the listing name as the chunk name. It uses the noweb package macros merely to generate the chunk header and navigation arrows to the previous and next chunk. This avoids the need to use noweave for most uses, and removes for me the need to use the literate lyx file format. I'm still lacking is noweb markup of <<includes>>. I need to get lstlistings to spot these... Because Lyx doesn't show the listing name visually while editing it can be very confusing to remember which listing has which name, so I've added a Chunk style which sets the listing name. This will be persistent until the next Chunk or the next explicitly named listing. I prefix most of my listings with a highly visible chunk header. The Chunk style pretty much just prefixes "name=" to it's text, and then calls \lstset, but it does some magic to preserve the chunk name into the listing. I also have a newfangle awk script which works like notangle, but nicely recognizes the latex that listings uses to pick out the code. The awk script has less bugs that notangle, can manage -L output without breaking the indents, and manage to suck C code into multi-line macros without breaking. I can now do literate programming with much more simplicity than before. I hope that Lyx will be able to accept this work, and make literate programming work more naturally. I've been "are we nearly there yet" for the last 6 weeks, and I think my answer is "yes" (the 3rd time around). I'll post the code tomorrow, I need to run, but I didn't want anyone working on my latex macro question now I solved it. Sam
