C / C++ / MFC

Re: regular expression problems

3-Oct-02 14:17

If you want to match a string with an arbitrary number of letters and commas, you will not succeed I'm afraid.

What you want is a context-free grammar for your input "language". I'm not sure if you are familiar with BNF (Backis-Naur Form), but here's such a grammar notation:

<string> ::= <string> <letter_comma_group>
         | (empty)
<letter_comma_group> ::= ['a'-'z'] ','

Regular expressions can only handle languages which follow something called regular grammars. Regular grammars cannot handle self reference - thus you will not be able to do what you want.

Please see Chomskys grammar hierarchy[^] for more information on this topic.

If you want to solve your problem, I suggest you look into writing a recursive descent parser http://www.wikipedia.org/wiki/Recursive_descent_parser[^]. But before you start implementing the grammar I mentioned above, please make sure you rewrite it to be LL(n) compatible. (Left-recursion in an LL(n) grammar is a big no-no! Try it and you'll find out why. Smile | :)

<string> ::= <letter_comma_group> <opt_string>
         |   (empty)
<opt_string> ::= <letter_comma_group> <opt_string>
             | (empty)
<letter_comma_group> ::= ['a'-'z'] ','

Good luck!

--
Please state the nature of your medical emergency.

Jon Hulatt3-Oct-02 21:52

Jon Hulatt

3-Oct-02 21:52

There is no easy solution. How i've solved this in the past is to replace the matched pattern with something else, and loop, like this (perl) :-

$p="a,b,c,d,e,f,g,";
$matchcount=0;

while ($p=~s/([a-z],)//)
	print ++$matchcount.": ".$1."\n";

Note that I lose the * from your pattern; you only want to match 1 at a time. =~ is the pattern operator in perl; // are delimeters around a pattern; s/// substitutes the first pattern with the second, ie nothing in this case. $1 is a special var which contains the first stored match.

Incidentally though, if your pattern really is as simple as a comma separated list, you may as well use a Split type function (theres bound to be one in your regexp library. In perl again:-

$p="a,b,c,d,e,f,g,";
@results = split (/,/ ,$p);

foreach $result (@results)
{
  print $result."\n";
}

Hope that helps.

Signature space for rent. Apply by email to....

*.bmp *.wbmp conversion

3-Oct-02 12:27

Hi,

have you ever come across converting a *.wbmp (wireless bitmap) into normal bitmap or *.jpg and vice versa?
Any answer would be highly appreciated!
Thanks

R.

Re: *.bmp *.wbmp conversion

Jörgen Sigvardsson3-Oct-02 13:35

Re: *.bmp *.wbmp conversion

3-Oct-02 13:35

Exactly what difference is there between a wireless bitmap and a normal bitmap? Please excuse my ignorance.

--
Please state the nature of your medical emergency.

Anonymous3-Oct-02 13:59

Anonymous

3-Oct-02 13:59

A wireless bitmap is, like the small brother of a normal bitmap, small, primitive, monochrom, and on top of all it is of different structure than *.bmp because no MS tool (including IE or Paint) is able to show it.

It is used on wireless devices (handy). WAP pages are written in so called WML (Wireless Markup Laguage) and they can only have references to *.wbmp pictures.
I would want to show these pics on IE if I knew how to convert them to human readable format.

R.

Re: *.bmp *.wbmp conversion

C^hris L^osinger3-Oct-02 18:00

C^hris L^osinger

3-Oct-02 18:00

this will do it: http://www.smalleranimals.com/thumb.htm

there's code on the codeproject site, too (i think), if you need that.

-c

bit, n:
A unit of measure applied to color. Twenty-four-bit color refers to expensive $3 color as opposed to the cheaper 25 cent, or two-bit, color that use to be available a few years ago.

Etch-a-sketch!

This 'works' but would you ever want to ??

RedZenBird3-Oct-02 12:21

3-Oct-02 12:21

The question came up today in our shop "can I call one constructor from another just like I can in java" and 'out of curiousity' tried to figure out if it can be done...After a few attempts (at what I thought were direct 'and sensible' approaches that only made the compiler whine) I came up with the following....Now, don't get me wrong, I'm not advocating that anyone *should* do this....only that it appears that you can and that it looks like it 'works'.... Big Grin | :-D

#include "stdafx.h"
#include "stdlib.h"

int flag;

class CtorTest
{
public:
CtorTest( void )
{
int x = 1;
}

CtorTest( int iFoo )
{
new( this ) CtorTest;
int x = 2;;
}

~CtorTest( void )
{
int x = 1;
}

void operator delete(void* p, CtorTest* poOuter)
{
free( poOuter );
}

void* operator new(size_t s, CtorTest* poOuter )
{
return poOuter;
}

};

int main(int argc, char* argv[])
{
printf("Hello World!\n");

int aNum = 5;
CtorTest x( aNum );

return 0;
}

Just trying to keep the forces of entropy at bay

RedZenBird3-Oct-02 12:49

3-Oct-02 12:49

okay, thought a little explanation is due here as well.....in main() we get a CtorTest object and the CtorTest::CtorTest(int) constructor gets called. Then we jump into the placement new function that effectively returns 'this'....Then from that point, we jump into the CtorTest::CtorTest(void) function ( invoked right after the return from the local placement new function. After that, control returns to the 'first constructor' call and we return to main. To support the assertion that there is only one CtorTest object created (?), the destructor is only called once......strange...very strange...

Just trying to keep the forces of entropy at bay

Jörgen Sigvardsson3-Oct-02 13:47

3-Oct-02 13:47

Sure, why not? It's perfectly legal to do replacement new like that. However, I'm sure it would be frowned upon in a programming team. Wink | ;)

One way to do it with a more java-like syntax would be:

template <typename T>
class java_lookalike_helper {
protected:
    typedef T this_t;
};
 
#define this_ new (this) this_t
 
class Class : java_lookalike_helper<Class>  {
    int x, int y;
 
public:
    Class(int x_) : x(x_) {
    }
 
    Class(int x_, int y_) : y(_y) {
        this_(x_);  // Check out the spiffy java like syntax!
    }
};

--
Please state the nature of your medical emergency.

RedZenBird3-Oct-02 14:46

3-Oct-02 14:46

thanx for that variant of this idea. I'll add that to the 'tool kit' and play around with that as well.....this arcane-chicanery must be what causes some advocates to lean toward the 'managed code' camp...where we likely can't have this kind of 'fun' with the language Big Grin | :-D

;)

Just trying to keep the forces of entropy at bay

Jörgen Sigvardsson3-Oct-02 15:06

3-Oct-02 15:06

Hey, exploring seamingly nutty ideas may indeed evolve into something really cool. Just look at ATL and WTL - now there are some wild and crazy ideas that have evolved into something cool and useful Smile | :)

And the advocates can bite me and my gotos. Wink | ;)

--
Please state the nature of your medical emergency.

RedZenBird3-Oct-02 15:16