Re: [Boost-users] boost::xpressive - Regex stack space exhausted

On 7/22/2010 10:01 AM, Pavol Supa wrote:

On Thu, Jul 22, 2010 at 2:33 PM, Eric Niebler wrote:

...

On 7/22/2010 5:29 AM, Pavol Supa wrote:

...

On Wed, Jul 21, 2010 at 10:52 PM, Eric Niebler wrote:

...

On 7/21/2010 2:39 PM, Pavol Supa wrote:

...

Hi,

I need to match a hex-written byte array, optionally separated with spaces. So i tried:

boost::xpressive::sregex r = * ( * blank >> repeat<2,2> (xdigit));

Wait a minute. What does the data you're trying to match look like? Do you know that xdigit only matches a single hex character? Your data should look something like:

00 11 22 33 44 55 66 77 88 99 aa bb cc dd ee ff 00 11 ....

Is that right?

i used data just without spaces, 000000000000...00000

the data are most often without spaces, so i decided meanwhile to use the workaround (it doesn't throw): expr = * repeat<2,2> (xdigit)

Then why not just "expr = *xdigit" in that case?

but if i move the inner expression to a function, it starts throwing (at similar length), i.e.: sregex byte() { return repeat<2,2> (xdigit) } sregex expr = * byte();

Yes, these are different beasts altogether. A nested regex match creates a nested match_results object, with nested submatches. By assigning to a regex object, you have erased the type information that xpressive uses to optimize matching. In this case, xpressive no longer knows that it's quantifying something of fixed-width so that backtracking is unnecessary.

...

...

...

...

smatch match; regex_match (input, match, r);

when i use input of approx. 150 hex pairs, i get an exception "Regex stack space exhausted" (i use default stack size by Visual studio 2008)

This pattern looks quite simple, so I'd like to know, if there is some fundamental problem with this expressions.

Yes. See http://www.boost.org/doc/libs/1_43_0/doc/html/xpressive/user_s_guide.html#bo...

Not only will this pattern tear through stack, it'll run very slowly. Try this instead:

* ( keep(*blank) >> repeat<2,2> (xdigit))

So, i tried it. It throws exceptions at ~230 hexdigit pairs.

That doesn't sound right. This could be a bug. I'll look at this today.

i tried to debug the parsing, but it looked ok - when there is a non-trivial expression inside (as mentioned above), it goes into recursion, about 12 functions on callstack, and it eats whole (default) stack at those mentioned quantities

See above. I hope it makes sense now. But still, *(keep(*blank)>>repeat<2,2>(xdigit)) should work just fine. I'll look. -- Eric Niebler BoostPro Computing http://www.boostpro.com