Re: [Monotone-devel] net.venge.monotone.deregexp

[Zack Weinberg]

On 2/27/07, Nathaniel Smith <address@hidden> wrote:
> On Sun, Feb 25, 2007 at 01:32:36AM -0800, Zack Weinberg wrote:
> > Specifically, uri.cc no longer uses 'em, and a
> > spurious #include of <boost/regex.hpp> has been removed from
> > netsync.cc. This could probably be flushed to mainline all by itself,
> > but I didn't.
>
> Any reason why not?

I wanted another pair of eyes on it, especially now that there's a
conversion of packet.cc too (which code I am not entirely happy with).

> > * diff_patch.cc: for the show-containing-function feature.
>
> ...How can we possibly eliminate this?  It's necessary if we want to
> be consistent with diff(1), and I can't think of any other approach
> that would be nearly as convenient anyway... (i.e., trying to use
> globs here does not seem like a winning plan).

We *could* make this operation a lua hook (more of one than it is now,
anyway) and use lua's string.match thing.  That implements a very
basic regular expression, as far as I can tell.  It has some odd
syntax (% instead of \) but the big thing it seems to be missing is
alternation.

I'm not all that enthusiastic about this idea, but I think it's worth
considering.

> if we are keeping some regexp engine around, that
> might simplify some other things.  (Don't have to rewrite the globish
> matcher by hand, etc.)

I was thinking about grabbing the fnmatch() implementation from gnulib for that.
[ http://www.gnu.org/software/gnulib/MODULES.html#module=fnmatch ]

Weren't you worried about unbounded recursion with pcre?

zw