"Eric Niebler" <eric@boost-consulting.com> wrote in message news:4924898A.4020101@boost-consulting.com...

Michael Goldshteyn wrote:

...

What is the best way to test for an empty boost::xpressive::regex object?

That is, if I have code like:

boost::xpressive::sregex re;

...

// Test if re is empty and should not be used to perform a regex_search if (?????)

There is, although it's not obvious from the documentation that it's possible. You can see from the postcondition on the default constructor of basic_regex that this->regex_id() == 0 for default-constructed regex objects and != 0 for non-default constructed regex objects. So your test above can be:

if( 0 != re.regex_id() )

HTH,

Thanks for the speedy reply. I did notice the postcondition on the constructor and the fact that the regex_id() will return 0 for a non-initialized regex, but was hoping that there was a more intuitive way to go about this test, since readers of the code may get confused without a comment. How hard would it be to add an empty() function with a signature that is similar to the one in boost::regex before Boost 1.38.0 comes out, since this functionality, at least in my humble opinion is very useful. The signature for the function in boost::regex is: bool empty() const; Thanks, Michael Goldshteyn

"Eric Niebler" <eric@boost-consulting.com> wrote in message news:4925187A.6080402@boost-consulting.com...

Michael Goldshteyn wrote:

...

"Eric Niebler" <eric@boost-consulting.com> wrote in message

...

if( 0 != re.regex_id() )

Thanks for the speedy reply. I did notice the postcondition on the constructor and the fact that the regex_id() will return 0 for a non-initialized regex, but was hoping that there was a more intuitive way to go about this test, since readers of the code may get confused without a comment. How hard would it be to add an empty() function with a signature that is similar to the one in boost::regex before Boost 1.38.0 comes out, since this functionality, at least in my humble opinion is very useful.

The signature for the function in boost::regex is:

bool empty() const;

boost::regex goes to some length to present itself as a souped up container of characters. You can assign a character range to it, get begin() and end() iterators for stepping through the characters with which the regex was initialized, etc. Given that, see if you can answer this without checking the docs:

assert(std::string().empty()); // OK assert(std::string("").empty()); // OK assert(boost::regex().empty()); // Is this true??? assert(boost::regex("").empty()); // How about this???

Yes, the behavior of the last assert is surprising, so I do see your point. "empty" is certainly not the right name for such a function, at least in the context of standard container functionality.

Unsurprisingly, regex::empty() is not part of the standard regex interface in C++0x. I don't like regex::empty() and I'm not inclined to add it. Sorry. You already have a way to get the information you're interested in. If you would like to give it a pretty name, by all means...

template<class Iter> bool is_invalid(xpressive::basic_regex<Iter> const &rex) { return 0 == rex.regex_id(); }

empty() may not a good name for this function, but agree with me that encapsulating the functionality of testing whether an re object actually holds a regular-expression should be more intuitive than: 0 == rex.regex_id(); // Test if rex actually contains a regular expression Perhaps something like unfilled(), bare(), or as you suggested is_invalid() should be added to the actual implementation, instead of user code? I would further argue that something so trivial and tightly coupled to the implementation (i.e., regex_id() being zero) should be a member function and not stand alone. Opinions on this topic from others would also be greatly appreciated! Thanks, Michael Goldshteyn

Eric Niebler wrote:

Michael Goldshteyn wrote:

...

Perhaps something like unfilled(), bare(), or as you suggested is_invalid() should be added to the actual implementation, instead of user code? I would further argue that something so trivial and tightly coupled to the implementation (i.e., regex_id() being zero) should be a member function and not stand alone.

What is it about checking the regex_id() that bothers you? Is it the constant 0? You could instead write it like ...

if(rex.regex_id() == sregex().regex_id())

Default-constructing a regex is very cheap, FWIW.

I'm not very familiar with xpressive, but having a look from aside it would never come into my mind that "rex.regex_id() == 0" is an emptyness (or whatever) check. Sorry, this looks more like a hack to me. I think a clear statement is needed: is a default-constructed regex valid? If not, then why does it have a default constructor? If it is, the default-constructed state should be easily detectable, since this is a common practice in just about any domain: - Boost.Regex, containers, strings, Boost.Function provide empty() - Boost.Optional, Boost.SmartPtr provide unspecified bool conversions and operator!, or equivalent facilities. Personally, I like empty() naming, but if a more precise name can be found, I would be fine. And I agree with Michael, such a function should be a member of the class. As for application, the empty state of regex may be useful when the expression may, or may not be compiled. It's a sort of implicit optional, but a simpler and more efficient one. However, I'm sure it's not the only valid use case.

I have submitted Ticket #2519 to more formally report this matter and its suggested resolution. https://svn.boost.org/trac/boost/ticket/2519 Thanks, Michael Goldshteyn