Does bsearch ever miss?

**MAtkins** · 02-16-2011

I've got 2 lists that I'm comparing (attached).
If I do a straight strcmp it shows a match which bsearch is not finding.

I've attached the code file. It's a bit complex. It's a derivitive of something that Salem wrote for me.

It's missing on
hornypotter.com and on
www.herbsbuy.com
It finds 8 out of 10 and I can't figure out for the life of me why it consistently misses on those 2.
Again, a straight strcmp shows a match in the x loop below.
the bsearch right below it misses on those 2.

The function in question is main. It is intended to be used as a dll.
The problem is in the code starting on line: 272.

You can see the problem by running it and seeing that it deletes 8 records from UTest.txt instead of the 10 that it should.

Any ideas as to what the problem might be?
Does bsearch provide inconsistent results?

The commandline run should be SLSSeoUrlFunc MTest.txt UTest.txt UUTest.txt

Code:

//Shows a match on the 2 URLs in question.
	for ( int i = 0 ; i < nUpdate ; i++ ) 
	{
		for ( int x = 0 ; x < nUnknown ; x++ ) 
		{
			printf("\n%s : %s : %d", dbUpdate[i].base_url, dUnknown[x].base_url, strcmp(dbUpdate[i].base_url, dUnknown[x].base_url));
		}	
			
	}
	//Misses on the 2 URLs in question.
	for ( int i = 0 ; i < nUpdate ; i++ ) 
	{
		db_st *f = bsearch(&dbUpdate[i], dUnknown, nUnknown, sizeof(*dUnknown), sortfn );
		if ( f ) 
		{
			strcpy((*f).base_url,"");
		}		
	}

**Adak** · 02-17-2011

When you're failing on a string search, the usual suspects include:

1) One string has a newline char on the end (fgets() puts one there), and the other doesn't (fgets doesn't put the newline on it, if there isn't room for one).

2) One string has an end of string char, the other "string" is just a bunch of letters, w/o an end of string char \0' at the end.

3) The data is goofed - one string somehow has an extra space two underline char's, that looked like one, or one string has a letter from another language, like u with two dots above it, (German), instead of just a u, a n with the tilde stripped off (Spanish), etc. One string has a space ' ' char added to it.

On the second Wednesday of each month, we set a random char to fail in binsearch -- lots of panic and fun to watch.

Yes, binsearch always works, but the list much be sorted according to ascii values, NOT lexicographically (like a dictionary). There are differences.

**MAtkins** · 02-17-2011

Well, I figured out that the problem was *not* bsearch (whew!).

Strange though, I don't know why this is:
It had to do with setting the base_url to "".
I guess it's one of those 'undefined behavior' situations.

Well, it's solved now . . .

Code:

for ( int i = 0 ; i < nUpdate ; i++ ) 
	{
		f = bsearch(&dbUpdate[i], dUnknown, nUnknown, sizeof(*dUnknown), sortfn );
		if ( f ) 
		{
			//strcpy((*f).base_url, "") //<- this caused the problem - I have no idea how.
			(*f).DomainFound = -1; //This was my work-around 
			//f = NULL; //This did NOT help the problem
		}		
	}

**nonoob** · 02-17-2011

I have found bsearch to go flakey in the case where the count of elements is zero. It blew up royal. I had to code a check for that instead of expecting the function to return NULL. Sorry I don't remember the compiler I was using at the time but it was late 90s.

**Bayint Naung** · 02-17-2011

Code:

			//strcpy((*f).base_url, "") //<- this caused the problem - I have no idea how.

is base field char pointer? Did you allocate memory?

Thread: Does bsearch ever miss?

Thread Tools

Search Thread

Display

Does bsearch ever miss?

Similar Threads

bsearch problem

Nostalgy

bsearch( ) and qsort( ) not agreeing

Problem Using bsearch

bsearch