david fries wrote:
My my concern was how you would perform random access in a functional parser. You're points are interesting too. I guess if we really had wanted to work with parsed objects, retaining the shared references would have been a must.
Out of curiosity, was there *really* a need for random access? It sounds like you could've written things to do it in one forward pass, keeping a table on the side of the targets for dangling pointers, and then back-patching once you've reached the object being pointed at. You may even be able to use laziness to do the back-patching instead of needing STRefs. The only problem I could see with that approach is if there was funny bit-packing going on so that objects were overlapped in the same memory region. -- Live well, ~wren