This morning Bill suggested that create_doc.c should use a parser generator like peg/leg. Since I had criticized the code somewhat harshly before, I felt I should undertake the trivial exercise of switching it over to an automatically generated parser.
Long story short, while this really wasn't hard, it was also not completely straightforward (just a bunch of "soft" difficulties related to writing a working grammar and stuff like that). I spent much more time on this than I probably should have, but there is a working version in my gsoc-doc branch now.
After that I thought about matrix temporary allocation jumbles. I think Bill's solution (allocating temporaries of sufficient size to fit every intermediate result, and using cheap resizing before each assignment) will work. I started the fmpz_matxx class, but haven't really gotten anything done yet. But everything is ready to go tomorrow.
No comments:
Post a Comment