Re: A regex question

From: Daryle Walker (darylew_at_[hidden])
Date: 2001-07-20 17:43:45

on 7/19/01 7:58 AM, John Maddock at John_Maddock_at_[hidden] wrote:

>> Let's assume that I have several regular expression patterns, p1, p2, ... pN.
>> I have an istream & (or a streambuf &.)
>> 
>> I want to determine which of the patterns, if any, matches the longest
>> sequence of characters at the current stream position (if there are several
>> that match the same number of characters I want the first), and then leave
>> the stream in the state it was before (if possible), that is, I either set
>> its position back if the operation is supported or call unget() repeatedly
>> the appropriate number of times (I'd have to know how many characters the
>> regex library has extracted.)
>> 
>> How can I do that using the boost.regex library? I don't see a simple way to
>> accomplish this.
> 
> No because the library is not stream based, and doesn't work with
> input-iterators (since then it couldn't report what marched as a pair of
> iterators), a more serious problem though is how long is a piece of string?
> Depending on how many characters are extracted, you may or may not be able to
> successfully put characters back (unless I'm misunderstanding how the iostream
> lib works).  Probably what you would need to do is create a kind of "caching
> iterator" - something that's constructed from an istream, and which carries
> out the seek (or putback) when it gets destructed, something like:
> {
>   boost::stream_container sc(mystream);
>   boost::regex_match(sc.begin(), sc.end(), my_regex);
> } // stream position reset here...

I think you would need something like a "caching iterator" instead.  As you
said, there's no guarantee of how many characters can be pushed back, if
any.  Even your reset idea may not work.  Maybe a wrapper iostream that uses
something like a string for an extra buffer could be used.  (Hmm, another
new class for more_io.zip.)

-- 
Daryle Walker
Mac, Internet, and Video Game Junkie
darylew AT mac DOT com